Oct 122010

How Retargeting Can HURT Sales

Advertising money can be wasted, but can it ever actually hurt your sales?

Yes! Particularly if you have the wrong attribution management system in place.

Let's take the case of re-targeting. At SMX this past week I heard one agency rep say Google re-targeting generated a 7 to 1 ROAS for his clients. He was asked what sort of controls he was using to determine the incremental lift and he said: "er, uh, none."

I'm willing to wager that re-targeting actually reduces sales for this company.

How could that be?!?

Simple: Some fraction of buyers will place an order on the first visit to an advertiser's site. Some fraction will leave and come back later. If all the people who leave before buying are shown re-targeting ads, re-targeting will get credit for a big slug of orders that would have happened anyway in addition to orders legitimately inspired by re-targeting.

Re-Targeting Heroics

Herein lies the rub: if all of the credit for those orders is given to re-targeting efforts then credit must also be taken away from the channel that drove the initial visit. If those driver channels are being run to efficiency metrics, then the ads that drove the orders will effectively be bid down as they are perceived to be less valuable. That, in turn, leads to less of the high quality traffic that re-targeting served to cannibalize.

By stealing credit from what actually drives incremental business, re-targeting poorly measured can actually lead to fewer total sales.

We see similar impacts with other touches that tend towards cannibalism: affiliates, email, brand search (paid and organic) and sometimes Comparison Shopping Engines. Clients that credit paid search only when it's the last touch end up shrinking their online sales as a result.

In one dramatic case, moving from same session only credit to smart credit attribution led to doubling the size of an already huge paid search program, and that had a corresponding positive impact on the more cannibalistic channels. The rising tide lifts all the boats.

Re-targeting can and should be tested. Showing 10% of the traffic a public service announcement ad will give a clear sense of the true lift created by re-targeting ads. Correctly handling fractional order allocation will give the right credit to each program preventing the advertiser from overspending on re-targeting and underspending on the marketing channels that drive the visitors that re-targeting feeds upon.

As online marketing becomes more convoluted, having smart systems in place to address attribution becomes more important.


