The definitive Guide to AdWords remarketing for e-commerce
A note for the reader
This is an extensive guide for remarketing. It contains information for the new marketer that wants to enable his first campaigns and the advanced marketer that wants to level up his game.
If you are a relatively new marketer, I would advise to follow the content in the way it is ordered. I have left the advanced topics for each chapter at the end, and you can freely skip it the first time you implement.
If you are an advanced marketer, you could go directly to some interesting topics such as
What is remarketing
In this chapter
What is remarketing
Remarketing (also known as retargeting) is the campaign type that we can use to target people that have already visited or taken action on our website. This is extremely valuable because people that have already visited our website have a significantly higher probability of conversion. This conversion probability can be further boosted by our ads that can be based on the exact actions that the visitor did.
Remarketing ads are displayed as banners on various websites across the web. In regards to Google remarketing, all these websites are part of Google Display Network. Up to now, we have said that with remarketing we can show ads to people that have visited our store and place those ads on 3rd websites across the web.
As I am a believer of “show, do not tell”, some examples of remarketing that you have witnessed as a user are:
- Many ads from a specific store after you have first visited it
- Ads with products that you have either browsed on a store’s website or that are closely related
- Ads with a specific promotion or coupon code after you have added something to cart but have not yet completed the purchase
What is not display remarketing
I already explained what is remarketing but I want to be sure that we are talking about the same thing. A good way to do this is to also tell you what is not (display) remarketing. So please find below some relevant terms and what they really are. For sure they are not what we are talking about in this guide.
- Remarketing lists for Search Ads (RLSA): RLSA is the concept of remarketing applied on Google Search. By using RLSA you can target, bid and use different ad copy for people that have visited your website.
- Email retargeting: Email campaigns that are sent automatically to email subscribers that perform a specific action on your website/apps. The most common example is to send an email to all cart abandoners reminding that they have still something on their cart (or give an incentive for them to convert).
- Similar audiences: Similar audiences are automatically generated audiences of people that have not visited your website generated by Google. They are based on a specific list that you have created (eg Similar audience of “converters (30)”). Although frequently positioned as a way to expand your retargeting campaigns, they are not remarketing.
- Prospecting: By defining prospecting, we clearly see what remarketing is. In short display advertising is broken down into remarketing (people that we “know” in some form) and prospecting (people that we do not know).
- Customer match: This is the late answer of Google to Facebook’s Custom audiences. We are now able to upload data from our CRM (eg past customers) to target people on Google’s networks.
Non-AdWords display remarketing
The focus of this guide is AdWords remarketing for e-commerce. But Google is not the only company offering a remarketing solution for eshops. Some other popular solutions are:
- Facebook: As in all display ad solutions, Facebook has a similar offering to Google’s remarketing solution and its own valuable inventory. In our days, there is no point in discussing what is better; we need to use both.
- Criteo: Criteo is an ad-tech company with a focus on dynamic retargeting (it is the equivalent of dynamic remarketing by Google that you can read about below). My personal opinion is that given that close to 80% of its inventory is Google/Facebook and it makes a 40% margin on top, it does not really make any financial sense to use.
- AdRoll: Another retargeting company focused more on the SMB space. I have the same opinion as for Criteo.
Before you start
In this chapter
How to setup the cookie
In order for remarketing to work, we have to change the tagging of our website so that Google Ads can collect audience data from our website visitors. We will then be able to create lists upon this pool of data inside the Google Ads UI such as “All website visitors” or “All converters”, to name the most common ones. There are two ways to do this use the Google Ads tag or take advantage of our Google Analytics setup.
We can use either Google Ads or Google Analytics for the remarketing setup and it will work fine with both. But by using Google Analytics we have two main benefits, namely easier/faster tagging and opportunity for more advanced ad targeting. What I mean with “easier/faster tagging” is that you do not need to add an extra tag for Google Ads, you can use the one you have from GA. One less tag, one less thing to worry about. But is not only the tagging benefit but also the more advanced targeting.
In short, with the Google Ads tagging, you can create lists based mainly on what pages people visited. But with GA, oh GA…, you can do magic. You can create any advanced segment that you want within Google Analytics and then make this a remarketing list. You can create a list for morning visitors or for people that have clicked on X banner 1 month ago or for people that bought something from Y category. You wantthis, you need this. Go that way!
To my mind the only reasons to go with the Google Ads tagging are :
- You do not have Google Analytics on your store
- Different team managing GA in your company and you have too much company politics (ouch…)
If you have to go this way, then check for the new Google Ads tag.
How to setup GA for remarketing
You can do this easily by following the Google documentation. In short, you have to edit your code and enable the feature on the GA interface. It goes without saying that in order to use this, you need to have your Ads account linked to Google Analytics. There are multiple benefits in doing this, so if you do not have this, just do it.
More for dynamic remarketing
As you will see in the structure part below, an important part of remarketing for e-commerce is dynamic remarketing. To be able to use dynamic remarketing you need some more things to setup in your GA.
What you need to take into account when you start advertising with remarketing campaigns is that any potential audience that you want to target should contain at least 100 users within the last 30 days. The reason Google has set this rule has to do with privacy because it would be kind of creepy to be able to target 1 specific user with a specific message (think “John Doe, why have you left products in the cart?”).
As I have mentioned the 100 users minimum applies to each list you advertise and you have to think about it every time you want to create more lists or make your lists more granular. For example, if you have a low-traffic website, it does not make sense to create lists for all your product category pages because most of these might not have enough traffic.
What to do if I do not yet have the audience to remarket?
Use other customer acquisition tactics in order to build your website traffic so that remarketing can firstly start working and secondly has a substantial volume. Targeted search campaigns, Google Shopping campaigns or Facebook Prospecting ads can be the solution on all this.
Product Feed setup
If you are planning to run Product ads (Google dynamic remarketing), you also need to create a product feed so that Google can match your products with your product pages and also have product information display on your ads.
There are 2 ways to set up your feed in order to use it for dynamic remarketing. The first and most popular way is to create a Google Merchant Center account and upload your product feed based on the specifications provided by Google.
The second way is to upload your feed on the “Business Data” section of your Google Ads accounts and then assign this feed into the campaign that you want to run (see image). This way is usually used by non-retailer accounts, but can also accommodate retail, so it is relevant to all people reading this guide.
If you plan or already run Google Shopping campaigns, then go the Merchant Center way. If not, then the 2nd technique is easier and faster to implement.
Campaign & Audience Structure
In this chapter
You have now set up the technical side of things so you can enable/scale remarketing for your web store. How you go about this? On a high-level you have to do 2 things: identify the audiences that you will run with your ads and set/create your ads to run. In this chapter, we are going to focus on the first part; the audiences.
I am going to start by explaining some key concepts and go through an example of how all this can work. Then we are going to dive deep into each segment and I am going to close this chapter by giving you the exact cheat sheet on how to structure your campaigns to better target your audiences and what are advanced techniques to use so that you can get even better results.
The idea here is one very familiar to marketers, the one about funnels. So given that we are discussing about remarketing, the start (or the top as they say) of the funnel is someone landing on our website. At that point, our remarketing cookie has been placed and this person is now included on one or more of our remarketing audiences.
The end (or bottom) of the funnel is the completion of the purchase. But this is not actually the end because we can then have (and should try to have) a re-purchase. Repeat customers are really important because of unit economics and lifetime value. You already know that is really expensive to get one new customer online. No matter what the acquisition channel (PPC, SEO, social, PR, etc) these channels cost and the cost per new customer is high. In fact, in most industries, the 1st order is not profitable if you include the customer acquisition costs. So what you care about then is LTV or lifetime value. And the easier way to increase LTV is to get people to buy more from you. This is why repeat customers are important and this is why it does not make sense to limit your funnel to the purchase but add one more step for the re-purchase.
in most industries, the 1st order is not profitable if you include the customer acquisition costs.
Now we have set the ground that we start from the 1st visit and we stop on the repurchase. This is our funnel and our funnel has different steps, different phases that we can use to segment our future prospectivecustomers. A really easy, and hopefully familiar way for you to understand this visually is to look at the various steps of the shopping funnel on the enhanced e-commerce report of Google Analytics (if you are using GA and do not have enhanced e-commerce setup, you are losing a ton. Go fix it!). There you can see the steps as; Visitor -> category page -> product views -> cart -> checkout -> conversion. And this is really a great break-down if you just add the repurchase we have what we need to build our funnel.
Not surprisingly this type of funnel is not only a remarketing strategy but it is the foundation of many growth hacking tactics as it is actually the ARRR framework for growth, as described by the founder of 500 startups, Dave McClure. This is a tried and proven way to think about your audience.
Why is thinking about my audience in the right way important? Because people at different stages of the conversion funnel need to be shown different messages in order to help them come back to the websiteand proceed further into the conversion funnel. The classic anti-example of this is marketers setting up dynamic retargeting (ads with products shown based on your browsing behavior) that target “All visitors” and ultimately show ads for products that people have already bought from the website. Not good as an experience and, of course, not optimal in terms of advertising effectiveness. Given that we have understood the segmentation, let’s do one example
A user sees an ad for our website that sells socks online at a bus stop and searches on google for ubercoolsocks.com and lands on our homepage (a cookie is now set and this user is part of our “all visitors” or our “homepage visitor” audiences). We show ads showcasing the main benefit of our website, which is that we breed our socks to be happy so they can make their owners happy in a beautifully designed banner ad.
The user sees the ad, remembers our website and comes back and this time starts browsing the different categories (the user is now part of “Category: sports”, “Category:boxing” audiences) and then leaves the website. The next day the user sees an ad about “Boxing socks that rock!” and comes back to our website directly at the boxing category where he goes deeper in the product search and visits a few product pages with boxing shoes (and now gets into the “product pages” audience that will fire dynamic retargeting).
A few days later, we run a promotion and the user sees an ad with the sock he has been browsing and see the discount in the ad and comes back to the website and adds 5 pairs on his cart. He goes on to start the checkout process but then Maria talks to him on Facebook and leaves the process as is. One hour later gets one more ad that drives him back to the website and starts and completes the checkout process.
Some days later the socks arrive at his house and now the user is an uber cool guy wearing the happy socks. And then guess what… He sees an ad with some socks that complement the ones he bought before and comes and buys again.
This is a long example of how all these fit together. It goes without saying that we would like our user to buy these socks without clicking (and us paying) for all these ad clicks. And most of the times this is the case. Hopefully ;)
Break AdGroups by
Audience time size
Who belongs to this audience
As the name clearly implies, users that visit the store homepage get into this category. As also you can clearly understand, this is the top of the funnel so the conversion probability is lower. And this is why the bids will also be lower.
How you can set this up
Go to Google Analytics and create an advanced segment with:
- Page exactly matches /
If you think about all the different steps in the conversion funnel of a user buying from an e-commerce store, you will notice that the aforementioned audiences do not include one important one; the category pages. Category pages are the pages that list various products and usually have filtering capabilities on the sidebar. Users browse these pages in order to find the products that are interesting to them so they can click in and learn more about these products.
If you are not already convinced that these pages are important, I can also show you the data! If you have implemented “Enhanced e-commerce” on your Google Analytics, you can go to the Shopping behavior report and get a data hint on the power of category pages. As you can see below in the section “sessions with Add to carts”, there are many of these where people did not pass through the previous step, which is the product views. But how on earth can people add something to the cart while not seeing it? It can happen because in most stores people can add something to cart from the product page. And depending on the nature of your business this number can be significant or very high (eg in online grocery retailers this is more than 50%!).
So now you know, category pages are important. How can we include them into our targeting? There are 2 ways that differ in ease of implementation and potential impact. The easy way is to make one audience named “viewed category pages” that includes all users that viewed such a page but do not make a distinction as to what page. The easiest way to do this is to add a GA page type tag on all your category pages (from your e-commerce software) and then create a Google Analytics audience based on this. Now you have this audience. The benefits of this audience are that you can bid differently on this audience and also get the segmented reports so you can further improve your bidding. The bad thing, however, is that you cannot make a targeted ad for this audience as you do not know what the users were looking for, you can just use generic store ads. So let me also share the other way.
The 2nd way to create category page audiences is to build different audiences for the different categories. You can have for example “Viewed category: laptops” or “Viewed category: laptops - 15”. As you saw in my example, you can make your decision at what level of categories do you want to target. The big benefit of this method is that given that you know what they are interested in, you can use targeted ads, eg by showing a “15” laptops on offer” ad and these ads have better CTR and thus lower CPC, ultimately leading to more efficient conversions. The bad thing (why should there always be a bad thing?) is that you need to put more time into building this. You have to create different audiences manually, create multiple AdGroups and also create different ad creatives. My opinion is that the benefit far outweighs hurdle in most of the accounts.
If your store is small, it may happen that a category page has so low traffic that you cannot actually retarget. Check your data
My closing recommendation is to use the 1st method (so make a generic categories audience) and then build individual targetings for your top-10 categories. If you have enough traffic and you see that the performance on the more targeted ones is really good, you can start going deeper when you have some spare time.
You have to define what “category” means to you, as there are many levels of categories in most web stores. The issue with going at the 1st level is there are a few categories and most probably there are really different products in each of these so you cannot make a fully relevant ad. The issue at going really low (eg <3rd) is that there are too many categories and you can either hit GA limits (2.000 audiences) or make the whole campaign really difficult to manage. Ideally, go for the 2nd level category.
You always have to think of the cost-vs-value comparison for any further optimization/break-down. If by going more granular the impact is really small and it takes a lot of effort (and all effort can be measured into $) then do not do it.
Who belongs to this audience
Any user that visits at least one product page will get into this category. Getting to a product page means that the users are further down the conversion funnel and want to see in more detail a specific product (eg zoom in the images or read the product specs) so he can make a decision.
How you can set this up
No, you do not need to make different audiences for each product, as practically this could not happen.
If your are using GA for tracking, you can use an option to set a special type of page and then we use this to create the advanced segment. If you are using Google Ads tracking, you can set up the appropriate tagging and then start using this information on your ads.
Shopping cart abandonment
Who belongs to this audience
Ah… the cart! The most used remarketing list of all and for good reason. Having a user add something on their cart and then not buying is not good. Imagine what would this mean in the offline world. A customer walks into the supermarket, strolls through the different sections, add products in the trolley and when they reach the cashier, leave the trolley and go. If you were the store manager you could easily grab a lasso to get them back in. In the online world, you can do this via remarketing.
How you can set this up
There are 2 ways to get this audience. They differ in sophistication and in audience size. The easy way is to just get all people that visited the cart page. You can go into GA and create a page segment where the value exactly matches /.
The other way, that sits on top of the previous is to also get all people that have added something to cart but have not visited the cart page. In order to do this, you have to set a special tracking on GA so that you can catch the “event” of the successful add-to-cart. After you get this, then you can go into GA and create an audience that fired this event. By the way, you can go even more sophisticated by catching the value of the products that added to cart.
Many times someone clicks on the add to cart button but this is not successful, eg because the size is not available (in fashion stores). Be careful, you do not want these.
They say that it is 3 (or 10, or whatever each study states) times easier to get a repeat purchase from a customer than to get a new customer. And you can use remarketing to work towards that goal. The rationale is easy, your customers already know your store (awareness) and have already seen the purchasing process (experience) and all these make it easier for them to buy from you again. There is a good way and a bad way to do this.
The bad way is to keep using the same ads that you were using when these prospects (now your customers) were higher in the conversion funnel. You can tell them again how great your store is, or even worse show them product ads that they have browsed while searching for their previous purchase. The last one is a really bad tactic and one that builds negative brand awareness to people.
The good way is to find the best offer you can make to them to come back to your store. And usually, this is a cross-sell to something they already bought. For example, if they just ordered a 50’’ screen TV, it makes sense to offer them cables, a DVD player.
It is really easy to set this up as most times AdWords has already generated the “All Converters” list for you. It is advisable, however, to not just use the standard list but to create more lists such as “Converters: mobile phones” so you can do cross-sell campaigns. In order to do this, you can follow the same logic used in the “Category pages”.
Advanced setup tips
Up to now, we have covered the main audience segments you should build your campaigns upon. No matter your store size or the size of your budget, you should have these. They are both for the one starting their campaigns and the one running campaigns for some years now and looking for optimizations.
But there is more on remarketing. And in this section, we are going to cover some more advanced audience segments. They are not for all stores and you do not need to do all these to start. But they help, and they help a ton.
In software engineering, they have a nice concept called “catch-else”. Basically, in everything they design, they also build something that can handle all different values that might appear. This is a nice concept to also apply to our audiences.
By creating a “catch-else” audience, we make sure that if we miss something, this will also be targeted. And this has 2 main benefits. The first one is that we do not lose the valuable audience that will not be retargeted. The second one is that we are able to dive into the performance reports and understand what we are missing, something that might indicate an issue with either our analytics tagging or our audience build up.
In all the examples above, we are using a standard “time size” of the audience for all the different audiences (eg cart, product, etc). You can start with the default 30-day tag (or see below and find your optimal size). In any case, up to now, we are discussing about one size fits all case.
This is not the optimal, however, as you can easily understand. The propensity to buy is higher for someone that viewed a product today than someone that viewed a product 14 days ago and has not visited our store again. This user is searching different stores and possibly stopped coming back to you. This is a really good case for remarketing. So what we should do?
We need to actually make multiple audiences for each audience category (eg cart, product, etc). So we can make a “cart-1”, “cart-7”, “cart-30” or even go “1, 3, 7, 14, 30,60, 90”. Irrespective of how granular you go, by doing this you are able to bid differently on these audiences and maximize your returns. In order to do this, you can use the remarketing grid in order to build your audiences. I could easily list this here, but I like to give credit where credit is due and the remarketing grid is a really good one out there.
Pro tip: Beware of the exclusions due to small audiences. Better to just select the days you want, without the exclusions and just make sure the bids are in the right way to funnel the right audience in the right group.
Break AdGroups by
Audience time size
Setup the optimal max time you retarget a user
Google allows us to build remarketing audiences for users that have visited our website for up to 540 days earlier. 540 days! This is kind of insane and because many advertisers use such audiences, the whole industry is taking a blame. You know this case where you visit one store to buy something, you buy from another and you get retargeted for more than a year! Yes, you know.
So we know that we should not use the 540, this is the upper limit. What is the lower limit? One could consider 30 days as the lower limit. This is the default window and for some reason. So we are looking at something that is 30<x<540, but what is this “x”? There is a data-informed way to decide.
You can use a Google Analytics report to find the optimal window size. The report is called “Time lag” and is under the Conversions tab. It shows how many days a user first came to your website before completing a specific goal (you need to select the e-commerce transaction).
Go to the report and select to see a max lookback window of 90 days and limit the report to show data from “Google Ads”. What you will see then is a break-down of the transactions based on how much time passed from 1st visit to the transaction.
After seeing this report, most likely you will be amazed. You will be amazed because although as users we visit multiple times a store to buy something when we think about our customers we think that they came 1 or 2 times on our website. Although many times, they came more, way more as you can see in the example below.
Ok, we got it. People visit many times. Now let’s go back to our technique. How you can define the optimal size? I would say that it is safe to retarget people up to the time that ¾ of the people made the buying decision. In order to find this, you drill down the 30-60 and 60-90 sections of the report and start measuring from the bottom by adding the transaction % till you reach. When you reach the 25% mark, then that day is the optimal size to retarget. Use this as your max audience.
Lastly, as you can understand the ¾ threshold is a rule-based one and is not based on some data analysis. One could play with this by either increasing or decreasing and noticing the impact. As you can understand the more you increase it, the more aggressive you are becoming and the most people you will annoy. I would go with 75%.
There is value in targeting people that have visited your website in more than the optimal window time. You can target these people in periods eg 1 week every 3 months or something to get them back in. But I do not include this into this guide, because on a high-level I consider remarketing going back to someone on this initial consumer intent. If you target them later, it is more prospecting than remarketing and this is why we will include it in a future guide (coming soon!)
High potential products in another feed
Earlier we discussed about the audience of people that visited product pages. As you can understand this is a super-valuable audience as these people have already shown interest in a specific product, so they are lower in the conversion funnel. This audience is also super important because it powers up the dynamic remarketing campaigns (will explain this lower in the guide) that is one of the higher-performing ad campaigns you can run for an e-commerce store. But all products are not the same, they do not have the same value for you. You may define as valuable for you because it may be high price, high % margin, high absolute margin, you have a ton of stock to sell. No matter how you define you have some “HiPo” (high-potential) products and it makes sense to put an extra effort on this.
A way to do this is to break this audience into people that viewed this “HiPo” products and “A/O” (all other). This way you will be able to bid more on this “HiPo” products and you will also be able to segment reporting based on this dimension. Although I list this in the advanced tips on audiences, technically this is implemented creating a new dynamic remarketing campaign and applying a new product feed that contains only these products. This works. A lot!
If you are new into dynamic remarketing or not very technical, it is totally ok to not get this the first time. Ultimately you will. If not, you will get this and will have an “aha” moment. If you have any questions, shoot me an email at email@example.com and I will be happy to help.
In this chapter
Remarketing Ad formats
When we are talking about remarketing, or any other form of online advertising, the first thing to come in mind is the actual ads and this is what we are going to focus on this chapter. Up to now, we have discussed about why to do retargeting and how to structure your campaigns to target the right audiences.
The reason we set such a granular campaign setup is to allow us to show different ads to different audiences. In a form this is about personalization, it is not 1:1 personalization but it is about adapting our message to our target audience. And why would we want to do this? This is simple; it is all about relevance. If our ads are more relevant, we will get both more people clicking on them and ultimately buying and our costs for doing so will drop given the auction based pricing nature of the major ad platforms. Relevance is king.
We are going to start by showing all the available ad formats in our toolkit and then I am going to tell you what to use in each case, combining everything to the campaign structure I have recommended in the previous chapter.
- Text: Text ads, such as the ones on the Google Search Network.
- Image ad: The notorious “banner” ads. These can be either static or have some form of animation
- Dynamic remarketing ad: This is a special form of image ads that display specific products. It is used for dynamic remarketing (on Facebook it is called DPA, or else Dynamic Product ads)
If you go through the documentation of Google Ads, you will see references to another format that is called “lightbox ads”. Do not use these for e-commerce! These are good for engagement and are mostly used for branding. They are not cost-efficient in order to drive sales as they are priced based on engagement.
Text ads are the first ad format that Google offered many years ago as a way to target people on the Google Display network. It was the natural extension of the technology it has from the Google Search network.
Although they lack in terms of appearance and branding (as they say an image is 1,000 words and 1,000 words are way higher than our character limit on a text ad!), they can perform really well most of the times due to the extra inventory that is available for them.
This is the standard ad format, a generic image ad. You can show almost anything you like in this ad as you let your designer make them.
If you want to think what are the main image elements that your ad should have these are:
- Title: the main message of the ad. This could be a benefit/differentiator of your store (The biggest collection of X), an offer ($20 off on all laptop cases) or the digital version of a brand campaign (“clever kids buy from us”)
- Subtitle(optional): Sometimes, you need to write some more text to explain or enhance your main message. This usually works on bigger ad sizes because you need the space.
- Main image: It is called image ad for a reason, you need visuals. And your main image is the most visual part of it. You might show a product, a store, a mascot or whatever both gets the attention of your target audience and helps you pass the message that you want. Sometimes the “image” is basically the title in a nice visual format.
- CTA: This is your call to action (CTA). It is the “learn more”, “buy today”, “get the discount” message that actually urges the person to make an action. It strikes me that still in 2018, we see ads with no call to action. This is wasted of opportunity. By definition, if you ask people to do something you will get a higher % of them doing this. More clicks, lower CPCs. It does not make sense to not use it. A clear example of how good this works is that Google does not allow you on text ads to actually include “click here”. Why? Because it actually works.
- Logo(optional): It is a good practice to include your logo on the ad. If you are a known brand, this will increase the visibility of your ad. If you are still a new brand, this will help you become a known brand as your ads will be out there. The only reason to not use this is if your ad is already cluttered and your logo does not add so much value there.
- Target URL(op) :
Tips to make them better.
Better ads, better results. While we instinctively know that this is true, traditionally it has been really difficult to show/prove the value of great advertising and its impact on the bottom line. The great thing is that with online advertising we can actually run a test and actually see the impact.
Please find below some tips in order to make your ads better:
- Always keep testing. Test, measure, refine. And keep testing. Test image, titles, CTAs, etc.
- Unless your ad is the digital version of a big ATL campaign, do not be too “clever”. You have fractions of a second to get a person’s attention and as Steve Krug says “don’t make me thing”. Be clear and to the point.
- The landing page should have a relevance with the ad. Do not lead people to something completely different as this could alienate them and actually bounce. Which means lost money for you.
Dynamic Remarketing ad
Dynamic remarketing ad is a special format of image ads both in terms of design/visuals and in terms of functionality. This ad format is used for dynamic remarketing, targeting people that have seen products on our website with products they have seen or products similar to them. In AdWords, it is actually that 90% of product impressions on dynamic retargeting ads is for products relevant to them.
In terms of visuals, we do not have full freedom as to what to put inside but rather we just select some basic principles for the design template. Each image ad is different completely personalized to the person and the generic template is filled with products, their descriptions, and their prices. In terms of functionality, these are ads have different click areas on the ad that can lead to different landing pages. So whereas if you show 3 products on a standard image ad, the visitor will lead to the same landing page no matter where he clicks, on a dynamic remarketing ad each product shown can lead you to a different landing page.
For starters, I would say that dynamic remarketing is a MUST. If you are not using this, you are losing money. I do not care if your developer says that development takes time, your feed is not ready or if someone from your Team says that these ads do not work or they get annoyed by such ads. These ads work.
The main reasons that these ads work are that 1) we do not buy on the spot, so getting ads reminding us about your store help, 2) showing fully personalized ads, such as the ones on dynamic remarketing increases the relevance exponentially.
On summer 2018 Google released Smart Shopping campaigns that are the equivalent of Universal App campaigns for e-commerce stores. These ads combine Google Shopping with dynamic retargeting. Although we have to admit that is clever from the point of Google (Facebook has played many more tricks such as the notorious view-through conversions) so that it can show better results, my view is that you cannot mix these 2 types of ads together and judge them as a whole. It goes without saying that remarketing will have better ROAS but you need to know this ROAS. Up to now, the industry has not fully solved attribution so it makes sense to have different ROAS targets for the different channels. Moreover, you want to know what AdWords remarketing is getting you back so you can compare, for example, with Facebook remarketing.
“Category ad” is not a standard ad format that you can find at the documentation of Google but rather a standard image ad with a different objective/offering. Given that this is so different than the standard generic ad, we are creating a category just for this.
You see we have the generic image ads where a store can advertise a benefit or a storewide promotion. And then we have the dynamic remarketing ads where we are advertising specific individual products. But if you think the user behavior and its flow as also seen in the enhanced e-commerce report there is a gap there and the gap is on categories. What do we show people visiting a category page, but not individual products? We show them “category ads”.
Most of you have already created “Category ads” and this is usually done with some custom made creatives for the X top-selling categories of your store. And you have seen that this works. The reason I am mentioning this here is that yes it works and yes you should find a way to scale this. But before we go into the scaling part, let’s define what a category ad is and what are its elements.
If you want to think what are the main image elements that your ad should have these are:
- Title: It is either the category name (“keyboards”) or the category name along with your main benefit (“Keyboards at great prices”, “Find all the keyboards”)
- Subtitle: No need for this
- Main image: In the main image you have to show a visual that showcases the category and guess what better to have there than the actual products. You can create a custom visual for the category or you can create a template and add individual products that highlight the category. The first way is the better but it has issues in scaling as it will be a lot of work that also needs to be refreshed. After all eg, you cannot show the same smartwatches for one year, as your ad will look old. The other way to do this is to add the individual products. If you do it this way, then you have the question of which products to add there and these could either be some nice products or the top-selling ones.
- CTA: This is your call to action (CTA). It is the “browse products” or “buy today”.
- Logo(optional): As already mentioned it is a good practice to add a logo and the only reason not to add it is if the ad is becoming too busy.
Difference of category ad vs dynamic remarketing ad. If you choose to display the category ad as a template with individual products, then one can ask what is the difference with dynamic remarketing. Let me give you the answer:
- The landing page of a dynamic remarketing ad is a product page whereas for a category ad is the category page.
- The target audience is different. With category ads, you target category page viewers, whereas with dynamic remarketing you target product viewers.
- You have way more creative flexibility on the category ad than on the standard dynamic remarketing ad.
- You can select which products to show on the category ad (based on your own “algorithm” or preferences) whereas you let Google decide on dynamic retargeting.
Category ads is an extra ad type so the question is if you are getting incremental value by creating these. My answer is yes you are getting because:
- You have more relevant ads for category viewers that are higher in the funnel. More relevance, better results.
- Most stores run dynamic remarketing, but few actually have category ads. By having this for your store you have a competitive advantage. And in the ultra-competitive environment of online advertising, you should take advantage of everything you can.
Ads for our campaign structure
After we reviewed all the different ad types, we now want to tie these with the audiences we have created so we create coherent campaigns. You can see in the table below how all these fit together. As you will see this is the full thing, but you can start with something smaller, so feel free to not use a row or some ad type but always do follow this structure.
Break AdGroups by
Audience time size
Time (eg 1, 3, 14)
How to implement
Ok, we reviewed all the different ad types that we can use to promote our store in the most relevant way to each audience. Now we need to see how we can implement these ads.
Not a long time ago, you had the option to create each of these ad types into Google Ads and optimize these yourself. These days, Google has introduced the “responsive display ad” format. In this new (as you can see also on the image) ad format available, you provide the main elements (images, text) and Google algorithmically generates different ads that are relevant on different places (“native ads”).
As you can see, when you go to create a new ad you provide Google with both the texts for a text ad (that will appear on the Display network) and also you can upload “ad assets” that Google will combine to create different ads.
If this campaign is targeted at product viewers and you want to use the “dynamic product ads” format (dynamic remarketing), you have some more options to further improve the appearance of your products.
As you saw, however, in the initial screenshot apart from the responsive ad format, we have the option to upload our own image ads. What should we do and why?
When you should choose responsive ads?
- Run dynamic remarketing: Unfortunately, dynamic remarketing is only available through responsive ads, therefore if you want to run this campaign (and you definitively should!), create a responsive ad. This means that at times, you may show a text ad, but you cannot have all!
- You only care about performance: I will start by saying that responsive ads get better performance and this is by no coincidence. They use all the machine learning power of Google and are able to show different ads to different individuals (1:1 targeting), something not available as targeting with standard ads.
- You do not have time/budget to create your own image ads: The great thing with these ads is that Google uses “ad assets” your provide eg your logo and creates image ads, so you do not have to.l
When not to use responsive ads?
- Your care about your branding: Let’s be real. In order for Google to customize your ad assets and create ads for different ad formats/devices/publishers, it makes some adjustments that many times do not look nice. If you are a big brand or if you are a retailer that care a lot about your perception (eg luxury fashion), these ads will not cut it for you (or your Marketing Director) so just don’t use these.
