Adference’s session at the “Google Digital Workshop on Campus” this Friday was streamed live from Stanford University where Adference co-founder Burkhardt Funk is active as a visiting scholar this semester. The professor opened his session explaining one of the first things he learnt about presentations in Silicon Valley: Pictures matter. Photos of Stanford in the golden hills of California, Google cars and Larry Page’s palm tree decorated with Christmas lights took the workshop’s participants from Leuphana to the Valley for a brief moment.
Between virtual reality and space-tech, Stanford University is one of the leading research institutions in the area of data science that investigates search engine advertising (SEA). In the workshop, Professor Funk brings together science and online marketing practice.
Why is search engine advertising of scientific interest?
- Search is a playground for economists, as millions of auctions that can be measured and observed take place. An overwhelming amount of information is readily available for statistical analysis – this differentiates Search from other areas in online marketing
- On the one hand, data from SEA is used to test and develop hypotheses on auctioning processes, developments on the market, purchasing behavior and more – on the other hand, it is also used to develop marketing strategies for individual advertisers
Which decisions are made in SEA?
- Which bid should I place for which keyword? Which keywords will help me show my ads to users that are ready to buy? Which copy texts, account structures, adjustments for device types, regions and budgets will work?
How does science offer support?
- One of Burkhardt Funk’s work groups – SEM-A² – researched ways to improve auctioning processes. One hypothetical approach of the group involved lowering the costs of SEA auctions by aggregating a number of accounts. At the moment, every auction participant is encouraged to place the maximum bid that they are truly ready to pay. With linked accounts, advertisers could outbid other participants only slightly. The price span between all ad ranks would become much smaller and costs for SEA could be reduced.
- However, another technique developed by the team proved more practical – namely, the statistical prediction of conversions on the basis of historical data. These predictions are used systematically to increase or decrease keyword bids.
- Problems in the longtail: Low conversion rates make predictions unsafe. Statistic measures to increase the data basis include the definition of similarities between keywords. Similar keywords that are similar in conversion rate, position in the campaign hierarchy, user behavior and the like are grouped together in clusters.
- The founding idea of Adference – a Leuphana spin-off – arose from these findings. The team was able to use their results to develop a tool that helps online advertisers manage Google AdWords and Google Shopping campaigns automatically. Co-founder Tobias Blask joined the workshop to introduce Adference and their journey from research project to company.
Google organized the “Digital Workshop on Campus” at Leuphana University for participants interested in improving their skills in social media, mobile marketing and search engine advertising. The training was directed at students, alumni and local businesses.