Leave a comment
Get the GH Bookmarklet

Ask GH

There is surprisingly little written about how to effectively analyze the effectiveness of retargeting ads - yet most people agree that it is the most cost effective form of advertising. Part of it is an attribution challenge. Retargeting can easily be considered part of conversion rate optimization rather than customer acquisition. I'd love to see a discussion on some of the best ways to analyze retargeting ads.

  • JG

    jordan gutierrez

    about 5 years ago #

    This is a great question. VTC is really overrated and CTA is really underrated

    This is what we did at Wishpond

    Using your favorite A/B testing tool we created 2 variations, one with the Retargeting JS and one with no JS, then we started the test and compare the results

    We realize that people being retargeted converted around 35% better than non retargeted, we've also found that Youtube video Ads are by far the most effective retargeting tool our there

    • SE

      Sean Ellis

      about 5 years ago #

      Wow, really interesting about the Youtube video ads for retargeting. Also really like A/B testing with Retargeting JS on one but not the other. I spoke to someone else today who was doing the same thing. He said retargeting was less impactful that he thought when he tested it that way.

      • JG

        jordan gutierrez

        about 5 years ago #

        A good retargeting campaign is also an art, the typical "4 banners and point them to the hompage" just don't cut it anymore.

        I've explained the automation + ads on a different post, but here is another great hack that we used

        Wishpond used to be great at doing instagram contests, we have a lot of great instagram marketing articles that rank really high on SEO, so this is what we did:

        We created 52 different videos on different instagram tips, here is the master video:


        Then we created 52 different audiences, each audience was users who visited the page between day a and b where b is a + extra day, and then each audience had a video. The result, you will get a retargeting drip campaign on youtube! cool eh?

      • DR

        Dave Rekuc

        about 5 years ago #

        This test does understate the value of retargeting a bit though. Good retargeting campaigns should have an opt-out pixel on the confirmation page. So, what this test doesn't take into account is the users that place an order/inquiry on the site without ever having had a chance to receive retargeting banners.

        So, theoretically, these single session conversions should be the same in each segment. So, you have these single session conversions mixed in with the audience that took multiple sessions to convert. The audience that takes multiple sessions (or at least sees at least 1 retargeting ad) is the audience that you want to isolate in the a/b test.

        I've looked into this a lot and I don't think there is any way to build a perfect A/B test. So, what Jordan did is more or less the best out there.

    • SJ

      Sebastian Johansson

      about 5 years ago #

      "one with the Retargeting JS" - Can you explain it a bit more for us tech-retards?

      So you do an AB-test and in version B you remove the fb-pixel that builds your audience (or whatever js script is used to track visitors for retargeting).

      Then you do a Facebook retargeting campaign with the fb-audience built with version A. And the retargeted users should be sent to version A since the AB-test has placed a cookie on their computer since the first visit.

      And finally you see which version has most sales?

      • JG

        jordan gutierrez

        about 5 years ago #

        Most pixels are added using JS (javascript), most of the code between the tags is a form of JS... I would recommend any growth hacker to know basic JS logic, Jquery, css, html, basic sql and some sort of basic back end lang (Ruby, Pyton or PHP).

        The second part is mostly correct.

  • RG

    Ruben Gamez

    about 5 years ago #

    Most people are wasting their money with retargeting, and they don't even know it.

    On Bidsketch we ran a split test for retargeting (Facebook ads) and were surprised at what we found. Half of our visitors where tagged in KISSmetrics with a "yes" property and shown included the JavaScript for retargeting. The other half "no" and no JavaScript was included.

    Inside the retargeting app our retargeting stats showed great ROI, which for some reason is what everyone relies on. No one I talked to actually tests retargeting in this way (to find how exactly how effective their spend is).

    Anyway, we ran that test and found that it made no difference in conversions at all. Take a look: http://cl.ly/3k22210J2R00

    Basically, this means that those visitors that were being retargeted and counted as conversions because of retargeting, would've converted anyway (as the data shows). Since then we've turned off retargeting (we had it running for over a year before that) and have seen no difference in conversions (and have saved thousands).

    I'm not saying that retargeting doesn't work for anyone, but I feel like it's super important to know exactly whether or not it's working for you. You can't rely on the reporting that retargeting apps show you -- they're counting conversions that would have converted anyway.

    Even if it does work for you, you'll want to run a split test to know exactly how well it's working, to better manage your retargeting spend.

  • BC

    Brian Carl

    about 5 years ago #

    I really liked @jordangk 's answer above. I believe running A/B tests is important to understand how remarketing is working for you.

    I've also run a different type of A/B test, which I'll call the "PSA Test".

    The idea is to segment users into two groups and serve one group your remarketing ads and then serve another group of users an ad that is completely irrelevant. The irellevant ad can be an ad for a non-profit or even a different product that is owned by your company.

    The obvious disadvantage here is having wasted spend. For this reason I think A/B testing retargeting is more effective, but I bring it up because doing a "PSA test" can be very enlightening for team members. This test reveals how often VTC's are attributed to conversions that didn't even see a relevant ad. You will likely be surprised how often this happens. On the other hand, you also get a good idea of how often remarketing adds to your total conversions.

    It's important with any of these tests to understand they are giving mostly directional data. The effectiveness of this channel will ultimately change and fluctuate as your marketing mix changes and time passes. These tests help you understand if the channel is effective or not, but applying a CPA to retargeting based on these tests will never be truly accurate over the long term.

  • AS

    Alexis Safarikas

    about 5 years ago #

    As an agency we are really tracking final Conversion Rates. We based everything on final CPA (using Google Tag Managers, Facebook Pixels, etc...).
    For one of our ecommerce customer (B2C - one of the major soccer team in Belgium) , we have tested different stuffs :
    - Facebook remarketing (based on customers list) : lot of impressions and clicks but very low final conversion (here is the point to track final CPA ; because Facebook CPA always looks better...).
    - Display remarketing with Google.

    We have discovered that this strategies are good for branding; using the attribution model we are able to track (within GA) the results of branding campaigns for final conversion.

    For direct conversion through remarketing we are now working with Criteo and using dynamic remarketing :
    - For visitors : we track their interest based on their visit of the website and propose the product they have visited + upselling.
    - For customers : we track their other interests into products + use the products tag to link new products to what they have bought (cross-selling).
    - For acquisition (this is a specific case) : we have add the cookie tracking system to the webpage of the team (eshop is on a separate website) to attract visitors from their audience.

    Our methodology here is CPA oriented. We have decided that the CPA can't be more than 15% of the AOV ; based on that budgetisation we are just maximing the budget regarding that metric (not fixed budget).

  • IE

    Isaac Ezra

    about 5 years ago #

    Excellent topic indeed Sean and l'd love to see more comments!

    I'm the founder and CEO of one of the largest conversion optimization and email retargeting platform from Brazil with more than 300 clients, called ShopBack.com.br and I've been asked this question almost once everyday.

    So far l believe @jordangk 's answer is one of the most effective ways to analyze it and in fact we encourage all our clients to do it however sometimes GA's attribution modeling may require a more in-depth analysis cos it might trick you. I'm mentioning that because our business model here is totally on CPA basis, as @Safarikas was mentioning, and always following our client's GA which is on the "last-click" attribution but on our dashboard we uses VTC just to prove the effectiveness of the campaigns and indeed we can see a much higher conversion rate than on GA as most of you may know.

    Anyway, what we've seen so far in tests is that a consistent personalized communication and customer relation do increase conversion rates thus most of the time retargeting optimizes conversions as @bradflora also stated. And what we've seen is also an average 35% increase in conversion rate but it varies a lot and you must find what works best for you!

