An ever-greater challenge for anyone working with digital communication is how to measure the total effect of digital campaigns.

Most people working with digital media today agree that “last click” or “last interaction” methodology is inadequate for measuring the total effect of digital channels. Still, these are the standard methods applied for evaluating digital attribution to effect.

The media channels themselves are partly to blame. For years the industry has promoted digital media as the most measurable of all media channels, defining the click as one of the most important KPIs.

The problem with “last click” and “last interaction” methodology is their inability to measure the attribution of all impressions and interactions throughout the consumer journey. They only give a snap shot of the last part of a process, instead of showing the full picture.

This is in particular a problem for media selling display ads, since display’s role in the consumer journey is getting attention to and creating interest for the product or service at an early stage of the consumer journey.

Today’s standard effect measurement methods disregard the value of any unclicked display ad, taking into account only the final activities — activities that benefit from awareness and brand building created earlier in the campaign.

Scope

Norway’s BN Bank, represented by Marketing Manager Torstein Torjuul, wanted to get a better understanding of the digital consumer journey.

Together with their media agency Mediacom, they joined forces with Storby (a media sales cooperation between Aftenposten, Bergens Tidende, Stavanger Aftenblad, Adresseavisen, and Fædrelandsvennen) and Business Science Nordic (BSN), a leading provider of sales modeling and conjoint analysis in the Nordic region.

Our scope was to analyse the digital consumer paths leading from exposure/interaction to lead generation, with particular attention to the display ad’s role in the process.

Method

BSN’s attribution model distributes effect to all interactions during the accumulated consumer journey, thus taking into account the impact of impressions that led up to the final click or interaction (assisted leads; Fig.1).

We applied BSN’s path-to-conversion analysis to BN Bank’s spring campaign, where the goal was to invite people to switch their mortgage to BN Bank (Fig.2).

The consumer journeys were tracked and accumulated using cookie data from all exposures and interactions connected to the campaign, during the five-week campaign period. In addition, we continued the tracking for five more weeks to take into account the long-term effects.

In total, more than 85,000 unique consumer journeys were recorded across display, pay-per-click (Google), organic search, and owned media (home page, Facebook page, and CRM-activities) (Fig.3).

The algorithmic attribution model was then applied to this data, to distribute the effect of all interactions leading up to a conversion (people who applied for a mortgage online).

The model’s ability to predict a conversion or non-conversion based on the consumer journeys was tested, showing a 99.6% accuracy in predicting conversions and 95.9% in predicting non-conversions — an overall accuracy of 96% on all exposures!

Results

The analysis of BN Bank’s campaign confirms the fact that last-click methodology is an inadequate method for measuring digital effect. Comparing the results from BSN’s attribution model with a last-click effect model showed that display media were underestimated by 35% (Fig.4), and compared to a last-interaction model, by 58%!

We are very pleased to being able to demonstrate the value of digital interactions taking place before a purchase is made and the huge impact that display ads have in the long process leading from first impression to a purchase or lead.

To quote Erik Kristiansen, digital manager of Mediacom Norway: “We were very pleased working with Storby in this campaign. By using attribution modeling, we can estimate which channels have contributed to effect for BN Bank with great certainty. The application of modeling removes the need for last-click as an evaluation method — a method we all know gives a skewed impression of the consumer journey.”

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .