The Promise & Challenge of Behavior Targeting (& Two Prerequisites)

crown I am a NPR junkie (and support my local station KALW). Driving to the airport last Monday I was surprised to hear a segment about "behavior marketing". NPR covers everything of course but for some reason it was surprising to hear of the current technology cool thing on NPR! Now that NPR has covered this I feel I am late to the party!!

Behavior targeting has been the news a lot recently. The old story for BT was about targeting ads, though that does keep getting better (unwarranted plug: get adblock!). The new story, and news, is all about using different methodologies to target the right content, promotion, message etc to you on the site you are visiting.

Omniture's purchase of TouchClarity has brought automated targeting to our side of the pond. Kefta does some very interesting things in terms of your ability to input business rules to drive targeting, they were just acquired by Acxiom.

On the multivariate testing side, Offermatica has announced "affinity targeting" (though I have to ask Matt for a non-marketing blurb of what it actually does, meanwhile here is our friend Jonathan's excellent write up).

In a world that is as "action challenged" as our world of web analytics, we should welcome behavior targeting with open arms .

Not the "lets figure out how to serve you the best ads from your browsing habits" but rather "we have all this data about you in our web analytics tools, why don't we show you relevant content regardless of if you are there to buy or get tech support".


The challenge with web analytics always has been: scale, data and diversity. They all plot actively against our ability to take fast action.

Here is how they do it….

~ Scale: There are thousands of people ("visitors") on our websites at all times. And thousands more are coming. That creates a unique problem of huge numbers that impacts analysis (not reporting, analysis).

~ Data: These visitors and pages going back and forth generate tremendous amounts of data, at a level where it is really hard for a human, or even our tools, to parse it all fast enough for us to take action quickly (see my point of view on real time data needs ). Or even take action every few days.

~ Diversity: Most website owners are pretty bad at understanding all the reasons why people come to their websites (see my recommendation on how to measure primary purpose ). Trust me people are using sites for purposes that you did not intend, or created your site for and that complicates data analysis and action.

Behavior targeting done with the right tool for you means that you can overcome the Scale, Data and Diversity problem by, simple put, automatically understanding your visitors as they interact with your web presence and showing them the most relevant content.

Sometimes the content targeting is based on business rules you input into the "automated system" (like say in Kefta). At other times the content targeting is 100% automated, based on complex "machine language algorithms" (as in the case of Touch Clarity).

It is the ultimate holy objective: Right thing to the Right person at the Right time.

BT has been around for a while. What has changed is how much more accessible these systems are, and in the grand scheme they are getting cheaper (yes "cheaper" is relative to your own context!). :) The promise is tempting and you can do a lot if you do it right.


While the promise is tempting, and results often very positive for clients, I find that people seem to treat BT as a panacea. God's answer to all problems you could have on your websites.

It is important to remember that the principle of Garbage In, Garbage Out applies here, more than it does any other situation.

For one simple reason: At the end of the day what makes a behavior targeting platform tick, and produce results, is not its own native intelligence and smarts even though that is critical. It is your ability to actually feed it the right content which it can then target.

garbage in garbage out

You feed your BT system crap and it will quickly and efficiently target crap to your customers. Faster then you could ever have yourself.

So you should absolutely positively investigate behavior targeting platforms and how you can use 'em. In fact if you are a Fortune 1,000 company it might be a crime against your customers for you not to look into BT.

But before you implement make sure that you cover these two pre requisites (no matter which vendor you want to go with):

#1 Invest in rock solid customer "listening posts":

    Think about it for a second. What are you actually trying to do with BT? Right content to the right person at the right time. Holy grail.

    Now take a step back and answer this question: Beyond having a web analytics tool and using it do you actually have ways to hear from your customers and understand who they are, what they need and what kinds of problems they have (with your website, with your products and services, with working / dealing with you)?

    I suspect that most of us don't have active listening posts. Surveys, follow me homes / site visits, usability studies, remote testing, market research or phone surveys or listening to call center conversations. Qualitative data [for more details read chapter three in the book ].

    All of the above help you get a solid picture of who your customers. Understanding them means that you can actually come up with all the content that your BT platform needs to intelligently target to the right person at the right time.

    No amount of technological coolness from your BT platform can make up for your lack of good content. In the end end you still have to get off your quantitative high horse and get out and talk to customers.

