Norway is currently a tough nut to crack for car importers. Varied geography, financial incentives and other demographic variables affect the behavior of Norwegians when it comes to car purchases. Particularly financial incentives are linked to the car you drive, which in turn is influenced by where in the country you live. It is therefore challenging for car importers to invest their market resources so that they hit potential customers that are most relevant without going into the troublesome retargeting trap, which has also become even more limited since GDPR came into force in summer 2018.
The challenge: The danger of optimizing to death, in a "last-click" world
In most cases, targeting, optimization, and bidding strategies are built based on the ad buying platform's self-generated campaign data. These are often based on "last-click" conversion data, which excludes the effect created by other channels earlier in the customer journey. A good example of this is the Facebook algorithms, which work towards the best possible results by doing more of what has previously done well on Facebook.
The same principle applies to other social media, and programmatic purchases of display or video ads. Over time, the purchasing algorithms within each purchasing platform become narrower and narrower, with the risk of losing track and potentially excluding highly relevant users who do not match a predefined or historical behavioral profile. The user pool hit by the ads quickly decreases, the exposure rate increases and so does the irritation. Poor user mood probably also means bad business for you as an advertiser.
It is therefore challenging for car importers to invest their market resources so that they hit potential customers that are most relevant without going into the troublesome retargeting trap, which has also become even more limited since GDPR came into force in summer 2018.
The Norwegian Mindshare and Xaxis team came up with the hypothesis that we could leverage Ford's website and behavioral data in a more holistic way, and possibly improve programmatic purchase of display ads. The aim was to increase the impact of digital investments, based on a broader and more realistic data base.
- More broadly, because we chose to spend one full year of anonymous behavioral data from Ford's website, regardless of the original traffic source (organic, search, SoMe, display, native, referral, etc.).
- More realistic, because the database reflects long-term interest patterns from the market, both paid and organic.
One of the challenges with this working method was that we plow our way through just under 10 million data points to find the answers we were looking for. Enter IBM Watson; an artificial intelligence tool. After a data cleanup, feed and analysis in Watson, the team was able to identify clear patterns per (item) car model, depending on geography, day, time of day, type of conversion, etc. These were used to build model-specific purchase algorithms, which were significantly different from the previously used media-centric capabilities that the buying platforms offer: domain targeting, formats, viewability, frequency, time management, recency, etc.
The result: 45% lower cost per conversion
Updated "rule sets" were implemented in Xaxis' AI-based purchasing solution Copilot. These have exceeded all expectations.
Utilizing free and GDPR-secure first-party data, along with a more holistic approach to programmatic advertising, is proving to provide better impact per invested dollar than other ways of data enriching or optimizing purchases. Media CPIs (CPM, CPC, CTR) were significantly improved, but the clearest effect is seen in a spectacular decline in the cost-per-conversion business-critical CPI. The decline, versus previous comparable campaigns, is between 25 and 75%, depending on the model you are advertising. Across all campaigns, cost per conversion is down 45%. On an annual basis, this represents a huge potential for added value, with a lower risk of segmenting themselves to death and of bothering people with advertising for cars that are not relevant.
This way of informing programmatic purchases, based on first-party data, has now become the basis for all model-specific campaigns - with great success and returns for Ford Motor Norway. Following a successful implementation in the Netherlands, the methodology is now also being scaled to several countries in Europe - which will potentially increase returns by many millions of Ford advertising crowns in European context.
This case has won several awards
INMAS PERFORMANCE AWARD 2019:
Sølv i kategorien "Online Performance"
DIGIDAY GLOBAL TECH AWARD 2019:
Gull i kategorien "Best Personalization & A/B Testing Platform"
DIGIDAY MARKETING AND ADVERTISING AWARDS EUROPE 2019:
Gull i kategorien "Best use of AI"
Read the original article in Norwegian on Kampanje
