Attribution Needs To Walk Before It Runs

david-dowhan"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 David Dowhan, president at TruSignal.

Today, most trackable human behavior occurs online in a digital context. Much of this data may be the raw material used in an attribution model. Using cookies or device IDs, brands can market directly to a single consumer, targeting their messages with unprecedented precision.

Marketers have become so accustomed to this level of detail that the possibility of mapping the path to purchase end-to-end seems seductively within reach. This has led to the expectation of measuring the incremental value of each impression with the same precision with which we targeted them.

Brands have tried to understand the influence of marketing on sales and measure complex media mix models since before the Internet. However, brands have never been able to attribute the impact of multimedia campaigns with 100% accuracy.

It’s not hard to understand the desire for a fully mature attribution model. But the truth is that we are quite a ways from that reality.

What Marketers Get Wrong About Attribution

I see several mistakes that marketers make to perpetuate attribution challenges. First and foremost, they are addicted to the last-click or last-touch attribution metric. These methods give all credit to the click or impression that immediately preceded conversion, while ignoring the six or more impressions before it. Last-click attribution is easy to track, but it oversimplifies the situation and can be gamed easily. Most marketers agree that last-click is not the right metric, but in many cases it’s the only one they have.

Marketers also miss offline sales and baselines. Multitouch attribution models attempt to weigh each impression, based on its influence to a sale, within a certain window of time. But there is often not enough data to calibrate an empirical model, so marketers have to create artificial rules and algorithm weights based upon gut, instead of hard data. Finally, many multitouch attribution solutions mostly track online exposures, actions and sales, thus missing up to 90% of offline sales.

Another challenge is failure to establish a baseline of sales that would organically occur without marketing, making it impossible to calculate any sort of incrementality. The baseline is especially important for companies that have consumers with repeat purchases, such as consumer packaged goods or apparel.

Finally, marketers use arbitrarily short time limits. Whether it's linear, time-decaying, view-through or click-through, all current attribution models are constrained by the arbitrary time limit put on them. And those time limits are typically too short, which creates a bias of recency over true impact.

Who’s to say that ads are only effective if someone purchased within three to seven days, or even 14? If we take a longer view of the consumer’s actions, we can see that advertising may have an impact weeks or even months beyond the period that most marketers consider relevant.

What Marketers Can Do To Get It Right

So what should marketers do? Several things can help them improve the accuracy of their attribution measurement in the near future.

The first area of improvement is expanding beyond bottom-funnel metrics. The industry has historically fixated on the bottom funnel and bottom-funnel metrics that lead to a sale – last-click, last-touch or view-through. While these are important, branding and upper-funnel marketing to create awareness and drive consideration are also important. The next generation of attribution needs to understand the different stages, objectives and metrics of campaigns as they fit into the full path to purchase. Until we become proficient at mapping each stage and its relationship to sales, a fully closed-loop attribution model is unrealistic.

The second area looks to the past: the media mix model. A key component to media mix models is the development of a baseline. What actions do consumers take organically that marketing should not claim? For example, I may need to buy jeans because my old ones don’t fit or they're outdated, as my wife would politely say. I would choose to buy my new jeans at the store where I purchased the old pair out of habit, not because of the company's TV, search or display advertising. My purchase cannot be attributed to marketing efforts, which is evidence that baselines are requisite to truly accurate attribution.

Lastly, as we expand beyond the immediate bottom-funnel focus and create a baseline, we also need to expand our time limits. Most metrics today use a three- or seven-day window, even though the path to purchase may take weeks, if not months. Marketers should look at the full sales cycle and the lifetime value of the consumer.

The consumer’s path to purchase is long and nonlinear. If a consumer reads an article that leads to a video, before searching for the brand and buying a product weeks later, did the article, video or search drive the sale?

Running to a multitouch attribution solution without fixing the current challenges will perpetuate confusion and unrealistic standards among marketers.

Follow TruSignal (@TruSignal) and AdExchanger (@adexchanger) on Twitter.


  1. Hi David, solid headline. Attribution is a worthy concept, but its implementation is typically flawed to such a degree that it can drive negative value.

  2. David, I agree that full-funnel attribution is where we need to focus. The marketplace currently seems confused about what quality attribution is. Instead of only offering vanity metrics, sales attribution platforms in the market today can and do deepen the understanding of conversion path to purchase. Beyond baseline products like first-click or last-click attribution, there are technologies that allow us to transparently view a consumer’s full path, sorting and ranking the true influence of each channel’s contribution – paid search, email, third party websites, display ads, organic search or brand sites. Frankly, I think marketers need to add a line-item in their budgets for analytics. If you're not measuring, your competition is going to eat your lunch.


Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>