    Do this for atleast three months before you implement your BT platform.

#2 Do A/B or MultiVariate testing first:

Readers of this blog are aware of my love for testing (of any type). Shifting the power away from you or your HiPPO's deciding the website experience to your customers dictating it.

Before you jump into BT it is a great idea to have implemented and done multivariate testing for a while. With multivariate testing you are not actually doing BT but you are trying to come up with various bits or pieces, or, I'll betray my loyalty, "mboxes" of content and throw them on a page to see what works. That process is very powerful in what it can teach you.

You will learn how unbelievably hard it is to come up with good content that will improve your customer experience (or conversion rate for you).

You will learn how painful it is for your organization to actually come up with creative and content (the marketing run around, the displeasure from sales in actually approving promotions, the road blocks from legal about what you can say and put out there, the displeasure from your "cool" creative designers to anything simple, the run around from your IT teams, and so on and so forth).

Then there is the learning around process.

Testing of any sort is not ad-hoc. For it to work in a systematic way you will have to create a structure and repeatable process with defined steps and roles and responsibilities and organizational clarity. As you struggle and fail and succeed with MVT you will figure all this out and if you are committed you will come out stronger at the other end.

This will be a priceless experience. You are now ready to move from you creating a few relevant customer experiences to automated system doing it for you at massive scale.

Do this for atleast three to six months before you implement your BT platform (and the initial three months can be in parallel to doing #1 above, establishing solid customer listening posts).

In summary:

You can't expect to get massive results from your Behavior Targeting efforts by simply moving from having a web analytics tool to implementing a complex targeting solution. You need to evolve first and ensure you actually have a process for understanding your customer needs / wants and ensure that your organization has the process of feeding good stuff into your targeting platform nailed down.

Update: A lot has evolved in the behavior targeting space, and a lot has not. Please see this comment for more current thoughts.

Ok I have had my humble (or not so humble) say, its your turn now. Do you agree? Is any of the above making any sense? What am I missing? If you have successfully done behavior targeting did you have the above two prerequisites? What else has worked for you?

Please share your feedback and critique via comments.

[Like this post? For more posts like this please click here, if it might be of interest please check out my book: Web Analytics: An Hour A Day.]


  1. 1

    Avinash, great post! Your call to do some simple testing first especially resonated with me. A company has to be comfortable with experimentation (and the possibility of failure that goes along with it) before building an automated program to do it.

    Given how much buzz is being generated around the subject, I think the vendors will have to demonstrate some degree of restraint here (I can't believe I just said that). Recommending a full BT program to a company that barely has its analytics in order will result in an expensive site that creates perfectly customized versions of pixelated Hell for each segment of visitor.

    The good news is that many of the vendors I speak with appear to have their customers' best interests in mind. I spoke with an account manager at Offermatica recently and was quizzing him on their new BT capabilities on behalf of a client. At the end of his explanation, he said to me outright, "You're not going to start out doing this, are you? 'Cause that could be….tricky. I'd really urge you to start with some simple A/B tests…."

    I wanted to hug him.

  2. 2

    I think your comment about testing is spot on, though I would go further. It is not enough to test before you start, you must continue to test (though I suspect you assume this, you don't say so explicitly). Decisions, in this case the decision as to what to target someone with, cannot be optimized once and then left. The world moves on, customers change, markets change, competitors respond and so on.

    Yesterday's perfect decision is today's dud. You must constantly test your decisions and refine them using something like adaptive control. Implementing an infrastructure that allows you to continuously test different approaches – a "challengers" with a minority of your customers while still feeding most through your "champion" approach will position you to constantly refine these decisions and ensure both better results and rapid response to changing circumstances.


    Author of Smart (Enough) Systems

  3. 3
    Patrick says

    If I understood this correctly behavior targeting/marketing is pretty much a new buzz-word for "personalization"?

    I guess I'm one of the few persons who are not all for this. I think it works great with e-mail marketing or Amazon's book offers, etc., but when it comes to content on websites, I think there's one problem that might be neglected:

    In SEO everybody's a bit afraid of the trend of personalization of the search results (that Google is obviously working on). Whereas, I think that delivering relevant content is always a great idea (SERPs, websites, e-mail), I think there is one really big problem with this:

    I don't know how many times this has happened to me, but I'm sure everybody knows this situation where they just want to show something to a friend. You're over at their house and are like: Hold on I saw this on the web yesterday, let me show it to you!