21 Responses to "How Retargeting Can HURT Sales"
Jim Novo says:
I was wondering when someone would bring this up...and George, I figured it would be you! When is controlled testing going to become the standard? Perhaps now that a lot of web analytics companies have been acquired by companies with assets in the BI area, we will start to see movement towards this. Until then, people are simply going to waste a ton of advertising and margin dollars on "selling" to people who would have bought anyway.
Thanks Jim. The other challenge to progress on this front comes from pricing models. Several of the attribution management solutions are priced as a percentage of ad spend. This means the attribution manager has an incentive to encourage overspending, rather than smart marketing. The analytics firms may develop smart BI but they may not if it hurts their revenue stream. Some folks in this space are much more interested in crediting ads with "assists" in addition to creating other ads with the order, than they are in doing fractional order allocation. These assists then act as additive benefits of advertising, encouraging advertisers to spend more, where fractional allocation encourages them to spend smartly.
Great article George, So what the point to test retargeting if we actually can't see if that is a result of cannibalism or it is an extra sale? I'm pretty sure that as a channel remaketing won't bring more then 5% of sales basically for any market. So any results will be around statistic mistake or deffinitely not enough to make any conclusion .
Andriy, thanks for your comment! Re-Targeting absolutely can create real lift. We're hoping to find that done well, it can create real lift at acceptable ROI, and further that the volume will be enough to make it worth the effort. The jury is still out on this. You may be right, if it ends up driving too few sales to be worth the management time/cost then time is better spent elsewhere. However, for some folks 5% of a paid search program would be enough to justify a full time employee. It all depends on scale and the category. Let us know what you find if you try it out! George
Jun says:
Hi George, A few days after re-targeting/re-marketing was introduced by Google, my team immediately requested a meeting with our Google rep. Our boss was very interested in it and I was very skeptical about it, since I felt it needs a very robust analytics system to first, know what to re-target, and second how to measure it. After an hour of discussion, we decided it's not for us. For the time being anyway. But I guess would be back on that discussion table anytime soon. Great post by the way. Cheers! Jun
George, spot on with this post. Re-targeting is getting a ton of credit (stealing credit of organic sales) and it's only a matter of time until we see structured testing en-mass. We're doing some of this testing now within the agency but certainly needs to be done at a mainstream level. Spot on.
Great article! Your analysis of Retargeting truly exposes a major shortcoming of this practice. Retargeting is really in its early stages, however, so there is time for effective marketers to streamline their practices to squeeze out the best ROI. I do agree with your point on potential of retargeting to drive down bids, which in turn dilutes the quality of traffic sources, which in turn affects the overall sales potential. There is a fine line there, and a large scale drop could severely cripple some of the smaller ad networks who feature Retargeting. Still, as Andriy says above, the jury is still out on this. PPC is a stats driven industry which lends hope that Retargeting analytics will see swift evolutionary cycle. Regards, Cleo
George, another great article and spot on. A drum not banged loud enough. Retargeting has it's place, but it's not the panacea that many think; indeed - it's like taking too many vitamins so they become harmful to the base of your health.
Jun, Darren, thank you for your comments. Particularly if you re-target people who actually purchased, you may end up crediting some of your natural lifetime value to re-targeting instead. Again, nothing wrong with re-targeting. Catalogers mail books to people who've bought in the past to gain wallet share, and it works. Email marketing can work on this same principle. One just has to be careful how we think about and measure the drivers of our businesses. George
Cleofe, Chris, thanks for your excellent comments.
Samir says:
Hey George, Good writeup, as attribution is definitely a big issue when it comes to retargeting. Google's 'remarketing' solution isn't optimal - it uses a view through window of 30 days, meaning that all view through conversions that occur within 30 days of the user seeing their first ad will be counted as a conversion. With that said, companies that specialize in retargeting (ReTargeter specifically [full disclosure: I work there]) can change the view through window to whatever the advertiser desires. For ReTargeter specifically, the view through window is 24 hours by default. Even without view-through attribution, we've seen a lot of companies receive a tremendous ROI from click-through conversions alone. Above this, the branding value that retargeting provides shouldn't be ignored either. I'm actually in the middle of writing an article specifically for SaaS pricing models and how retargeting's branding value can really help companies with these models. But I definitely agree that more intelligent and vigorous tracking needs to be done. Starting with a small-test is a great start, and I know that some analytics platforms are working on things specifically for retargeting. Additionally, advertisers should give a smaller attribution to view-throughs -- I'm not sure if Google's remarketing allows that. Lastly, I guess I should refer back to that recent comScore report: http://www.mediapost.com/publications/?fa=Articles.showArticle&art_aid=136236 The tl;dr version of this article: Brands that use retargeting start to receive a much higher search volume over any other kind of media buying strategy. Yes, attribution analytics are important when tracking retargeting campaigns, but you should also compare the direct traffic that you're receiving before and during your retargeting test. I'm actually gathering data for another post on this topic as well. Thanks a lot - feel free to reply/email me if you'd like to discuss this further! - Samir
Thanks for chiming in, Samir. Certainly crediting only click throughs on retargeting helps clean up the problem, but there is no substitute for controlled tests.
I do agree with you here George and this is a key part of the attribution model debate. And you're also right when you say that testing is the vital element to determining what level of re-targeting works for your brand. I would argue that re-argeting can be a vital component to keeping your brand front of mind for the customer as they proceed through their purchasing journey and encourage them to repeat visit and ultimately make that purchase from your site. Re-targeting = customer reminder = re-emphasis = return visits = sale.
Great article. I agree that testing against a 'default' ad makes sense, but the science behind the testing approach also needs consideration. There is really only one way to test properly and that's by using a fully automated control group that randomly selects which visitors should be shown the control ad versus the re-targeted ad. Also while 10% might be a reasonable rule of thumb for control group size the right number is highly dependent on the campaign volume. Be wary of any company that ‘proves’ a difference between the re-targeted ads and ‘control’ ads if they don’t provide you with a statistical confidence rating. So my view is – we need a control, we need to calculate how large that control sample needs to be and we need to show the difference between re-targeted and control is statistically significant. Shameless plug - that’s what we do at Cognitive Match.
Excellent point, Glen. The random assignment is key AND the system needs to 'remember' to show the user the same version each time to keep the test clean. On the flip side, with control tests in place, it's important to evaluate the success of a re-targeting test program base on the run-rate trends. A smart re-targeting algorithm learns as it goes and will produce much more efficient returns at the end of a test than at the beginning. Advertisers should look at the "steady state" efficiency achieved at the end to determine whether the test succeeded, recognizing that they spent X% on placebo ads which they wouldn't have to spend going forward.
Dave Rekuc says:
George, I agree wholeheartedly with your concerns regarding re-targeting cannibalizing other marketing channel sales. However, I think identifying whether re-targeting has provided a lift in sales or not is only the first step. If the re-targeting program is indeed a success, the advertiser needs to take the necessary steps to identify which marketing campaigns initially drove the visitor to the site. Without this information we wind up starving our upstream marketing campaigns and losing market share over time. We need to move away from isolating our campaigns in a narrow-minded same session attribution model and look at them as a whole. The best approach for this scenario is to develop a multi-touch attribution model that examines the full cost of a conversion, not just the cost of the converting session. Either way, great article addressing an often overlooked issue. Thanks!
Thanks Dave, we agree, and have worked hard to build exactly that type of fractional attribution system. Check out today's posts for more thoughts on attribution done badly.
Thanks I really enjoyed reading this post and the point you made. So how exactly can one set up a retargeting campaign so that it does not take credit for sales that would have otherwise happened anyways?
Owen, thanks for your comment. The gold-standard approach is to run controlled tests in which a public service announcement shows some fraction of the time instead of your firm's display ad. You'll see that some fraction of the folks who see the public service announcement will return to the site naturally, and that is the baseline that the real display ads have to beat. The display ads should only get credit for the incremental traffic driven, not all of it. Once that tuning fork has been established using controls, a smart attribution system can incorporate those findings to prevent you from having to test constantly.


Check out what others are saying...
[...] This post was mentioned on Twitter by Rimm-Kaufman Group, ana peña, SEM Software, search tool madman, Keyword Grouper Guy and others. Keyword Grouper Guy said: How Retargeting Can HURT Sales: Advertising money can be wasted, but can it ever actually hurt your sales? Yes! P... http://bit.ly/99C6M0 [...]
[...] How Retargeting Can HURT Sales, Rimm Kaufman [...]

Leave A Comment