- You want to show a specific message to a specific audience in a specific way: Let’s say you are running a cross-channel campaign (Google, Facebook and even ATL like tv) and want to show a consistency across channels to get your message through. You cannot do this with these ads.
All in all, I believe responsive ads are the way to go. And I have faith, that as time passes Google will improve these creatives so we do not have any hesitation to use these in all cases.
Closing this section, I am also showing what are the ad specifications if you want to upload your own image ads. Google is really strict about these, so when you send the brief to your designer be sure to include these also.
Advanced remarketing ads
Given that this guide is meant to be the ultimate guide, I want to also include some more advanced ad types, as I also did in the audiences section. If you are just starting or if your account is small and thus you do not have much time to invest, you can skip these for now. But if you have the time, creating also this helps and helps a lot.
X-sells or cross-sells are ads that are meant to help you increase your basket size/value. Imagine that someone is looking for a TV. X-sell ad is an ad that will also show them a DVD-player or a beautiful stand. Technically these are just some more image ads, but semantically these are a different type of ad.
If you have created category audiences and created category ads, you can use these ads and create cross-sell adgroups.
Once more, although this is technically not another type of an ad, semantically is. And it is a really important one. Promotions are really commonplace on e-commerce. You have promotions on special days (Black Friday being the biggest one) or ad-hoc promotions that you run to boost sales.
Another type of special ads are the “category promotions”. These ads serve the purpose of promoting a specific promotion that runs on a specific category or promotes a message that makes a generic promotion more relevant to a specific category.
Break AdGroups by
Audience time size
Time (eg 1, 3, 14)
Cart, x-sell, category-promo, Promo
Dynamic ads, Promo
X-sell, promo, generic
In this chapter
Selecting the right KPI
"If you can't measure it, you can't improve it," said Peter Drucker, the father of modern management (if you have not already read any of his books, do read the Effective Executive. Thank me later!) And who am I to disagree?
If you can't measure it, you can't improve it
You are already tracking conversions in your AdWords account (if not, stop reading and go do it). And since our focus is on e-commerce, you are already measuring conversions and their value. Given your setup, there are tons of metrics that you have available in order to measure the performance of your campaigns. Now it is our job to select the right metric. Yes, that is right, I said metric and not metric(s) and let me explain why.
The whole idea of a KPI (that is so heavily abused these days) is to have a KEY performance indicator that clearly shows you if you are winning or not. As a result, you cannot have as your target 2 or more metrics. You should always have one metric that you measure success. You can use, however, secondary metrics to show you the scale of the success (1st: ROAS, 2nd: conversions) or to set minimums (ROAS as the KPI but with a minimum of $12.000 ad cost/month). Since we have agreed that you have one, now we need to figure out which one.
I will not even discuss anything other than conversion metrics (although I still audit accounts that have as their KPI the CTR%!). At the moment, Google Ads have 20 conversion metrics available at the campaign level under the “Conversions” tab of the “Columns section”. The 2 more relevant are i) Cost/conversion (CPA) and ii) Conv value/Cost (ROAS). The difference between these metrics and the resulting actions that one can take after monitoring these is huge when the value/transaction can vary significantly. And given that we are discussing about e-commerce, this is always true. You can have a transaction of $20 and a transaction of $820 in the same day for the same store. This is why I strongly recommend going with ROAS as your KPI.
Without a doubt, ROAS is not the perfect metric. There are more aspects of the value that you can include and a quick one is the gross margin of the business (implementing ROAS on the “gross margin” not on the revenue). Other more advanced aspects of the customer lifetime value and offline conversions, but I am saving this for another guide. Although not ideal, ROAS is a good way to go.
You need to have a target for remarketing (and all other parts of digital advertising). Period.
One of the biggest benefits of online advertising is that it is measurable. We know how many impressions, clicks, and conversions each of our ads generated. And this is great. Given that we know and we will be able to know also in the future, it makes total sense to set targets for our campaigns. We do not want just to advertise and get the results as they are because this is for sure not an optimized process. We need to set a target and drive all our actions in order to achieve this target.
What gets measured, gets managed
As Peter Drucker, the father of modern management said “What gets measured, gets managed” so we should think hard about what to measure because then all our activities will go that way. You may have different goals that seem like the right ones but you should select the one that is most relevant for your business. To make an example out of the fitness world one can irrespective if you want to have muscle growth or get ripped you have to train and eat right. But depending on your exact goal you should do things differently. So it is imperative to be clear about your goal.
On a high-level, on PPC you have two types of targets:
- Get the best results based on a fixed budget
- Spend as much as possible with a fixed effectiveness target (eg CPA, ROAS)
Different businesses have different targets, but if you are asking for my opinion if you are into e-commerce it makes no sense to not follow the 2nd way. If you target setting is right, you will be making money with on the ad spend you are investing and thus does not make sense to not try to scale it, while you remain profitable. Moreover e-commerce most of the times has cash-flow advantages because although you are collecting the money asap, you usually pay your vendors and Google on a net-30 or net-60. So it is even better. It happens, however, to have a Client that has a budget target. Usually, these Clients do not really believe what they are seeing so it is your job to excel at their target setting and then educate them on the potential of going with unlimited budgets as long as you stay profitable.
If you are in a good position to have an unlimited budget based on a fixed effectiveness target it is really important to set the right target. In order to help you with this, I am sharing some thoughts:
- If you have a perfect attribution model, use the same ROAS/CPA as with all of your other campaigns
- The ceiling for your target should be the break-even point where the gross margin equates with the advertising cost (do not go with LTV on retargeting)
- A good rule of thumb is to set the target +40% your Search Generics target
Note the inverse proportional relationship between spend and effectiveness. For the example of ROAS.
Up to here, you have selected the right KPI (think ROAS) and you have set a target ROAS that you are trying to hit with as much ad spend as possible. Now let’s go to one of the biggest levers you have in order to reach your goals; Bidding!
What we want to achieve
On a high-level, you start by setting your strategy/KPIs, then set up your campaigns and then perform various optimizations in order to consistently reach your KPI. From the different optimizations that you can perform, bidding is one of the most impactful.
The reason is that with bidding is so impactful on reaching your KPI is that you effectively control 2 things: the average cost (per visit) and the exposure of different ads on different audiences. With your bidding choices you control the average cost and because your KPI is usually affected by revenue and cost, cost control is one great factor. Also with bidding, you control the exposure of different ads because you can bid-up (boost) the ads that have good performance and bid-down (limit) the ones that do not perform so well, this way you are optimizing the overall performance of your account.