    Then you type it into a search box or type in the domain name trying to show a certain thing to your friend(s), but can't see it. You get a bit angry, continue to search but have to come to the realization that you're not gonna find it again, because the results are personalized and the person you wanted to show it to has different browsing habits. Let's go over to my computer and I'll show you is probably not very good usability!;-(

    Maybe this sounds a bit exaggerated, but I'm wondering what type of consequences it could have if webmasters/search engines take the content personalization idea a bit too far.

    I think if it's overdone it could make "word-of-mouse" less effective which is so great on the web. If you tried to show a website to a friend a few times on instant messenger, but they were never able to see what you wanted them to see, maybe after a few times you'll think "I should send this to…or well he wont be able to see it anyways").

    Or if people dislike personalized search and Google introduced it too fast they might lose market share to yahoo and msn, if they still have the same search results for everybody? I know many people in the search space dislike the idea of personalized results (not exclusively from an SEO perspective, but also from that of a user!)

    I think all in all this BT (=personalization?) idea is a great one, but I'm wondering if there are not only advantages to it, but also disadvantages. For ads (just like for e-mail) it should work wonderfully, of course. But with content it might be a different, more complicated story.

    That being said on youtube I think it works perfectly (just like on Amazon). I watch a music video and then I see similar music videos on the side and I find myself clicking through those videos (and watching them) A LOT and there's no problem with it b/c if I want to show it to a friend I'll remember the name of the song/video, of course.

    I would really be interested what you (Avinash & fellow blog readers) think of this?

  4. 4
    Patrick says

    Let me stress that I was not referring to word-of-mouth (though its important, too) but more to viral marketing on the web. Sending stuff to friends in an e-mail or on instant messenger.

    One of the most important principles in viral marketing is "if you make an idea is too spread it'll spread far. If you make it hard to spread it won't spread far". If your URL can hardly be rememberred b/c of 3 dashes and 4 keywords its chances of spreading far are bad and I think (overdone) content personalization could have the same effect.

  5. 5

    Avinash, in your inimitable fashion you have struck right to the heart of the newest buzzword.

    Having been in the thick of the last round of Personalization (think LikeMinds, NetConversions, Personify), and having been in the heat of the testing revolution, I have my own point of view on how far math can go in marketing.

    Regardless of whether behavioral targeting is the holy grail or just a solid tool in the toolbelt, it serves us all to think about Occam's Razor and find the simpler option if it is available.

    It is my fervent belief that the most important thing a marketer can do is engage with their consumer. As an exercise, start with the simple act of changing something purposefully that matters and see how it affects your customer. That is the simplest thing and still among the rarest.

    Once you can do this, introduce segments, discover segments, pull data from analytics, go crazy with math. But do it deliberately, and constantly check and see how it effects your customer.

  6. 6

    James : I love the title of your book!! Brilliant!!!

    You continue to test but a lot less, perhaps only radical changes or "interesting complex" changes.

    I tend to think of BT as a continuous methodology, it is always there. You understand your customers, you create content, experiences, promotions, "stuff" and as a result of it you are constantly feeding your BT system. It in turn, I know touchclarity does this (others do it as well I am sure), tests the new stuff, learns from behavior and automatically targets "stuff" to the right people at the right time.

    So maybe "constantly feeding" content is a stretch. Maybe you do it every week, month, couple months depending on your business and your ability to understand your customers. But you don't have to test as much.

    I hasten to add that this is a nirvana outcome. It is tough to get there, it takes time, you have to overcome your limitations to understand your customers and your ability to create content. Not easy.

    Patrick : Behavior targeting is not personalization.

    In some cases people call personalization behavior targeting, but they are completely different ball games.

    You can use a BT system to personalize if you want but what is exciting about TouchClarity now, and about Kefta's system from before is the ability to know nothing PII (Personally Identifiable Information) about a person and yet be able to target content.

    This is my personal opinion: For the most part personalization is really really hard to do. It is difficult to understand each person, amongst the million who are on your site, takes too much time and too much effort. Then it is really really difficult to respond back to each person in a "personalized" way. You can hack it, but it is rather obvious to your customers that you are hacking!

