As the programmatic campaign setup heavily relied on re-targeting, there was a clear need address any selection bias in the attribution model. Although a path-to-conversion analysis was conducted by an external company, when the programmatic bidding strategies are setup to aim for “the paths” with highest conversion probability this would naturally make retargeting appear to contribute more than what might be the case. Therefore, an experimental approach was initiated together with Integral Ads and their Causal Impact analysis.

The creativity of the “causal impact” approach derives from the use of an industry waste issue - from impressions that were paid for but never seen by intended consumers, the so-called unviewable ads. The analysis uses this industry wastage to construct a control group which replaces the need to run a randomized PSA (Public Service Announcement) banner or other A/B test.

This data analysis then provides valuable insights to the advertiser for deciding on the optimal media plan for maximizing incremental effect. And with this attribution in place, different types of re-targeting strategies could be evaluated and optimized toward maximizing the incremental lift.

On the first visit to the SAS site, customers generally provide a range of useful information (desired travel dates and destinations, for instance). To identify business travellers — SAS’s main target audience — Xaxis targeted the browser language for customers using a desktop OS. For instance, one strand of the campaign involved targeting customers in Germany whose browser language was set to Japanese. Users who didn’t convert on the first visit were segmented by time to desired departure. Based on historical data, a value was assigned to each segment (for example, travellers who wanted to depart within the next 24 hours had a higher average booking value than those planning to travel within the next seven days).

With this analysis, low-funnel retargeting was optimized against these search queries to maximize ROI while also using this data to excluded users where the departure date had passed to cut down wasted impressions for users that are not in the market. Due to the complexity of these rules, AppNexus’ Programmable Bidder was applied to upload a custom algorithm that captured all these insights.

The results met or exceeded the campaign’s targets. Based on internal numbers from SAS’s attribution data, Xaxis’ work with SAS was directly responsible for triple-digit increases in revenue over a three-to-four-month period.

Success metrics include:

• 203% increase in revenues

• 346% surge in marketing ROI

• Decreased cost of sales by 78%

• 3 times higher conversion rate due to development of custom algorithms

• Expanded activities to 16 markets using fewer personnel hours than was the case for previous campaigns, thanks to a centralised planning, buying and optimization process.

“For users that have been to our site and searched for a flight, say from London to New York, but then didn’t book, then we go out and target the user with relevant creative (with a New York background). Before, we only had 400 generic messages. Now we have 1.3 million. The results were amazing. We saw website revenue increase by 203%, return on investment increase by 346%. And we saw cost of sale decrease by 78%.”

Ian Björkdal, Digital Marketing Manager, SAS