Third-Party Data Works. It’s Just Not Priced Correctly

alan-pearlsteinData-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 Alan Pearlstein, president and CEO of Cross Pixel Inc.

The current pricing models for purchasing third-party data are far from optimal, and are actually holding back usage by advertisers.

Most data sellers and buyers use a cost-per-impression (CPM) pricing. My company, for example, is paid a CPM rate for every ad impression delivered to the audience we provide. While this model offers pricing certainty, it is highly problematic because it places a fixed value on an audience, regardless of the context and value of the media placement. The model assumes that all audiences are equal regardless of where the ad appears.

The reality is that reaching an audience with a display ad on Facebook (FBX) is not equal to reaching the same audience on, and the price of the media reflects this. How many ads are on each page? How long is the user on the page? Where is the ad placed? Is it below the fold? What is the content being engaged? All of these variables matter and are directly related to the value of the audience as well as the impression.

The CPM model is also holding back the growth of third-party data usage by the industry.  The biggest new supply of inventory in RTB is from FBX. CPMs on FBX average $0.45. Once you add in the approximate cost for your typical “in-market” audience data, which is about a $1 CPM, the price tag of the FBX media more than doubles. In order for the data to justify that extra cost on FBX, it would need to outperform advertising using no data by more than 200%.

I think we all know that this degree of lift is unlikely, and use of third-party data on FBX and other low-priced CPM media, such as Yahoo Mail, is very limited or nonexistent. The industry needs a new model so that valuable data can be utilized in all campaigns.

A New Model

A few DSPs and data sellers have started to address this issue by offering a new model that should be the industry standard: percentage-of-media pricing. Data is priced as a percentage of the CPM of the media. This model aligns data costs with media value and identifies the right value of the audience.

Display media, particularly in the RTB environment, is purchased via auctions, and has found its true value. With percentage-of-media pricing, data is priced accurately. More importantly, usage of data on low cost CPMs like FBX is now possible. When data is priced at 20% of the value of the media, third-party data works very well.

The broader adoption of percentage of media pricing for third-party data should lead to a significant increase in usage by advertisers, improved targeting and, most important, superior campaign performance. Every DSP and DMP needs to have the capability to manage percentage of media pricing for third-party data to work at scale.

Follow Alan Pearlstein (@alanpearlstein), Cross Pixel (@crosspix) and AdExchanger (@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. Ramsey M

    Given that data comes in many forms – offline, online, PII, non-, the value of the data is all over the place. In either case – CPM or % of media buy, the data company and media buyer are negotiating an effective price (either specifically or in aggregate). For example, if you license auto data, you may pay multiples of the media cost or you may buy it on a CPM and apply it both to low value and high value media. If you manage the overall data costs and media costs, does the % of media buy model change anything? Interested to hear more about how this changes the dynamics.

  2. Nitin Chowdhary

    I don’t agree with the logic of this argument.

    If anything, the focus of the audience data firms should be to make this data perform better i.e. increase value addition so that it creates noticeable lift so that the $1 CPM is justified. At the moment, the performance I see on a DSP shows hardly any lift between a untargeted campaign and a audience data targeted campaign. If, at a high level, the goal of this line item of the media plan is to generate top of the funnel interest in the product, without a performance lift there is hardly any logic to increasing cost even by 20%. I’d rather buy 20% extra impressions and depend upon the DSP to optimize click throughs or conversions.

  3. Couple things here …

    (1) The argument above is assuming an agency is purchasing 3rd party data + media through a Trading Desk. If the agency were to work with a data vendor who also sells media, rate-card pricing (that Trading Desks are held to …) wouldn’t apply. Therefore, data costs would be significantly lower, making the buy desirable and in-line with the Agencies goals.

    (2) Although the 20% data cost-to-media ration is one opinion, in reality, this would be very difficult to track through a DSP, and would require the Trading Desk to figure out. As of today, logistically, it would be a mess to sort out.

  4. @Ramsey M – thanks for the comment. You are certainly correct. Media buyers and data sellers are negotiating a rate, regardless of a CPM model or % of buy and a buyer can apply it accordingly to low and high priced media and manage the aggregate costs. The issue we are facing is that our DSP’s partners generally have you set one CPM rate for the audience data. Therefore we are never really getting a chance to negotiate a rate with the buyer based on the media that is being utilized. We are “stuck” with the predetermined average CPM rate, which may or may not be appropriate for the media type. In the end, we lose out on business (and advertisers don’t get to take advantage of the lift our data can bring) because we are not being utilized in many of the lower priced media campaigns because our average price is too expensive and cannot deliver a commensurate lift (particularly FBX only campaigns). This is what led me to suggesting a percentage of media pricing – it allows you to utilize a rate that can work for media of all different values when individual negotiations are not possible, as is the case with most of our partners where fixed rates are plugged into their systems.

    @Nitin it sounds like you have not have success with data, regardless of the pricing model. We have a long list of success stories at Cross Pixel regarding usage of our audience data. We would be happy to share them with you at your convenience.

  5. The first key point is that third party data is generally over-priced compared to the uplift it generates so the cost needs to come down.

    Then a % of media cost model for data introduces auction-based pricing for data by proxy. This incentivses data providers to sell valuable data for which advertisers will compete. Fixed price CPM models help the aggregators buy the data from the data-owners but they don’t let the market find the data’s true value efficiently. Until the market operates properly without this bias towards the seller third party data usage will continue to be limited.

  6. Great point here – if you price based on a % of the media, you can far more easily get lift.

    As a display DSP, we tested a lot of third-party data that provided lift, but not enough to justify its cost. Meaning, we were better off NOT targeting – this applied to just about every data provider we worked with because only one major one struck (what they described as a very non-standard) a % of media spend deal with us. And they told us they’d been burned before by at least one DSP who inaccurately reported their spend figures for that deal. So it’s understandable that people don’t want to rely on the partner’s reporting for figuring out how much money they made.

    The other point here is that some ad buyers just feel better knowing they have a target, even if the ROI is worse – we saw this a lot with buyers working with some of the BlueKai categories in their early days, pricing on a ‘cost per stamp’. The pricing was a multiple of the media costs, and yet because they could go to their boss/client and say “I’m targeting the People who eat Almonds category, aren’t you happier now?” even though they’d have been better off not buying the data at all from an ROI standpoint.

  7. Rob – it strikes me that generally data providers are subject to a partner’s reporting when using any delivery-based pricing model, whether CPM or rev share. The only data management platform I’m aware of that solves this problem is Google’s – the other DMPs all lag in this area, which is really a shame.

    As for ROI, it seems that when most people complain about data performance, what they really mean is that it didn’t improve CTR (a dubious performance metric at best). But even with a more appropriate performance metric, the challenge is that the methodology behind the data and the performance metric usually aren’t aligned. Pricing isn’t the challenge – the challenges are 1) picking a proper performance metric and 2) picking data which is specifically designed to optimize to that metric.