Lastly, the great thing about birding is that if you select the right bidding method and a suitable bidding schedule, bidding does not need to take a lot of time. If on the other hand, you choose a non-suitable method it will take both time and cost you a lot on performance.
Available methods of bidding
These days, Google Ads offers a variety of bidding options and this gives us the opportunity to find one that is in full alignment with our KPI. There are two big categories of bidding methods, the manual ones and the automated ones (now called smart bidding).
Let me start by explaining the manual ones, that was the first introduced by the then Google AdWords. On these methods, you are able to set yourself different prices (bids) on the actions that users can perform on your ads. You can bid on 3 different actions: click, 1.000 impressions or engagement with your ad (for immersive ad formats). In terms of performance advertising (a part of which remarketing is) we care about bidding on the “click” action because the other two are focused on branding (either showing your ads to many users or have some user interacting and effectively making a brand connection with one of your ads).
On top of the automated ones, Google Ads offers 4 automated bidding methods, now called as a whole “Smart Bidding”. These are part of the various Machine Learning initiatives by Google that aim to give us better tools to manage our advertising. On these methods, you do not set an actual price (bid) but actually what is your target value on the KPI and let the machine handle the rest.
Manual bidding methods
- Manual CPC
- CPM (no)
- Cost per engagement (no)
- Enhanced CPC
- Target CPA
- Target ROAS
- Maximize conversions
Where you can apply the bidding methods (AdGroup, Campaign).
But on what part of the Google Ads hierarchy (Account, Campaign, AdGroup, Ad, audience) can you set your bids? The bids can be set on the Campaign or AdGroup levels. And how can I choose which of these should I use?
I assume that your KPI is either a ROAS target or a CPA. In that case, my recommendation is to start with manual bidding (setting a CPC price for each click) and after you get your first 50 conversions do a pilot with the Smart Bidding methods. Give the pilot ample time to show its potential (3-4 weeks). If it helps your numbers, keep it, if not return to the manual method.
In order to make the most out of your manual method, add the “enhanced CPC” option (which effectively adds a “smart” layer upon your manual bidding) and perform periodic bidding optimizations (weekly, not daily) so you can optimize the performance. You can do this with a decision tree approach.
How to select your initial bid
There is no exact science on this, so you should make fast iterations and monitor in order to reach a bid that can both drive you enough traffic so you can get data and make the right decisions and not bid too high so that you do not overpay and kill your performance.
You can look at IS budget lost to ensure that you are not overbidding.
On smart display campaigns:
To be eligible for a Smart Display Campaign, an account needs to have at least 50 conversions on the GDN, or at least 100 conversions from search ads in the past 30 days. Daily budgets also need to be set to at least 10-15X the target CPA bid. As such, this format is only really available to larger advertisers spending significant amounts on AdWords.
If you have not used some automated bidding strategy (as listed in the previous section), bid modifiers are extra levers you can use to optimize the performance based on your goals. Your standard bid on remarketing is applied at the AdGroup level, and if you are following my proposed structure this means that you set a bid for a specific audience (eg homepage visitors of the last 30 days). But even these coherent audience can be further broken-down eg by the device they are using or if we want to target them on weekdays vs weekends. This is where bid modifiers come in.
Google Ads give us the option to bid differently on 3 different dimensions:
- Device: what device is the user using (mobile, tablet, desktop)
- Location: at what location is the user browsing from (location can be targeted with different granularity starting from the country level but down to a radius around a specific location)
- Day/time: at what combination of weekdays (eg Monday, Saturday) and hours (name every hour of the day) will the user see the ad
You are setting the bid modifiers as a % difference on the base bid. Eg if your base bid is $1.0 by bidding +20% for the state of California, you will actually enter the ad auction with $1.2.
In order to apply these bid modifiers, you have to look at the past performance and based on these numbers, apply the modifiers. The idea here is that you want to boost what is working well and decrease the exposure of what is not working well. The north star on this optimization is no matter what dimension are you using to break-down the results, your main KPI is almost the same across each value of this dimension. For example, if you are checking at the performance by device and your KPI is ROAS, you should have x3 across all devices. This means that you are advertising in the most efficient way.
Important implementation note
Do not go nuts in applying modifiers based on performance before looking at the volume of the data. You cannot set a bid modifier at something that has just 1 conversion. If you do not have a really big account, I would advise you to apply the bid modifiers at the campaign level (not lower), by using data of at least 30 days and by applying these modifiers for the top values. Eg if you are advertising in the US, you could apply location bid modifiers just for the top-5 spending states. Or if you want to apply modifiers for the hours of the day, break the hours into buckers (0-07, 08-14, 15-18, 19-24).
AdWords scripts are a way to do interact with your Google Ads account in a programmatic way. By the programmatic way, I mean by using code. By interact I mean either getting data (for reporting/analysis reasons) or making changes (such as changing your bids).
In short, I really love AdWords scripts. We live in the world of automation and AdWords scripts is bringing automation in our ads. It goes without saying, that this is code applied to our Ads account and we need to know what we are doing. But, do not be afraid. Luckily there are hundreds of scripts out there to help us better manage our campaigns.
In regards to bidding for our remarketing campaigns, the most useful one is the bidding on a day/hour basis script. You can use a script to track performance by hour/weekday and then use these data in order to bid intelligently multiple times/day, even on a by hour rate. It goes without saying that this works only if you fo with manual bidding.
Although this script is not restricted only for remarketing, if you have an ad account that spends more than $5k/month, I would urge you to bid differently on the main campaign types (Search, remarketing, display prospecting) as these have different behaviors.
Apart from selecting the right audiences and setting the bids that will improve performance, you have one more big lever that you can use to boost your sales. This lever is no other than “ad testing”, so optimizing what ads you are promoting.
Showing the right ad to the right audience is important because as we mentioned this boosts relevance and ultimately this is reflected your performance because you are actually paying lower CPC due to the nature of the ad auction.
