Lookback Windows: The Quiet Setting That Changes Who Is in Every Segment
Two analysts build a lapsed buyers segment from the same database and get audiences that differ by half. Neither made an error. One looked back twelve months for the last purchase, the other looked back ninety days. The lookback window, the period a segment examines, is the least discussed and most consequential setting in segmentation. It decides who exists in your audiences before any clever criteria get a say.
What the window does to common segments
A high value segment over a 90 day window finds seasonal whales and misses the annual bulk buyer entirely. Over twelve months it finds both, mixed with people whose circumstances changed a year ago. Neither list is wrong. They answer different questions, and campaigns fail when the window answers a question the campaign never asked.
Recency segments are pure window logic. Recently purchased in 30 days means something precise only if your category's normal repeat cycle is longer than 30 days. In a category bought weekly, 30 days is an eternity containing five purchases. In furniture, it is a rounding error after one. Windows must be set from the category's buying rhythm, not copied from a template's defaults.
Matching window length to the decision
Short windows suit fast signals: current intent, in season interest, cart and browse behaviour whose value evaporates in weeks. Long windows suit identity level facts: category loyalty, lifetime value tiers, replenishment cycles for slow consumables. A common failure is measuring a long cycle behaviour in a short window and concluding the behaviour does not exist, when the window simply never contained it.
Write the window into the segment's name and definition, where a successor can see it. Lapsed 180 no purchase 12m lookback is a definition another analyst can reproduce. Lapsed buyers is an argument waiting to happen in a meeting nine months from now.
Rolling versus fixed windows
A rolling window moves with the calendar, always examining the last N days. Membership churns daily as behaviour ages out, which suits always on campaigns. A fixed window examines a set period, a season, a sale event, and freezes the audience, which suits analysis and one off campaigns comparing like with like.
Trouble comes from mixing them unknowingly: a report pulled monthly from a rolling segment appears to show audience growth or decline that is really just the window sliding over a seasonal peak. Before interpreting any segment trend, ask what the window slid over. Half of segmentation mysteries end at that question.
Stacking windows for a fuller picture
Sophisticated definitions use two windows deliberately: purchased in the category within twelve months, and browsed in the last fourteen days. The long window establishes membership in the world, the short one establishes current intent, and the combination targets people who are both relevant and awake. That pattern, identity from the long window, timing from the short one, resolves most arguments about which single window is correct. The correct answer was usually both, doing different jobs.
Keep reading
Related guides on this site: Segment Entry and Exit Rules: Audiences Should Expire as Well as Accumulate Household Segments: When Three Customers in Your File Share One Kitchen Practical answers about personalization.
Frequently asked questions
What is a lookback window in segmentation?
The period of time a segment definition examines when deciding membership, such as purchases in the last 90 days or the last twelve months. Changing the window changes the audience, often dramatically, even when every other criterion stays the same.
How long should a lookback window be?
Long enough to contain the behaviour's normal cycle in your category. Fast repeat categories need short windows; slow, considered purchases need long ones. Set the window from observed buying intervals in your data rather than from platform defaults.
What is the difference between rolling and fixed windows?
A rolling window always covers the most recent N days, so membership changes daily. A fixed window covers a specific past period and stays frozen. Rolling suits ongoing campaigns; fixed suits comparisons and event based analysis.
Can a segment use two lookback windows?
Yes, and it often should. A long window can establish lasting membership or value, while a short window captures current intent. Combining them, such as a twelve month buyer who browsed this week, usually targets better than either window alone.