    Finally, regarding mainly email retargeting, it's known that customer service plays a major role on user's decision and reaching out this user by email when he's abandoned your website offering help, in a totally personalized way, sending the email as if it was from a real person with a personalized content to the user do count a lot positively. So this is why l like to say that email retargeting converts due to a good customer experience rather than being a "retargeting ad" itself.

  • SE

    Sean Ellis

    about 5 years ago #

    I'd love to see an answer from @bradflora (founder of Perfect Audience) on this one. Any tips Brad?

    • BF

      Brad Flora

      about 5 years ago #

      This thread is hitting all the low-handing fruit that we see driving results for Perfect Audience customers. In Perfect Audience we've got a feature we call "test patterns" that lets marketers slot campaigns into isolated user cohorts. This enables all sorts of fun things mentioned in this thread, with the two most common being A/B test and PSA "lift" tests.

      1. With an A/B test you might create two campaigns, each targeting the same audience with the same bidding strategy and same ads, but with different landing pages. You might create a "Test Pattern" with two buckets, each containing 50% of the Perfect Audience userbase and target one campaign to one bucket and the other campaign to the other. In this way you get true separation and results you can draw trustworthy conclusions from.

      2. With a PSA Lift Test you might create a Test Pattern with a 90% bucket and another with a 10% bucket, then build two campaigns targeting each bucket. The 90% campaign serves your standard ads to 90% of your desired audiences while the 10% campaign serves PSAs to the other 10%. Then you tally up the VTCs from each campaign and run the math to see how much incrementality there is.

      We're always looking for new features we can build to better enable this sort of testing. Retargeting works. Helping people find out how to get it working best for them is our daily work over at Perfect Audience.

      • CL

        Chris Lin

        about 5 years ago #

        Do you have an article on PSA testing? First I heard of it and it sounds interesting.

  • DD

    Daniel Daines-Hutt

    about 5 years ago #

    It depends on where in the funnel and the specific goal- if it's at the point of sale then we measure simple cost per sale and then tweak and edit - it's amazing how few people talk about how retargeting is technically a direct response medium and no one really talks about copy/design/ or measurement.... I'll see what I can do about it :)

  • JH

    Jon Hinderliter

    about 5 years ago #

    We segment a lot at UMSL by creating landing pages for every channel within the retargeting campaign. So organic, facebook, instagram, video pre-roll all have separate retargeting landing pages. It's a pain to setup at the start but we can now measure the effectiveness of retargeting and the efficacy of the demand generation side as well.

  • KS

    Kirill Sofronov

    about 5 years ago #

    I can only speak from app retargeting perspective but great question - the trickiest part about retargeting is to measure the uplift against organic "control group", whilst segmenting the audience.

    But essentially here are the "standard metrics" we use to measure performance:

    ROAS - return on ad spend (aim is to bring 2 USD per each 1 USD spent) - 200% ROAS

    CPA goal vs. eCPA - Total cost / # conversions (on avg. or per segment basis)

    Sales / Conversion uplift as noted above as opposed to control group

    As mentioned - its effective because of the usage of 1st party data, as opposed to 3rd party. If a user has purchased in the past and you want to up-sell him/her - he definitely has what I call "revenue driven behavior" - and for the group of previous purchasers the CPAs are much lower than for just targeting people who viewed / searched your product before. Thus its more about managing expectations and having the right goals / intentions in mind. Test, Measure, Fail, Learn, Test again, Measure, Profit - rinse, repeat :)

  • JM

    Joe Murfin

    about 5 years ago #

    Great question Sean,
    It's a difficult question to answer.

    Lots of agencies I've worked with just Hoover up any conversion that comes through Remarketing and claim it as theirs. Not just with retargeting but with large reach display adverts too.

    Suppose this is bad, but then you're trying to measure who would have bought anyway and who wouldn't have.

    Again with a lot of online marketing it can be hard to allocate each conversion to a particular channel. Especially when some conversions simply aren't tracked end to end due to a change of device etc. ( also when you're trying to hit the same people on different platforms)

    I recommend using google tag manager to segment each URL for retargeting and make your ads as specific as possible to each page.

