Personalized Retargeting Is Overkill

Data-Driven Thinking“Data-Driven Thinking” is a column written by members of the media community and containing fresh ideas on the digital revolution in media.

Today’s column is written by Alan Pearlstein, CEO at Cross Pixel Media.

I have been overwhelmed by Zappos retargeting ads recently. Apparently, I am not the only one. Michael Learmonth wrote a piece in Ad Age yesterday about the subject and I think it about it a lot. Data driven marketers are walking a very fine line, balancing what we can do and what we should do, and I believe that certain retargeters that focus on automated, dynamically personalized retargeting creative have crossed it. What’s worse is that there is no need to cross this line because I don’t believe that this approach to retargeting improves performance and I know it pisses a lot of people off and can hurt a brand.

I am a big believer in retargeting and we do a lot of it because it is an effective strategy. My concern is with retargeters that use tactics that scare the consumer by exposing too much information in the ads. One popular tactic is to utilize the last products searched as the basis for the ad creative. The theory is that the ad will perform better because it is being personalized (dynamically) to the specific products that were of interest. It makes sense, in theory, but it can be annoying to consumers in practice.

Since buying a new pair of very cool Puma sneaks from Zappos about 2 weeks ago (the author is going retro), I believe I have seen 300+ ads for Zappos that continually show me the same 5 pairs of sneakers that I was researching. Here are some suggestions that retargeters should consider:

  1. I already bought the sneakers they keep pushing, so I would recommend that remarketers remove buyer’s cookies from these types of retargeting efforts. A consumer that isn’t in the business would think that Zappos is a clueless company (“Why does Zappos keep bombarding me with ads for a product that I already bought?”)
  2. Frequency caps are even more important when you are dealing in personalized ads. I have seen the Zappos ads 10 times or more on certain days. Most marketers utilize a frequency cap of 1 or 2 impressions per day per unique. Dynamic retargeters MUST utilize frequency caps. My sneaker purchase was one small part of my life about two weeks ago. Based on the frequency I see the ads from Zappos, it feels like that one shoe buying event is defining my online existence.
  3. Dynamic, personalized creative can produce as many false positives as positives. If I was on an auto site and I looked up lots of Ferraris and Lamborghini’s does that mean I am “in market” for those cars? I wish that were true, but it isn’t the case. Dynamic creative can make many mistakes – the last products you looked at aren’t necessarily the ones you will react to and buy. What if the last products you looked at were a total turn off and made you leave the site? This is a risk not worth taking.
  4. Overkill – this is my biggest concern. Based on experience, automated, personalized creative in retargeting efforts does not improve performance. The reason this is the case is because you are targeting a consumer that is already predisposed to the web site that they visited. It doesn’t take much to get them back to buy goods at the site. Simple ads that remind them to return are all you need to get people to return and buy. I equate it to Search Engine Marketing and bidding on your brand name. When someone goes to Google and searches for “Zappos,” they are likely going to click on the Zappos search result regardless of what the text ad says – they were already predisposed to Zappos. There is really no need for an SEM firm to focus on the text ads for a “Zappos” keyword search. They don’t need to personalize them, or test varieties of copy. There is only so much fine tuning that is necessary – optimizing the ads for the keyword “Zappos” is a total waste of time and the same is true for retargeting. It is highly likely that the consumer will return to the site they already visited if they are interested. Personalized creative is overkill and unnecessary.

I am sure the retargeters using these tactics will disagree with me and I look forward to hearing why I might be wrong. Our data says automated, personalized creative doesn’t improve performance in retargeting and I know I am not the only one who is annoyed by this type of advertising. The data driven marketing community needs to show a little restraint if we are going to win over acceptance by the consumer.

Follow Alan Pearlstein (@AlanPearlstein) and (@adexchanger) on Twitter.

Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!


