Paid Search: PPC DayParting and Time Lag Between Click And Order

Digging into the data is essential to paid search success. Wrote yesterday on some findings we've publicly released on our ongoing click-stream research. Wanted to mention another ongoing research project tracking the time lag between clicks and orders and strategies for "day-part" bidding (time of day, day of week, and seasonality). We spoke on this work last year at the San Jose SES, and have slides from that presentation online: The Click-To-Order Interval "Time Warp". That talk received some attention from the blogosphere and a writeup by Jakob Neilsen. The gist of that talk: While most order following a paid click come quickly, there's a long tail. 50% come within half an hour, 75% come within 25 hours, 90% come within 12 days, and 95% come within 4 weeks. Your mileage may vary -- these numbers are based on a random sample of clients, a mix of B2B and B2C, low-ticket and high-ticket retailers, etc. Even if the numbers shift a bit for your exact situation, the take-away message stays the same. It takes some waiting for those last few orders to straggle in. Larger ticket orders tend to have more delay, as to do purchases in more considered categories. The implications for PPC bid management are significant: it makes zero sense to compare click costs last Sunday afternoon to online sales last Sunday afternoon. Why? Because a meaningful fraction of last Sunday afternoon's revenues came from earlier clicks, clicks on Sunday morning, clicks on Saturday, and even some clicks on Friday night. Dayparting PPC bid algorithms need to differentiate between time-of-click and time-of-order, matching each order back to the click responsible for it. Bid algorithms need to do their work in click-times, not order-times. Big big big idea there. We see many (some leading) public bid management platforms out there blow it on this key issue. It is subtle, but it is also important. We're continuing our research work on the the click-to-order interval and dayparting algorithms, collaborating with some really solid academics over at Univ of Rochester. We'll likely share some of these more recent findings at the summer or fall SES.

Join the Discussion