Retargeting For Retailers

Retargeting is a fairly simple concept. The goal is to turn potential customers into actual customers by bringing someone to your site who has shown an interest in your product or service.  The practice has gained a following in recent years because it consistently drives sales and revenue. Increasingly, due to strapped media budgets, marketers are relying on retargeting to boost the bottom line.  But here's the truth: While everything you know about retargeting is not wrong, it is likely incomplete.

When most people think of retargeting, they think of Site Retargeting, which is the practice of targeting users who have already visited your site in the hopes of getting them to return. For example, Amazon can drop a unique cookie on your browser when you look at a pair of shoes on its site. If you don’t buy the shoes, but then continue to surf the Web, a retargeting platform connected to an exchange using real-time bidding (RTB) can identify you (via your cookie) as someone who visited a specific page on Amazon. Amazon, meanwhile, will have already entered a bid that tells the ad exchange how much it’s willing to pay to serve you an ad for the shoes you looked at. Amazon will likely bid high, because you are a valuable potential shoe buyer that is clearly in-market.
Basic Site Retargeting of this sort is widely used and can be an important tool for marketers. But lost in the discussion are the six other types of retargeting, some aimed at persuading former visitors to return to a site and others that focus on acquiring new customers.  The most exciting of those six other types of retargeting is Search Retargeting, which boasts significant untapped potential for attracting new customers based on the search terms they've entered into Google, Yahoo, Bing, or other search engines.
At the most basic level of Search Retargeting, a company uses search data to target potential customers and bring them to their site for the first time. For example, users who search "speaker sale," and then browse around the Web might encounter display ads from an electronics retailer selling speakers. The key here is that the user never visited the site of the electronics retailer. The targeting was done strictly on the search term, making it possible to acquire a new customer. Targeting based on search terms is uniquely effective — just ask Google — because search terms signal a user's intent.
Given the recent rise of these new types of retargeting, one might think that traditional Site Retargeting has reached a mature, non-changing, static stage.  However, dramatic new approaches are promising to overhaul even Site Retargeting. Traditional Site Retargeting has focused on tagging sections of your site with cookies -- and this even got clever enough to show the last product you visited. 



However a new approach doesn’t aim to hard-code these segments by dropping multiple cookies.  Rather, a “programmatic” approach aims at pushing all the granular details about a users actions into a cloud-based user profile.  By keeping all user activities server-side, marketers can essentially query their site visitors in real-time and make up highly granular segments on the fly.

We call this practice Programmatic Site Retargeting (PSR), and it’s sparking a revolution in how marketers optimize their RTB buys.  By keeping all browsing and behavior data in the cloud, markers can examine each user's actions and determine the likelihood of different types of users to convert.  This "visitor score" can then be used to inform the RTB bidding – reducing bid prices on those users not likely to convert and increasing them for high value customers.  It works incredibly well because robust PSR data creates a sophisticated picture of a user’s value to a site, based on a wide range of data points including pages viewed, referral data, shipping address, environmental factors (time of day, day of week), items in cart, customer profile, and many others.

PSR transforms Site Retargeting from an intelligent guess into something much closer to a science. For advertisers, this means fewer wasted impressions and greater returns.
4 comments about "Retargeting For Retailers".
Check to receive email when comments are posted.
  1. Alex Ruschin from Buysight, September 20, 2012 at 1:46 p.m.

    Thanks for sharing Chris. True, there are many other forms of retargeting that should be considered by retailers to increase conversions, and ultimately lift the results of all their other marketing efforts. You mention there are 6 other types of retargeting, but then only go on to talk about Search and Site. I’d like to add some points that I think would be helpful for retailers who are considering retargeting.

    This article buckets Site retargeting as one type, but the reality is that many vendors offer different types Site retargeting, with a) different business models (CPC, CPM, CPA, Engagement, etc.), and b) different technologies that will deliver significantly different ROAS and true incremental lift. There’s also 3 types of Site Retargeting (Site Level, Segmented, and Product Level), again, all delivering different levels of results. “Site Level” is much more of a branded level ad, to a user that has been to the site in the past, and are then shown a more generic ad for that Brand when they are on other sites. “Segmented” speaks more to a specific audience of that sites visitors (ads tailored for Men v. Women, or Clothing v. accessories). And “Product Level” which is the most personalized and tends to have the highest engagement and conversion rate. Product Level is usually done with dynamically generated ads, and if done properly, combines recommendation engine technology, based on what the visitor looked at, and deep links them back to the product page on the advertisers site. Proper analytics to score true buyer intent is also critical in the success of these campaigns, and a few vendors can do this better than others.