  1. While I wholly support your comments around our industry’s need to respect consumers and not creep them out, I can’t more strongly disagree with your point about performance. Indeed, Criteo is paid on a CPC basis, so they are ultimately the best judge of what creative units perform best. (Remember an ecommerce marketer’s motto, “Long live the click!”)

    To wit, here’s some data from TellApart. Dynamic creatives with recent products from the user’s stream perform just over 2x better than the exact same unit when populated with top selling products from the category. For Hayneedle, for example, these ad units achieve over a 2% user CTR… *at scale* for hundreds of thousands of users on millions of impressions per day.

    Now, your point about overkill is also well stated. Frequency should be tightly managed, and we find that CTR peaks at just over 3 imp/day for the average user (but is retailer specific.) With Realtime Bidding, however, frequency “caps” are a simplistic relic of the past. Instead, every impression’s value should be calculated and bid… and less inventory will be won as the value drops significantly after the third.

    I think the reason so many Criteo-driven Zappos ads are pummeling users is because Criteo gets paid for every click. I encourage marketers everywhere to look instead for business models that compensate the vendor only on click-based conversions (hint, hint :-). Only then can you be sure that you’re not spamming your users with too many ads and getting tons of duplicate, non-converting clicks.

  2. Interesting post on a topic that I too have been thinking about for a while.

    I agree with some of your points, and disagree with others.

    1. Stop showing me stuff I’ve already purchased
    AGREE. Conversion pixels should be fired for purchased products, removing them from the delivery pool. If possible, other related products (the shoes you looked at but didn’t buy) should also be removed.

    2. Frequency caps
    AGREE. The best personalized retargeting campaigns I’ve seen make stringent use of frequency caps – typically somewhere in the range of 3x per day for a period of a week or so, depending on how high or low the product resides on the consideration spectrum.

    3. Potential to show false positives
    DISAGREE. Nothing’s perfect; there will always be some bad (retargeted products) to accompany the good. The trick is to minimize the former and maximize the latter. But to use this as a rationale to completely stop, as you seem to suggest, smacks of throwing the baby out with the bathwater. (On a semi-related note, there’s an almost certainly apocryphal story related to this involving an outraged Microsoft executive and retargeted ads containing sex toys. Said exec’s outrage evaporated quickly when he realized that the ad was merely a mirror. Hah.)

    4. Overkill / Doesn’t help performance
    DISAGREE. And there are documented case studies to prove it. Your position seems to be: customers who have previously visited a site are so likely to come back that even showing a generic ad will suffice. Thus, personalizing the ad is overkill and runs the risk of ‘freaking out’ the consumer for no additional gain. Respectfully, I could not disagree more. If you wish, I can come up with lots of different scenarios that point out the errors here, but for brevity’s sake, I suppose I’ll just point to the public case studies and use them as an existence proof. (The case study I was personally involved with more than doubled profitability, saw more than 650% improvement in click-rate, etc, etc – all over non-dynamic retargeted ads.)

    That said – I agree with your direction, if not your reasoning and conclusion. Restraint is absolutely key. The Zappos ad you saw could have easily said something like this: “Hello Alan, please buy the Green Pumas you looked at 1 hour, 42 minutes and 32 seconds ago. We’re certain that they’ll look great along with the new pants you purchased 3 days ago, and Rachel (who lives with you at the same address and is probably your wife) can coordinate with the green hat she bought yesterday!”

    I think we can all agree that such ads are wayy over the line. The scary part is that such ads are, from a technology standpoint, 100% feasible right now. The challenge for our industry will be in policing ourselves, encouraging those who are stepping over the line to stop screwing things up for everyone. Murphy’s law may or may not win out here. Alas.

