Cracking The Fractional Attribution Conundrum

marc-grabowski-better“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 Marc Grabowski, chief operating officer at Nanigans.

When I approached the register last weekend at a local Best Buy to purchase an iPhone charger, I was asked if anyone had helped me with my shopping. I said yes, even though the young man who assisted me did nothing more than point me in the general direction of iPhone accessories.

I assume he received credit for my purchase although he didn’t do much to drive that revenue. He didn’t create the need for the item – this was my doing since I left my charger in the office – nor did he create mindshare around the Best Buy stores.

This mirrors the global conundrum of attribution in programmatic media. The attribution challenge on digital channels has the most complexity, largely due to the measurable nature of the media.

Marketers must now rethink the “last-touch” or “last-click” model that was accepted not long ago and consider the confluence of events that create buyers and lifelong customers. Which events were responsible for their buyers?  How would the absence of specific events impact the face of their business? Savvy marketers also must recognize that attribution is more than tracking and must be applied to optimize future spending.

This leads us to a range of questions we’ll explore in this two-part series as I aim to unpackage and shed light on the discrete value of events that take place throughout the attribution chain.

First, let’s look at the attribution types and terms that float around in the market:

Last-Click Attribution

This model gives 100% of the credit to the click closest in time to the intended action, such as purchase, lead form, etc. This is closest to the Best Buy example above as it doesn’t account for any other experiences the customer had prior to this click.  Advertisers normally allow an acceptable window of time between click and conversion, such as seven days, and some even allow 30 days depending on the length of their consideration cycle.

Last-View Attribution

This gives credit to ad exposure whether or not the user clicked or interacted with the ad. There is a big push for ad viewability, but this still assumes the user recognized the ad and it changed the way that person thinks about a brand or product. There are methods to test impact on view-based attribution, but that is another topic for another day.

Normally this attribution type is given a shorter window than click-based methodology – two days or only several hours.

Fractional Attribution

This model gives some attribution to multiple items. Marketers employing this attribution recognize that buyers are created as a result of a number of events in the consumer decision-making journey. The methodology attempts to place value across events in such a way that each event is given appropriate credit for the ultimate action.

Multitouch, multichannel or full-funnel are all forms of fractional attribution that attempt to assess value of each impression and separate correlation of exposure from causality of an event.

Note that I did not include first-click attribution in the breakdown of attribution models. Please, by all means, feel free to let us know in the comments if you believe this model is still relevant to the attribution conversation.

Food For Thought

Before we continue the attribution conversation in next month’s column, let’s consider several questions:

1. Is your desired action a one-time purchase or are purchases linked to prior purchases?  If purchases are linked, how do you attribute weighting to previous purchases?

2. Which channels are most meaningful to your users, and does channel importance differ by user segment? In other words, does one audience trust a source more than another audience? Do younger members of your target audience care more about social context vs. older segments relying more heavily on established content originators?

3. Is some of the data used to target good users gathered in different stages of the campaign? Do you attribute the data-gathering phase to the conversion? This comes up most frequently with retargeting: What credit does the impression that drove that user into your cookie pool receive? Can I get to the bottom of the funnel without the work done at the top of the funnel?

I’d love to hear your feedback and thoughts on these questions in the comments section. We’ll discuss in next month’s column, and also test frameworks to better understand the value of events in the attribution chain.

Follow Marc Grabowski (@MarcTGrabowski) 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. The attribution problem is definitely fragmented. What is your thoughts on ad tech companies who cookie stuff in order to get that last click through?

  2. Grabo- thanks for tackling this subject. it’s hugely important because advertisers and agencies that ignore their measurement methodology are at risk for rewarding media companies for bad behavior (cheap imps, below the fold, e.g.)

    I’d like to add the following to your list: incremental lift analysis. All of the attribution models listed above neglect to address what’s known in statistics as “confounding factors”. Simply put, it’s impossible to go back in time to see what would have happened had the user been served no ads at all. Would the user have converted anyway? The only way to solve for this is through control and exposed test methodology.

    Control and exposed testing is extremely complicated when advertisers work with multiple vendors across multiple platforms, but there are statistical methodologies around this (see causal inference modeling).

  3. Agree on all points and Jeremy’s comments. To add, retargeting based attribution (eg last click) and aggressive CPA goals set at retargeting levels often persist throughout all advertiser display tactics including modeling and acquisition. This leads to the common effect of optmizing out the non-retargeting tactics and re-attributing the benefits of these tactics disproportionately to retargeting. The net net result is that I hear advertisers complain that while retargeting is great – that it only represents 20% of their monthly traffic and delivery/scale are challenged. More often than not, if I dive into attribution, goals and tactics, I find that simple retargeting is strangling out very valuable tactics that feed the top, middle of the funnel. Fractional attribution is a practical way to open the conversation but as Grabo stated can be used in conversion event analysis or LTV analysis. The latter tends to weight acquisition/mid/upper funnel strategies more equitably. I find looking at LTV returns better YOY growth for the full funnel.

  4. Upvote for Tony and Jeremy’s comments/praise. I would also add in that you should never underestimate the TV Broadcast campaign that ‘may’ be running in tandem.

    You can often see that both paid search and re-marketing are heavily influenced by that upper funnel TV spend so those channels (remarketing and PPC) actually have a much higher actual true-cost once attribution covers that TV activity also. VisualIQ have a good study on this here: