How does conversion lag mess up smart bidding, and what do I do about it?
Lag is the time between the click and the moment the conversion is recorded. The bidder sees an incomplete picture of recent days, so recent performance always looks worse than it is, and people react to a shortfall that fills in later. Long lag also slows learning. Import conversions daily rather than weekly, and never judge a period shorter than your lag.
Two kinds of lag, and only one is natural
Natural lag is the customer's behavior: they clicked Tuesday, called Thursday, booked the following week. You cannot remove it, only measure it.
Manufactured lag is yours: a weekly export from the field service system, a nightly job that fails silently, an offline conversion import that runs on Mondays. Manufactured lag is often larger than the natural kind, and it is entirely fixable. Moving an import from weekly to daily is usually the single highest-value change available to a service account running revenue-based bidding.
Measure your own lag before you tune anything
Pull the distribution of days between click and recorded conversion, not the average. Averages hide the tail that actually causes the trouble. What you want is the day count by which most conversions have landed, and the day count by which nearly all of them have.
Those two numbers become your reporting rules. The first is the earliest you can look at a period at all. The second is when a period is closed and can be compared to another closed period. Write them down, because everyone in the business will otherwise keep asking about yesterday.
What lag does to the bidder
Smart bidding learns from completed conversion data. Long lag means the freshest signal is the thinnest signal, so the model leans on older data and responds slowly to real changes — a new competitor, a price change, a service line you just launched.
It also interacts badly with sparse conversions. A campaign with few conversions and a long lag can spend most of its life in an unsettled state. That combination is the strongest argument for optimizing toward an earlier event and reporting on the later one in your own revenue reporting.
The practical compromise
- Bid on the fast event, report on the slow one. Use the qualified lead or booked appointment as the primary conversion and keep completed revenue as a reporting metric imported on its own schedule.
- Import daily, backdated to the click. Offline imports credit the original click date, so a daily cadence recovers most of the manufactured lag without changing anything else.
- Set a review calendar, not a review habit. Weekly reviews of a business with a two-week lag produce two weeks of panic followed by two weeks of relief, every month, forever.
Topics: conversion lag · smart bidding · offline conversions · reporting
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.