    Reach v. Scale is also important to how successful a retargeting campaign can be. Googles reach is probably the highest among all display networks, but it’s only one network among many, who also find users online. And, if you’re looking for premium inventory on a top Comscore publisher site (which will yield the best results), you’re not going to get it through Google. Some full service retargeting companies buy inventory from all the major exchanges, networks, and also secure inventory from top publishers directly, which will ultimately deliver you the highest volume of quality impressions.

    CRM retargeting is another interesting way to reach your target audience of existing customers through online display ads. This is great for promoting Seasonal or Holiday Sales, Flash Sales, etc. This is done by taking your encrypted email list of past customers, and allows you to reach them through display ads online anytime, not just after they’ve recently visited your website. This is a great way to augment what you’re already doing through your email campaigns.

    I’ve heard mixed results with Search retargeting , but I look forward to seeing how it continues to evolve. Would be great to see some white papers or studies on the effectiveness.

  2. gregory calvert from eXelate, September 24, 2012 at 12:08 p.m.

    It is encouraging to see that site retargeting is becoming more sophisticated and that publishers and brands are becoming more proficient at using data and segmenting their audiences. However, no amount of audience science will solve the problem of limited scale that is inherent with site retargeting. You can only retarget the people that have actually been to the site.

    Search retargeting improves scale but I can’t agree that it’s the most exciting type of targeting, especially with the rush of new companies I see jumping into the space. Advertisers need to carefully evaluate these offerings, because more and more I see that it’s just new ways of repackaging the same 3rd party data. As a result, there is a wide spectrum of performance.

    Based off of the strict performance numbers that I’ve seen, Predictive Modeling (Lookalike Modeling) has yielded the best results outside of site retargeting. Predictive Modeling can expand on the scale of site retargeting from possible audiences in the tens of thousands to audiences that are in the tens of millions. And it’s superior to search retargeting which only relies on one type of data, a Lookalike model built from the right platform can layer Search, Site, and Social data for a much more accurate picture of whom you should be targeting. The better data going into the model, the better it will perform.

    (If advertisers are considering Lookalike modeling, it’s important that they are not being sold “Overlap analysis” or “Overlap modeling” and being told its Lookalike modeling.)

    **I have to strongly disagree with the comment that “premium inventory on a top Comscore publisher” will yield the best results, especially for a Direct Response campaign with strict CPA goals. The value of audience data, especially if used to buy on any of the exchanges via RTB, is connecting the right person with the right ad, at the right time. If you can efficiently target the best audience, performance won’t be affected if it’s on a top news site or my neighbor’s cat blog.

  3. chris sukornyk from Chango Inc., September 26, 2012 at 4:36 p.m.


    Really appreciate your comments. I agree that site retargeting has limited reach because you are only talking to your existing visitors. That said - programmatic site retargeting is a new way of making site retargeting more granular and more efficient. With respect to your comments on Search Retargeting I would have to disagree with a few of the points you made. Marketers are telling us that Search Retargeting IS the most exciting type of media solution they've seen because it bridges the worlds of SEM & Display and most importantly it delivers phenomenal results. Also, I do not agree that what you call Predictive Modeling is superior because it uses more points of data. In fact, proper search retargeting campaigns use just as many points of data as look-a-like systems. The key difference is that in addition to using multiple points of data (ie. Social + Site), everything can be tied back to a search event keyword which assists with optimization and provides the strongest signal of intent.

  4. gregory calvert from eXelate, September 27, 2012 at 12:20 p.m.

    Hey Chris, thanks for the response and your thoughtful feedback.
    I can see your point and agree that marketers would be very excited about search retargeting when compared to other past targeting options. In an instance where a marketer has been leveraging contextual placements on ad networks, then in comparison, search retargeting would provide a huge boost to performance. So yes, marketers should be very excited about that.
    On your other point I didn’t quite follow the logic of using one type of data (search), then adding two additional types (site and social), you would still end up with “just as many points of data?”
    As with any targeting system, the quality and performance is predicated upon the quality and amount of the data going in. I can say that a Lookalike model built off of first party data, collected in real-time on 98% of the US internet population (billions of data points daily) benefits from leveraging Search, Site, and Social data and layering ALL THREE into one holistic audience.
    More data points do not make Lookalike modeling superior, the increased performance we see on campaigns with challenging targets makes the case for using multiple data types, especially if optimizing towards a specific conversion. (online sale, email acquisition, donation, etc.)
    Search data is a very important component of Lookalike modeling. Search retargeting by itself is effective and will produce “phenomenal” results for a lot of marketers. However, there are areas and campaigns where strict search retargeting has limitations and the additional layering of site and social data will cover that ground.

Next story loading loading..