    When I think of BT I think of understanding micro micro segments of customers (small groups) and then understanding what they need, responding back to them with relevant content (in terms of creating) and using a BT system to target them efficiently and on a large scale.

    I hope this helps.

    I welcome the other BT Gods & Experts out there to add their wisdom to the discussion. :)


  7. 7
    Rahul Deshmukh says

    Another great post….we were waiting for a post on BT.
    One item to note is the confusion around BT and MVT.
    Behavioral targeting and MVT are similar technologies (both aiming at improving conversion rate), there are inherent differences between these two. The water is getting too muddy for what what Behavioral targeting is about and what MVT does. At times, they are pitched as being used for both MVT and BT.

    Behavioral targeting in my mind is "Real time CRM for Online customers at a session/visitor level". MVT is page based and BT is visitor based. MVT soups content based on recipes, BT serves the right dish according to the individual’s taste buds.


  8. 8

    Great post Avinash.

    I am running A/B tests quite a lot recently and learning very valuable insights on user behavior. As you said, It will help me to prepare BT as a next step!

  9. 9

    Thx for the reply Avinash,

    unfortunately I have to say that I don't understand what the difference between personalization and behavior targeting really is.

    I just read this blog post on Anil Batra's blog, where it says this:

    "What is Behavioral Targeting
    Behavioral Targeting (BT) is the ability to target users based on their behavior on the internet. Most commonly it used to target online ads but the technique can be very well used to target products and content.

    Behavioral (Ad) Targeting promises to precisely target the audience that matter most. Hit the users with the right message, a message that they care about. It is all about audience."

    To me this sounds exactly like the personalization algorithm Google is trying to roll out for their search engine: collecting data from a users browsing history and that way displaying more relevant content to them.

    Or what e-mail marketers (or offline database marketers in CRM) do when they try to target users better with their ads.

    Maybe the main difference is that personalization is thought of as one-to-one marketing whereas behavior targeting is thought of as one-to-some marketing as "thinks" said in his blog post and as you (almost) say when you said this:

    "When I think of BT I think of understanding micro micro segments of customers (small groups) and then understanding what they need, responding back to them with relevant content (in terms of creating) and using a BT system to target them efficiently and on a large scale."

    Thus BT is about delivering the most relevant message (ad or content or product) to micro micro segments of customers (small groups) whereas personalization is thought of as taking it a step further and trying to target people in a one-to-one fashion?

    Sorry, Im just really confused now! lol

  10. 10

    This is one of the best threads I've read in a long time. The content is great, everyone's input very cogent.

    Three points that come to mind…

    – This is a very new industry, there's a lot of buzz around it, but little is yet agreed upon in terms of simple things like what each of the terms mean (i.e. targeting vs personalization vs testing). Being in the middle of this discussion from a vendor perspective (Kefta and Acxiom Digital), I can say that hosted services of every flavor want to distance themselves from the failed personalization approaches of the 90’s. They were heavy, IT centric applications that largely relied upon explicit statements from “customers” in order to change something. The focus was on delivering something different, not measuring the impact or lift caused by delivering that different message.

    – A thought for Jonghee Jo from Victoria’s Secret… If tomorrow your marketing team offered an incredibly deep discount coupon across all of the major deal sites, what would happen to the nature of visitor needs? My impression of your brand is quality, but would this coupon drive large quantities of visitors with different needs? Often times the behavioral learning that is found through testing is relative to the majority population of visitors that you see at any point in time. Internal or external events can cause a shift in your customer needs. Similar is the concept of doing testing prior to the holiday shopping season, only to find that consumer habits change once the season is upon you.

    – I thoroughly agree with Rahul Deshmukh… testing tools are page or asset centricity versus targeting solutions focus on marketing to the differences between site visitors, visitor centricity. Well stated!



  11. 11
    Rahul Deshmukh says

    Good thoughts…Personalization and BT are different. When I think of personalization, it has more to do with the added features/content you provide to the end user that facilitates the user to customize his experience on the site. For example, Google targeted me with iGoogle. This was BT…I might be in one of the micro group of persona that Google was targeting the product to. They had the right message for me.

    Next step…personalization kicked in – Google provided me with the tools/features to personalize my experience on the site. With iGoogle, I was able to pick and choose what I want to do.
    Expanding on the analogy that I had on my post, BT would be telling me what is the best Chinese restraunt, but personalization is picking your veggies (being considerate to my friend Avinash here) to make the stir fry that you like.
    I welcome your comments/feedback.