    A side note, I checked the pixels on the Zappos page. The personalized retargeting vendor that is so annoying you is Criteo. From what I know of them, they, unlike others in the space, notably their principal rival, Dotomi, charge for retargeting on a CPM basis. This incents Criteo to push as much volume as possible, which is possibly why you see the lack of “turn-off-the-ads-after-I-purchase-dammit” conversion pixels and no frequency caps at all in this campaign. One of the not-so-attractive things about the retargeting space is that because the impression volume is so small it’s hard to create a viable business on a CPM basis no matter how good the performance. This is almost certainly why Dotomi has adopted a model that is CPA(-esque) and Criteo is blasting away without frequency caps and other practices that would make them less annoying / freaky. I could be wrong, so Criteo people, if you’re out there – and I’m more or less certain one of you will read this – please correct me. 🙂

    • Eh – looks like Josh and I were working on posts at the same time. It sounds like he has better info on Criteo’s pricing model (CPM vs CPC vs CPA) – so I’ll defer to him on that point.

      (Unless Criteo wants to weigh in that is…)


  3. Vernon Niven

    Excellent read. Thoughtfully critical articles like this will hopefully push marketers to learn how to apply this new technology more intelligently, and to move a little bit faster towards that end.

    The irritating situation you experienced (I have, too) was probably caused by over-reliance on the simplest form of dynamic retargeting used by retailers, ie, pushing the SKUs you viewed on the website into display ads. Most DCO platforms can do this relatively simple (dumb?) task. I agree w/ you that it’s not very good marketing if overused or not used correctly.

    I am just guessing here, but Zappos may have been trying-out basic SKU retargeting without attempting to use the full power of their DCO platform. Knowing how smart Amazon and Zappos are w.r.t. online merchandising, my guess is they will get much better at dynamic creative marketing, and soon.

    Fact is, some (but not all) of today’s dynamic creative marketing platforms are capable of much more intelligent, sensitive forms of display marketing. They are most often not asked to do this by the agency/marketer for practical and understandable reasons (including the learning curve).

    For example, in your situation, Zappos could have leveraged other Zappos marketing content and feeds from parent Amazon to deliver the following targeted msgs/ads to you, instead of ignorantly pushing the shoe you just purchased:
    – 25% off running gear (like shorts)
    – invitation to a local 10k, sponsored by Zappos (clickthru Facebook or Eventful)
    – music for your iPod (Amazon mp3 store)

    To market the above, Zappos would have needed to input a few business/merchandising rules into their dynamic creative platform, and they would have needed to provide real time access to a few more data feeds / marketing content available within Amazon. Not all that difficult to do, but there is a learning curve required to design these campaigns correctly (new way of thinking about creative). The analytics also get more complex, but they also provide more insight.

    Bottom line, today’s dynamic creative marketing technology can do a lot more than most marketers are prepared/willing/ready to try. Most marketers are still experimenting. Retargeting DOES perform well, so they often start out there. But the marketers that figure out how to use this stuff to improve overall relevancy and brand appeal will reap the largest rewards, which go well beyond pure performance/CPA/CPO. Like improving brand loyalty.

    For this reason, I believe it is a mistake to limit the scope of discussions about dynamic creative marketing to purely tactical/performance goals like retargeting and A/B test automation. We need to talk about what it can do for brands and consumers, too.

    Based upon my experience at Tumri, I am fairly certain that today’s technology can be applied – right now – to transform display inventory into a new type of an “always on”, highly scalable communication channel where marketers can share relevant/personalized content, links and offers with consumers via any display-enabled device/channel… and return better quality insight back to the marketer re. what consumers really want to see (or not). I am not talking about 5 years from now. Some of the vendors out there (like Tumri) have these capabilities built-in to their platforms.

    Apologies if this sounds like an advert for Tumri. Just trying to nudge you/us to think beyond retargeting.

    – Vernon Niven

    Full disclosure: since 2005, I’ve been involved with, working on, and investing in dynamic creative technology primarily through my work with Tumri ( I don’t work at Tumri anymore, so these words are my personal opinions. Also, I don’t know anything specific about Amazon/Zappos’ online campaigns nor who they use for DCO.

  4. Vernon Niven

    wow, this article hit a nerve… excellent commentary by Josh and Pete (while I was writing mine)…

  5. All very good points. Frequency caps, proper conversion tracking/un-segmenting are all very important.

    Though I do see a problem in that many small-to-medium sized businesses are data driven — the value of branding is not something SMBs generally like diving into. SMBs want to see an ROI from their marketing money, which retargeting can deliver, but the branding aspect of retargeting is often ignored. With dynamic creatives, branding is further ignored in favor of conversion driven creatives. Advertisers think that the extra cost of serving dynamic creatives will be made up in the returns… but is that really true? Is the additional money being spent on extra conversions AND pissing people off actually worth it?

    I’m sure by now, with all the buzz around their retargeting campaign, Zappos is closely looking over their retargeting strategy. MAYBE they’ll move toward something a little less targeted and a bit more branding focused… or maybe the financial returns from their retargeting are good, and it’s just a vocal minority of people being pissed off by the hypertargeted creatives. I don’t have the answers, but time will tell, as I’m sure Team Zappos is reading all the buzz around their retargeting. Needless to say, we’re all talking about Zappos, and this is a very interesting, and perhaps the first big case study on the effects of personalized retargeting.

    – Samir Soriano
    Director of Marketing

  6. Nice to see familiar faces (Pete, Josh and Vernon)here.
    Feel compelled to comment here as our (Tumri) platform has been used very successfully by several leading marketers for their product/offer based re-targeting campaigns.
    The author is definitely right about frequency capping; however, he is absolutely wrong about the performance of these campaigns. Criteo might run into some long term challenges with their arbitrage business model; however, marketers can definitely benefit significantly from retargeting. Here are some facts:

    1. only 2-3% of visitors to ecommerce sites convert – the rest have shown intent to buy; then why not try to elicit a response from them using more compelling offers (mind you, this requires more sophisticated retargeting than just showing the product they abandoned).
    2. Retargeting with specific products/offers based on business rules (i.e. show cart abandoners a better offer, show category browsers a list fo products from the category etc.) have shown 4-6x lift from simple retargeting.
    3. we have seen more advanced online marketers are leverage the returns from re-targeting to fund their broader display ad buys.

    Net-net, retargeting is here to stay (provided marketers and intermediaries respect consumer privacy)

    – Hari Menon

  7. some really great points.

    Regarding performance – If we are going to say that personalized creative in retargeting “works” from a direct response perspective, I think we need to be talking about CPA, not CTR. The nature of these ads does elicit a high CTR (your a little unnerved and want to see what’s up, so you click), but I haven’t seen any improvement in ROI. Does anyone have case studies that show that these ads deliver a better CPA than retargeting creative that is more general? I also think we shouldn’t fall into a trap of “one solution fits all”. I could envision many scenarios where on type of retailer could gain no benefit and others could gain a lot of benefit. The campaigns we have run were for niche specialty retailers, not general retailers, so maybe that effected results.

    @vernon – I really like what you are saying about using the data to generate more intelligent, related offers. That’s where I see the future of using data in creative – this really helps advertisers move “up the funnel” to find new customers and expand their client base, not just retarget the same person with the same offer.

    I also think it is an open question about how this type of advertising is effecting a brand. I don’t see the Zappos ads anymore, so I guess they are evaluating this too.

  8. Alan: valid points. The advertisers we work with use CPA and see higher ROI. To Vernon’s point and to my earlier comment, advertisers who are seeing higher returns (lower cost per acquisition- CPA) are going beyond simple product based retargeting and using more sophisticated business rules. For example, one of our travel customers use different retargeting windows based on the type of the offering (i.e. airline, hotel, cruises etc.); they also use business rules like “if the user has been on both airline and cruises pages, first show cruise offer as it provides higher returns; – these are specific examples of business rules based retargeting

    -Hari Menon

  9. Hello all, great to see such lively discussion around retargeting. I wanted to clarify a few incorrect assumptions made above regarding the Criteo solution.

    1) Criteo works only on CPC and not on CPM. The only charge to the advertiser is when a consumer actually clicks on the ad and is redirected to the website product purchace page.

    2)Given that we charge only CPC, we have a very strong business incentive to maintain a strict capping policy, as it’s against our interest to show too many banners that don’t result in clicks. In fact, being click focused, we calculate the value generated for our clients and COS purely on post-click sales rather than commonly used view-through practices.

    Karen Dayan
    VP Marketing

  10. Actually, I’ve seen that the vendors who charge for view-throughs tend to show far fewer impressions. Why? Because all they have to do is show one impression (usually on ultra-cheap, below the fold inventory) to keep the cookie window live on that user & scoop up credit for sales that were going to happen anyway.

    Using clicks as the metric of success is exactly what retailers want, and I applaud Criteo for helping to lead this industry change. But there are plenty of junk, duplicate, non-converting clicks out there. What matters is the sale. Pay only on click-to-conversion.

  11. Enhanced remarketing no doubt produces very high lifts in all metrics – clicks, interactions, engagements, site visits, search activities and conversions. We indeed have data to prove it across a wide variety of clients. Basing these executions on user intent, and an intelligent assessment of what this user is likely to click/purchase requires the following functionality:

    a) Marketing finesse in terms of what should the next be in the sequence of impressions to deliver – copy, offers and experiences that are data driven.

    b) The ability of a platform/underlying network (like Criteo or others) to execute something like:

    “show the last 3 browsed products for 2 impressions, but exclude the products that the user has already purchased, and after that sequencing roll into related categories/upper level categories/brands/price ranges of products”.

    c) And constantly self-correct as the response rates show different patterns than the predicted business rules.

    d) This requires semantic analysis of the impression context, user context and the data/message/rich media asset feeds from the advertiser.

    We, at Tumri, have worked pretty hard over the past few years to deliver on an ad serving platform that does all of the above at very high scale and low latency – while applying marketer specified business rules.

    Addressing the “creepiness” problem requires full transparency, education and user control at the finest grained level. Again, this is next generation buy side ad serving functionality that we all must strive toward standardizing and perfecting.

    In my humble opinion, we owe this to the consumers that we all serve. It brings relevance to our marketing dialogs, and also makes the entire experience more pleasant. This need not be a zero-sum game.

  12. Alan – Joining many of the other comments, I agree that advertisers should be following personalized retargeting best practices, which include frequency capping and not pitching products the user already bought, best practices we follow at myThings.

    Regarding false results in dynamic personalized retargeting, the problem you describe exists for all behavioral and contextual targeting companies. Just because I’m reading an article about Hawaii doesn’t mean I’m planning a trip to Hawaii. But both these industries have grown exponentially because these targeting assumptions have proven to be valid.

    Concerning your last point – overkill – I beg to differ. myThings generates lifts in conversion rates of 200%+ because it does work – and most of our campaigns are on a CPA basis. Personalized and dynamic retargeting has been more successful for us than other forms of retargeting.

  13. AJ Kintner

    I want to bring up the observation that there are around 12 hard coded pixels on the Zappos site. Those 12 pixels then piggy back roughly another 54 other pixels for a total of 66 3rd party pixels. 12 which are Yieldmanager pixels which you could assume they are buying Yahoo inventory. Since this is a remarketing campaign this could result in 66 different companies bidding on each remarketing impressions. This could be causing the overkill in frequency.

    • @AJ don’t you think this is evidence that a lot of the RTB world relies on re-targeting and the actual site frequency and the site by site impressions need to be right priced? I think it would be an easy conclusion to jump to to just say, “well they need to run all of their re-targeting through one dsp and control their re-targeting frequency.” But the reality is that your campaign frequency cap should be a much higher number than a lot of people out there would recommend and that if you right price site and site frequency then your campaign frequency won’t matter as much.