Henning Hohnhold

Who they are

The sources read for this page establish Hohnhold as a Google researcher at the time of the paper, since it is hosted on Google's research publications site, and as the first-named author of the long-term ads study. They do not give a biography, and this page does not supply one.

The paper is internal research on a commercial product. The authors are Google researchers writing about Google search ads, which is useful to know when weighing it. Insider access is what allowed the long-term question to be studied at scale. The same position means the work reports from inside the business whose revenue it discusses.

Credit follows the byline. Hohnhold is first author, and O'Brien and Tang are co-authors. The Why We Project cites the paper as Hohnhold, O'Brien and Tang, and names Hohnhold as the lead.

Key papers

Hohnhold, H., O'Brien, D., & Tang, D. (2015). 'Focus on the long-term: It's better for users and business.' Proceedings of the 21st ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Sydney.

The paper argues that short-term A/B test results do not always predict long-term effects, because users change their behaviour after a product launches. The authors develop a method for measuring long-term user learning and apply it to ads on Google Search. They study ads blindness and ads sightedness, meaning changes in how likely users are to click on ads, and report that these changes are driven by the quality of the ads previously seen or clicked, including relevance and landing page quality.

They build a model that uses short-term metrics to predict long-term outcomes. They report two applications: a change to the search ads auction that increased the importance of ad quality, and a 50 per cent reduction in ad load on Google's mobile search interface. In summary the authors write that user satisfaction is paramount.

Only the abstract-level description and publisher page were read for this page.

A reader should hold the paper to its own claims. It argues that short-term A/B results do not always predict long-term effects. It does not argue that they never do, and its model exists because short-term metrics carry some information about the long run. The two applications it reports show the method being used to decide a real question: how much weight to give ad quality in an auction, and how many ads to show on a mobile results page. The sizes of the results are in the paper, and this page does not repeat them.

Key ideas

The first idea is that an experiment measures a moment while a product lives in a relationship. Users who meet poor ads adjust, and the adjustment appears after the test has ended. The paper gives that adjustment names, Ads Blindness and ads sightedness.

The second is that quality, not volume alone, drives the long-run response. The factors the authors identify are the relevance of the ads and the quality of their landing pages. This fits the title's claim that a long-term focus is better for users and for the business.

The third is methodological. The authors do not discard short-term metrics. They build a model that uses them to predict long-term outcomes, which treats the short-term number as an input to a forecast and not as the final answer.

For readers of the Why We Project, the paper is the clearest published case in which a company reports that its own short-term count could point away from the long-term result. It sits beside Good Abandonment and the eBay field experiments of Blake, Nosko and Tadelis as evidence on what Engagement Metrics leave out. That grouping is our reading.

---