    Facebooks new pixel is good for this too.

    Very few people will be seeing an advert and buying directly...it's usually a much more complex customer journey.

    Retargeting is effective because you know you're being consistent with your advertising.

    No easy answers though ( think there's room for a tool here)

  • NP

    Nilan Peiris

    about 5 years ago #

    One of my favourite interview questions is:

    You have 2 advertising campaigns:

    An acquisition campaign, that markets to people who have never visited your website. You spend $1000 dollars on this and it returns $1,500 dollars - 50% ROI

    A retargetting campaign that markets to people who have visited your website. You spend $200 dollars on this and it returns $400 dollars - 100% ROI

    Assume that the only way people find out about your website is from the acquisition campaign

    If I give you an additional $1000 dollars where do you invest it and why ?

    I've yet to learn the answer to this myself.

    The only way I run retargetting today is by running incremental uplift tests to understand the incremental uplift in conversion due to retargetting and then investing accordingly


  • RL

    Raggy Lau

    about 5 years ago #

    Retargeting works. Period. I tried different type of retargeting here at OnePlus. Next to A/B testing creatives, the targeting is really important:

    - Upselling retargeting: Bought a phone but did not buy any accessories? You'll get accessories retargeting ads (it is much cheaper than email AND you only pay per clicks).
    - Abandoned cart retargeting: next to an email you'll receive you'll also see related ads.
    - Youtube retargeting: as mentioned before it works really well.
    - Facebook video retargeting: works even better than Youtube retargeting (try to do a video less than 20 seconds)
    - (new) Viewed an instagram ad + FB retargeting: Instagram isn't really a platform for hardcore advertising. It is a platform for awareness (impressions), after they viewed the ad, you can retarget them on FB with a more product compelling ad.

    • SE

      Sean Ellis

      about 5 years ago #

      So are you suggesting people should just accept that retargeting works based on gut feeling? Obviously every prospect who is in a retargeting campaign is already in your funnel, so some of these people would convert anyway. How do you measure the increase in conversion rate that retargeting offers and compare that to the incremental cost of retargeting? Some of the other answers give some good tips on this, but for me "it just works" is a little hard to justify as an answer.

      • EK

        Eyal Katz

        about 5 years ago #

        You make an interesting point that can really be asked of any type of advertising. Would they have heard of us anyway and found us anyway? I think that walks the philosophical line and can only be truly tested if you have a lot of time and your margins are well padded enough to shut down all advertising and see what happens. But, IMO, marketing efforts, and retargeting especially, should simply be measured by ROAS.

    • MR

      Mike Ray

      about 5 years ago #

      Retargeting is cheaper than email?

  • MH

    Mark H

    about 5 years ago #

    We've done layered RT (day 1 shows one ad, day shows another, etc). And it's still had to figure out if brand search lift has happened due to this.

  • EK

    Eyal Katz

    about 5 years ago #

    First off, all of our campaigns have conversion tracking so that's our first and most important KPI for retargeting. Secondly, we also operate various analytics platforms on first click attribution to see how retargeting is working in support of our marketing efforts. Finally, IMO, retargeting should not be overly concerned with branding or awareness. Retaregting is meant for users who have already visited our website and did not convert or for users who have not been back to our web app in a certain amount of time. It's an action based campaign so it's actually not too difficult for to measure at least in our case.

  • AS

    Anupam Saha

    about 5 years ago #

    Quick question - can attribution work with retargeting and how to measure it? Not able to figure this out

    And this is a great question asked by Sean, its certainly got me thinking :)

  • CJ

    Campbell Jof

    about 5 years ago #

    Just I wan to know about Facebook retargeting ads. Do you think we should have a separate ad for Facebook retargeting ad or we can only include facebook audience (facebook pixel code only) to retarget visitors?