What elements of the ads you can test:
- Image sizes
- Title (both in text ads or the main message in the image ads)
- Call to action
The process in order to test different ads is:
- You start by defining what your KPI is (it is advised to use the same KPI you have for your campaigns. Only in instances of really low data volume, you can just use CTR as a proxy)
- Then you create different ads and applying these into the different AdGroups
- You make sure that you have enabled a “rotate evenly” campaign setting so that Google gives the chance in all your ads and does not start to prematurely optimize and show some of your ads more
As also with the bid modifiers, what you need to pay attention here is to not try to make winning decisions too early or with a really low volume of data. As also with website A/B testing, it is advised to let the ads run at least 1 week but ideally no more than 30 days. Lastly, you can aggregate the performance of your ads across multiple AdGroups in the same campaign.
As we saw above there are 2 ways to upload ads, one is by creating responsive ads and the other is by uploading your own image assets. All the above apply, only on the 2nd way whereas if you are using the responsive ads format, all these happen automatically. So no need to do ad testing.
Regarding responsive ads, we are seeing significant performance improvements and this is even more beneficial if you think that you are also saving the ad testing time. It happens, however, that sometimes Google is enabling some betas where it makes some crazy combination of your ad assets and you get weird looking ads (eg combining text from one AdGroup with an image from another). If you see such cases, contact Google support and they usually resolve these for you.
As we have already said, remarketing ads appear on various websites/apps of the Google Display Network. And as you can imagine, no two websites are the same. You get websites with great design and upscale, websites with a design from the 90’s, mobile applications that have ads all over the place, you name it.
Choosing on what placements (websites/apps) your ads appear is therefore pretty important. You can do this in 2 ways. The first is choose websites that align with your brand/color/design, etc. The second way is to choose websites that perform well in terms of advertising effectiveness. For remarketing purposes, my approach is to focus on performance after making sure that brand alignment is decent. Let me explain more.
In terms of brand alignment, I do not want the ads of my store to appear on “bad content”. Not around things related to sexuality, war, extremists, etc. The good thing is that Google has such a setting and we can remove all these placements with a few clicks. In the new Google Ads UI experience, you can make this setting at the account level (as you can see in the image below). My recommendation is to just select all these exclusions. This is for me being decent about brand alignment.
By doing this, your performance might drop. I could expect a slight increase in CPA and an increase in CPC. This will happen because you will be excluding low-quality, cheap inventory. Although this might hurt your performance a bit, it is beneficial for the bottom line of your Client. Why? Because of the concept of “negative branding ROI”, as described by Andy Sernovitz in his influential book “Word of Mouth Marketing”. What this concept means in our case is that you might be a losing a few cheap conversions, but you are not losing customers from associating your brand with extreme positions, or low-quality content. And your manager/Client will be happy about this :)
Now, in terms of performance, you can work on the “placements” level of your campaigns as you work on all other levels. Placements is one more dimension that you can optimize your campaigns. On a high-level, you want to do 2 things.
The first one is stopping placements that do not work at all (non-converters). They might not work because they are tricking users to click on ads when they don’t actually want or for other reasons. What we care is that they don’t work, so stop them. Given that your remarketing structure is really sophisticated and granular, we will have the issue of data scarcity, meaning that we cannot get enough data to actually do this on the ad-group level. In regards to the “non-convertersrdquo;, I would recommend to aggregate the data from all your remarketing campaigns and stop all the placements that have substantial click volume but do not bring any conversions. You can define substantial by calculating how many clicks you are usually getting to get one conversion and then double it to be on the safe side.
The second one is doing bidding optimization on the placement level in order to boost placements that perform well and reduce the exposure of others that do not work. In order to do this, you have to judge your campaigns by the KPI that you have defined and break the placements into performance buckets. The main bucket, for example, can be the one where the performance is +-10% of the average KPI. And then make buckets for every 10pp in the left and the right. Boost the ones that have better than average performance and reduce the bids on the others.
If you have a bigger than average budget, you can be a bit more granular and aggregate the data per ad type (dynamic retargeting, generic ads, text ads) or by campaign type (intent, retain).
How to deal with placements on mobile apps?
Given that many of these apps are of low quality and actually trick the user into clicking (and us as paying!), many times we want to exclude advertising on apps altogether. Up to September 2018, this was done byadding a placement exclusion for adsenseformobileapps.com. Going forward, this does not work and in order to exclude advertising on mobile apps, you have to deselect app placements on the settings of your display campaigns.
In the case that you are advertising on mobile apps, you have the option to exclude specific types of applications from your campaign settings. And treat individual mobile apps as placements so stop non-converts and implement bidding optimization.
In the era of TV advertising, there was the common knowledge in the ad industry that a message was “written” to the minds of the consumers after them being exposed to it 7 times. If you show this lower than 7, you have not gotten the message through and if you are showing more than 7 you are not efficiently using your ad budget and it is preferred to use the budget to increase your reach (how many people you run ads to).
How many ads a specific user sees is the metric of frequency and it is great that Google gives us the option to limit how many ads we show to a specific user. We can set this at the campaign level and we have the option to either let Google decide this or set this at the granularity we want.
Benefits of being specific about frequency caps:
- Do not get ad fatigue: Ad fatigue is the phenomenon where one user has seen your ad too many times and thus do not even pay attention to it. This means that you are wasting your ad impressions.
- Do not have negative branding effect: Showing too many ads to a user can even turn to a negative feeling for this user. I bet that you had at least one instance where an advertiser was showing too many ads on YouTube where you were constantly hitting “skip ad” on YouTube and wished to never hear about this company again.
- Better CPCs: People tend to get less interested in the ad message as frequency increases and this translates into lower CTR. What this actually means, however, is that you will be paying more for each click and actually hurting your performance.
Recommendation on the implementation of frequency caps:
- If you have not broken down your campaigns by time groups (remarketing grid) and your advertising for max 30 days, I would use 7/day. If you are advertising for more time I would go to 5, or even 3.
- If you are breaking down your campaigns by time groups, then you should set lower frequency caps as time increases (eg set 10/day for the “1-day” group, 7/day for the “3-day” group, etc).
Lastly, you can check the actual frequency of your ad campaigns by adding the “reach” columns at the campaign level reporting table on the Google Ads UI.
As I promised at the start of this guide, I shared all our learnings and recommendations on remarketing via Google Ads. These are the actual techniques we are using on the campaigns we run for our Clients and now you can also implement these in your accounts.
P.S. - At adaplo we are relentlessly working to automate the campaigns of online retailers to provide a performance boost. If you do not have the time to implement all the above, I would be happy to share how we can help.