The Official Blog of Cogmap, the Org Chart Wiki



Lookery in the Deadpool – I say “I told you so”, gently


Hmmm, I can’t seem to find my “behavioral data sales exchanges are doomed” blog post, but I discussed it with hundreds of people and I referred to it in this post, where I tried to make people feel better after a particularly harsh anti-behavior diatribe.

Anyway, here was my theory:

Data is only useful if you see tons and tons of impressions to drive targeting off of.  Monetizing a person requires a data point the advertiser values and an impression where the advertiser sees the value and the value of the data is limited to the difference between what you could have charged for that inventory and what the data makes it worth.  Anything other than conquesting automotive behaviors rarely bridges that gap enough to justify the expense.

The problem with a data exchange is that you gather tons and tons of data, dropping pixels everywhere, with little hope of seeing that person again and not a lot of hope of an advertiser buying that behavior.

Ad networks were able to be successful because, at the very least, they could take every impression they bought and show some crazy CPA ad.  In the data exchange business, you drop tons of pixels and never do anything with it again.

So here is my business model:

  • Ad network was going to sell 1,000 impressions for $1.00.
  • Data exchange provides behavioral data that allows the impressions to be sold for $2.00.
  • To generate the 1,000 behavioral data points, they drop 3,000 pixels on the 1,000 users (3x frequency – for fun – let’s assume that makes it a good behavior as opposed to a random behavior).
  • To get 1,000 users, they pixel 10,000 users – not every person pixeled gets seen for the campaign served, so you have to get a lot of “qualified users”.
  • So they dropped 30,000 pixels at a cost of $0.01 CPM – so $0.30.
  • The ad network wants to keep some of the upside (otherwise why sell the behavior), so they keep half the upside and give half to the behavioral data exchange.
  • So the behavioral data exchange just made $0.50 and it cost them $0.30 – BUT, they have to pay the behavior provider, and he wants 50%.
  • So every transaction costs the data exchange $0.05.

Now, if they lifted the value of the inventory more, they might make a little money.  But, if they have to drop more pixels because network reach is poor, then it costs more.

AND HERE IS THE KILLER: If they drop a bunch of pixels and no one buys that behavior, they just lost all that money.

So data exchanges drop tons of pixels and roam around trying to convince someone to buy that behavior.  Every second that passes, behaviors are losing value and more cookies must be dropped to replace people falling out of populations.  Can you really dig your way out of that?

So Lookery just went belly-up.  And Scott and Todd are smart.  (Although I think I told Todd all this stuff some time ago.)

Were they too early?  Maybe.  The creation of real real-time-bidding interfaces will allow people to cherry pick bigger networks, maybe creating a better opportunity for data sales.  Good enough to build a viable business?  Not sure.

Hmmm, I suppose Exelate is next?

5 Responses to “Lookery in the Deadpool – I say “I told you so”, gently”

  1. Greg Says:

    You didn’t need to make me feel better after your diatribe – after all, I had already scored some free schwag! That mug still graces my desk at world headquarters.

    Paying for your behavioral data doesn’t make a lot of sense to me – just offer some minimal sort of service and the publisher-side will give it to you for nothing. Don’t know for sure, but I’ve got a strong suspicion that’s what’s up to. (Or perhaps that’s just what I’d do with it.)

  2. Scott Rafer Says:

    What you’ve proved is that we’re early. We agree. Data exchanges can not exist independent of media exchanges. That’s not what we do/did.

    We make it easy for people to store and distribute their proprietary targeting sets. If you have a few million high-value profiles, how do you sell them to the specific customers who care and deliver them appropriately? If you want to cherrypick impressions off of RMX without dumping your data into their system and losing control over it, where’s the Profile SaaS that can handle the volume and latency of an exchange?

    That’s the slot we’ve tried to fill.

  3. Todd Sawicki Says:

    Scott and I apparently weren’t smart enough. To Scott’s point – Lookery wasn’t in the data exchange business and we had lots of concerns about the model so you won’t hear me argue against your point.

    A few points. Careful not to assume that data is going to come from ad networks – in fact the reality is that valuable data is going to come from publishers. I think we completely agree that ad networks are poor sources of data and thus would even be worse resellers of data.

    Thus if its coming from publishers it isn’t really behavioral the way that is often wrought with concern – the idea of tracking and mapping user’s online browsing across many websites.

    There are really two types of valuable targeting data – retargeting and profiles. Both have shown to provide demonstrable lift in CTR’s and conversion. Retargeting data is hot and as such acerno and blue kai are doing much better then Lookery. But long term, the idea marrying user information from offline and online profiles becomes another very attractive targeting option. Lookery enabled the later and as Scott has noted we were too early.

  4. brent Says:

    I assume the data is coming from publishers. They get that data payout in my model and some data provider payout happens anyway you slice it. I am assuming networks sell the ad space to the advertiser, using the data provided. Even marrying data up basically fits the same financial model I describe. Unless you own the ad space, it is just OK, financially.

    Retargeting is interesting and I haven’t followed Blue Kai closely, but I have always assumed that there is not enough data to build for truly monstrous scale. Maybe coming from a place that already has huge reach, retargeting data exchanges don’t seem particularly valuable.

  5. brent Says:

    Hmmm, incidentally, in response to Scott’s comments, I don’t think I actually proved much of anything. Scott and Todd proved something. Easy to sit on the sidelines and throw stones! The only thing that would have been nicer than my hypothesis being true would have been my hypothesis being wrong, because it would be great if data-layering added so much value that people could be exclusively in the data business. What a fun new twist that would introduce to our industry.