  12. 12
    Ned Kumar says

    A spot-on post and an excellent thread. I agree whole-heartedly with your pre requisites. Recently I heard a few folks talk about wanting to do 'BT' and 'Customize'/'Redesign' their site etc., but when I quizzed them about their business focus and customer needs/experiences and gaps, they didn't have much to say.

    As you rightly point out, the key to success using any Marketing tool and/or methodology is really understanding your business and your customers. To me, both are critical as sometimes because of your business focus, you may not be looking towards specifically targeting a certain segment of visitors — not to mention changes in your business outlook as a result of seasonality etc. Also IMHO, once you have mapped out the customer experiences x customer profiles against your offerings (through testing / surveys/ research), BT can be great to drive/address the basic CPRA (Conversion, Penetration, Retention, Attrition) on your site.

    And lastly, I too am of the opinion that BT and Personalization are similar but yet different. The former (in my way of thinking) is focused towards a segment of customers — the key is segment. The latter is focused at an individual.


  13. 13

    Interesting post Avinash.

    I have taken forward the discussion in my blog. It is available at:


  14. 14

    Let us not confuse behavioral targeting with attitudinal targeting. The "learning from visitors" that you propose does not really tell them apart and reconciling them is where the money is in advertising.

  15. 15

    Ok, Avinash, I couldn't resist and so here's a link to my reply:

  16. 16

    Hi Avinash – really great post (yet again – I mean it though). Really busy so don't have time to do proper comment. But in essence nothing replaces the hard work and graft of at least 3 -6 months of thorough A/B testing etc before you go down the BT route. Thanks again for your insights! Marianina

  17. 17

    We have encouraged companies large and small to engage in on-site behavioral targeting for four months since releasing Offermatica Affinity Targeting with mixed success. I should clarify. When companies DO affinity targeting they see KBM improvement. What remains difficult is getting companies to do behavioral targeting at all.

    The difficulty seems to stem from a belief that it is complicated. Unfortunately this post, Jon Mendez's post and others succeed in establishing the value of BT, but in communicating the significance of the opportunity, create the impression of major work and organizational change. To be fair, between no BT and full BT, there is work and change required.

    Getting started requires neither of these things. Simply removing content, like a subscription solicitation for someone who has already subscribed and using the space for a paid ad, can be done in a day, and with no hand-wringing from the brand or legal folks. Small changes like this also have the benefit of providing a success story that can be used to convince the powers that be to allow more experimentation.

    Like living healthier, a small step in the right direction with positive results can make the bigger commitment required for the full value much easier.

    Great post Avinash! Thanks.

  18. 18

    Im a starter in the world of analytics and thanks to all you people especially Avinash for starting the discussion. It seriously gained a lotta insights now….thanks once again…take care!

  19. 19

    Even 4 years back.. still relevant and as always a good read!

  20. 20
    Mario Rodriguez says

    Hi Avinash, seven years have passed from this great post, do you have the same opinion with available tools like Woopra, Clicky or the like? Dont you think "customer profiles" may really enhance conversion rates?.

    I have visited BT Buckets that you mention in your book, I think it gives interesting options, anyway I recognize that first of all it is necessary to have the customer consent for this kind of tracking. But I think that if one works well the sign up process and the follow up of customers, making them know about offers really meaningful to them will not be invasive, I would really appreciate your opinion.

    Thanks in advance

    • 21

      Mario: First a minor semantic clarification, neither Woopra nor Clicky is strictly speaking a behavior targeting tool.

      Woopra provides a way for you to create user profiles, but you might find Kissinsights to be substantially competitive and valuable if you want to play in that space (I'm not connected to Kissinsights in any way).

      Clicky just provides a mix of IT and Marketing features, but is is unclear why you would want it if you have a free Google Analytics account with right now data, advanced segmentation, custom reports and a million other things.

      That brings us to our delightful question about onsite behavior targeting. There is still no winner in this space, or even a decent tool because as this post outlines the challenge with onsite behavior targeting was never the ability of the tool to provide features, it was always a company's ability to collect the assets, create rules, set up a process to make updates at scale, and all that other stuff that is the reaction (in real-time!) to a customer's behavior. This remains a challenge, hence no alternative.

