We had a collection of hypothesis about how users would engage with the platform and one of the strongest signals we didn't anticipate was the time our sites were accessed.
Looking at our analytics we learned that while desktop did spend longer on a page, it wasn't by anywhere near as much as we predicted. The theory of users carefully researching menus before a visit failed and a new hypothesis was developed.
We also saw mobile traffic spike Friday to Sunday, quickly dropping off after 8 pm. We used Full story to unearth some of the behaviours and we discovered users were often looking for walk-in restaurants they were nearby. We developed some small tests to validate our rationale.
We also saw users spending more time, as expected, on items with photos. We launched a photo service for restaurants that were lacking quality images using this evidence and other data to demonstrate the need to restaurant owners. To this day menus with photos outperform those without in all our measures - they're a big deal.