Less is More: Pricing Your Way Out of Data Congestion

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

Today’s column is written by Omar Tawakol, CEO of BlueKai

Pricing a shared asset like data can be counterintuitive. Data is a non-rival good, like a toll road.  Even though people expect to pay for and share a toll road, if it gets too congested it becomes less useful. This is unlike a physical commodity, like a camera. If you and several others have the same camera, it doesn’t diminish the utility of the one you own.

The usual economics of pricing a commodity break apart in the world of data. A seller of a data asset wants the most yield from their data. The cost of distribution and replication of that asset for multiple buyers is near zero. So maximum yield usually occurs with lots of buyers paying the highest price possible. That much is obvious. The goals of the buyer, however, are more complex. A buyer wants data that performs well at a low price (these two goals in practice are contradictory). The cheaper a data asset is, the more it gets used.  The more a data asset gets used, the more it degrades in performance (like a traffic jam on a toll road).

As an example, let’s take a data attribute that is in clear demand and limited in supply – say, “luxury auto intender.”

When BMW is the only owner of the luxury auto intender attribute, lots of good things happen. However, when Mercedes and Lexus also have that attribute they start to bid for media against those users and the cost of the media rises. So even if the cost of the data is fixed, the bundled data and media cost becomes expensive due to bid-up. So too many data buyers is equivalent to having a toll road traffic jam.  Another cause of data congestion has to do with share of voice. Lets say “Jack” is about to buy a luxury car in the next week. If BMW were the only advertiser to Jack, BMW would dominate his attention. However, if Mercedes, Lexus and Cadillac all advertise to Jack, BMW would have a lower share of voice. Since Jack will only buy one car – the ROI for BMW decreases due to share of voice competition. The effect of share of voice competition on performance isn’t as direct as the bid-up effect on media because there are other ways to reach Jack (like contextual targeting) but it does have an adverse effect. Taking this all into consideration, BMW wants the price of the data to cause all other auto marketers to not use the data. Fewer buyers means more value for BMW (less is more).

The wrinkle in all of this is that one way to get fewer buyers is to raise the price of the data so that BMW buys it and others don’t. Obviously this has a limit. No one wants to spend their whole paycheck on a toll road, and a data budget is limited by the ROI of the campaign. There is an equilibrium price where incremental performance improvements from scarcity don’t justify the incremental rise in price. The more you raise the price of the data, the higher your share of voice and the less you have to bid against others in the media. In other words you will pay just enough on the toll road to get rid of the traffic jam – but no more.

Data brokers first became aware of this phenomenon when buyers would ask them to lower the price in order to hit a certain performance goal. Over time however, the lowered price wouldn’t satisfy their performance goal because the data was now getting more crowded and overused. What the buyer really meant was lower the price for me – but raise it for others. This would generally not be fair. But there is a way to get this to be fair and that is through forms of exclusivity.

Lets take the most common form of exclusivity — retargeting. Retargeting works well for several reasons – but one fundamental reason is the exclusivity of that data asset. It is like a highway built only for you. In general you bear a high cost to build the retargeting asset – but that is a sunk cost and is never factored into the cost of the media targeting. It is hard to compete with the perception of free. The problem with trying to seek exclusivity for a data asset that isn’t created by you — is that the seller can rarely justify selling a data asset to only one buyer. A single buyer would have to pay for the utility of all potential buyers in order to contractually get that seller to sell the asset exclusivity. That is possible but unlikely.

There is another way to approximate exclusivity. That comes from a look-a-like model. If a look-a-like model is built only for you off your exclusive retargeting asset – the data product is unique and can uniquely be bought by a single buyer. It isn’t perfectly unique because other look-a-like outputs could be partially overlapping with yours but it is more unique than a generic data asset that everyone has access to. A look-a-like model resembles a self-portrait. You may produce several paintings but none of them will look like your self-portrait and your self-portrait isn’t an exact replica of you. That means a look-a-like model doesn’t obey the same economics of a pure non-rival good.

As I said at the outset, the pricing of a non-rival good like data can be very counter-intuitive. Do you have any other examples of the upside down nature of data pricing? I would love to hear from you.

Follow Omar Tawakol (@otawakol) and AdExchanger (@adexchanger). 

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. Omar –

    Great article, but I think you missed one key point. In the world of RTB, we’re dealing with a finite amount of cookies, and therefore a finite universe of consumers that can be targeted with an ads (from any category of advertiser). Even if a user falls into an “exclusive” bucket of data based on a 1st party relationship, that does not exclude them from also belonging to other “non-exclusive” segments from other data providers, etc…

    While data exclusivity may dampen the affects of a crowded data marketplace, I do not think it solves the problem. As more and more bidders join the exchanges, the “bid density” problem only increases – especially when you consider that the number of “addressable cookies” will stay relatively flat.

    Data providers and advertisers a like have to realize that this finite universe of data will always lead to many bidders trying to target the same consumer – regardless of if they have “exclusive” data on this user or not.

    • Kyle-

      You are absolutely right. There are second order effects caused by out-of-category data. If we build on the example above, even if BMW had exclusive access to the data they would still have to compete with United, Chase, and Apple – which would all be competing for the same person who happens to be a luxury goods buyer. I call it a second order effect, because the in-category data is more reliable in its ability to concentrate competition across all the users in the category than out-of category data. Thanks for adding the insight.


    • Kyle,

      There may be finite cookies, but currently impressions on RTB world is nearly limitless given the newly opened FBX. I wonder how competitive the media RTB auction in USA is. If not that competitive, media bidding engines have many opportunities to cherrypick some specific cookies multiple times.