      We have made some progress. You mention BT Buckets, that is a interesting one that allows for simple behavior targeting. The ability to target is also now a part of some CMS platforms where if you leave something in the cart and come back maybe they can do something simple. And they do make things incrementally better.

      But the the grand promise remains unfulfilled, not because engineers are not smart or investment has not flown into behavior targeting, it is all the other things. : )

      PS: This is very similar to why real-time data is almost always useless. See #4 in this post if you are interested in that: A Big Data Imperative: Driving Big Action

  21. 22

    While one-to-one marketing has been oversold for years, behavior focusing on certainly makes a powerful situation for one-to-some.

    The "some" in such cases is every section you choose to create innovative for. The thing to remember about behavior focusing on is this


  1. thinks says:

    All about behavioral targeting (or the birth of one-to-some marketing)……

    Wow! What happened yesterday? Suddenly, it seems like every site I read is talking about behavioral targeting. In case you missed them:

    Om Malik discusses Google’s plans for behavioral ads
    Matthew Roche gives a general overview of optimizing
    Avinash Kaushik provides two excellent pre-requisites before implementing behavioral targeting

    While one-to-one marketing has been oversold for years, behavioral targeting certainly makes a strong case for one-to-some. The "some" in this case is every segment you choose to develop creative for. The thing to bear in mind about behavioral targeting is this:…

  2. […] Avinash perceives a problem coming down the road with behavioral targeting, that is, while the machine is smart, the results are only as good as the content you feed the engine.  Absolutely right.  If you run campaigns designed around static demographics on a behavioral platform you have created a way to “efficiently target crap to your customers”. […]

  3. […] Zachary Rodgers over at ClickZ has written a good overview of the recent addition of behavioral targeting by Google AdWords.

    This new feature is using previous searches to fine tune what ads are presented to the searcher.

    Behavioral targeting has had mixed opinions. Avinash Kaushik, who blogs on analytics at Occam's Razor, sums up the method as: "Right thing to the Right person at the Right time.." […]

  4. […] She sleeps little and obsesses at 4:45AM over analytic-flavored black coffee. So much traffic from StumbleUpon again? He wakes every morning , checks bid management software to see Google API charges month_to_date, and maybe shoots off a quick support inquiry (Support Ticket #25486705-4456-432) to Australia. There’s time for her to read 3 articles in Search Engine Land on her iPhone while going to the bathroom. She emails herself post title ideas. […]

  5. […]
    The scenario above is an optimal situation however personalized experiences can still be offered based on the content visitors view, content which they neglect, products they add to their shopping carts, and seasonality. Web Analytics plays a big role in behavioral targeting because it allows you to cookie your visitors so you can track their trends/patterns on your website.

    Resources for Behavioral/On Site Targeting:

    – Omniture TouchClarity
    – WebTrends Marketing Lab 2
    – BlueLithium – Anil Batra: Behavioral Targeting 101
    – Kefta
    – Clickz: Target Behavior on the Site Level
    – Avinash Kaushik: The Promise & Challenge of Behavior Targeting (& Two Prerequisites)

  6. […] – Avinash Kaushik: The Promise & Challenge of Behavior Targeting (& Two Prerequisites) […]

  7. […]
    Wow! What happened yesterday? Suddenly, it seems like every site I read is talking about behavioral targeting. In case you missed them:

    * Om Malik discusses Google’s plans for behavioral ads
    * Matthew Roche gives a general overview of optimizing behavioral targeting; and,
    * Avinash Kaushik provides two excellent pre-requisites before implementing behavioral targeting

    While one-to-one marketing has been oversold for years, behavioral targeting certainly makes a strong case for one-to-some. The "some" in this case is every segment you choose to develop creative for. The thing to bear in mind about behavioral targeting is this:

  8. […] The Promise & Challenge of Behavior Targeting (And Two Prerequisites) […]

  9. […]
    Another issue you have to be careful of is that behavioral targeting isn’t about brilliant segmentation, it’s about brilliantly relevant content. In this regard, we couldn’t have put it better than Avinash Kaushik of Occam’s Razor, who said “You feed your BT system crap and it will quickly and efficiently target crap to your customers. Faster then you could ever have yourself.”

Add your Perspective