Habits and buying windows

Moonbrush research note 2

Version

1.0

Published

October 20, 2026

Checked against the product

October 19, 2026

Authors

Moonbrush

Keywordshabit; past behavior; habit discontinuity; life events; recency; timing of advertising

Abstract

Gravity reads two kinds of behavior in two different ways. Habits and shopping manner are read from repeated behavior over weeks, relative to everyone else. Buying windows are read from a recent episode of research, are absolute, and close on their own. This note sets out the research behind that split.

Habit research finds that behavior repeated in a stable context becomes automatic, that a large share of everyday behavior is of this kind, and that for such behavior past frequency predicts future behavior directly. The same research finds that habits are most open to change when the context changes, as after a move, and consumer research links life changes to changes in brand preference. Field experiments on advertising find that response depends on timing and on where a person is in a decision, with effects that fall quickly after the moment of interest. Together these support reading stable dispositions and passing episodes as different things. The literature does not fix the length of any window, and Moonbrush does not publish its own.

1. Introduction

A person who has bought the same brand of coffee every week for years and a person who started researching a new roof three days ago are both showing behavior that matters to a marketer. They are not showing the same kind of behavior. The first is a disposition that changes slowly. The second is an episode that will end, one way or another.

Gravity treats these as different kinds of model. See The model catalog. This note asks what the research says about each, and why mixing them would be a mistake.

2. Habits

2.1 Past behavior and habit

Past behavior predicts future behavior through two routes. For behavior that is well practiced and performed in a stable setting, it becomes automatic, and how often it was done before predicts what happens next directly. For behavior that is less well learned, or performed in unstable settings, people decide deliberately, and intentions guide what they do (Ouellette and Wood, 1998).

2.2 How much of life is habitual

Habits are common. In two diary studies in which people reported what they were doing every hour, behavior done almost every day, usually in the same place, made up 35% and 43% of what was reported: between a third and a half of everyday behavior (Wood et al., 2002). A review of habit in consumer settings applies the same account to purchases (Wood and Neal, 2009), and a broader review covers the psychology of habit in general (Wood and Rünger, 2016).

2.3 How this bears on modeling

If much of what people do is habitual, then how a person tends to shop, save and spend is a stable property worth measuring, and it is best read from repetition over time rather than from any single act. It is also a relative property. Saying a person “waits for the markdown window” means they do so more than most people, so a habit score is a comparison with the population. In its simplest form, a statement of that kind is a rank in the population distribution of a behavior $x$:

$$h_i = \Pr\left(X \le x_i\right), \quad X \sim \text{the population}$$

which is how the national percentiles in the person view read. A percentile is a position in the population, not a probability of acting. How Gravity builds the behavior measures themselves is not published.

3. Buying windows

3.1 Life changes open windows

Habits are most open to change when the setting that sustains them changes. In a field experiment, an intervention to change household behavior worked better among people who had recently moved than among matched people who had not, and the advantage appeared only within about three months of the move, a window the authors describe as tentative (Verplanken and Roy, 2016). Earlier consumer research linked changes in life status to changes in brand preferences, and found households going through such changes more open to marketers (Andreasen, 1984).

3.2 Timing and stage

Advertising response depends on when and where in a decision a person is. In a field experiment with an online travel firm, ads showing the products a person had viewed generally did worse than generic brand ads, except once a person’s preferences had narrowed, for example after visiting review sites (Lambrecht and Tucker, 2013). In a randomized field experiment with an online home-improvement retailer, ads shown after a visit led 14.6% more people to return within four weeks, the effect weakened with time since the first visit, and a third of the first week’s effect came on the first day (Sahni et al., 2019).

3.3 Recency as evidence

Models of customer purchasing have long treated how recently and how often a person acted as evidence about whether they are still active (Schmittlein et al., 1987; Fader et al., 2005). The same idea applies to research: a burst of research that has gone quiet is weaker evidence than one that is still running.

3.4 How this bears on modeling

A buying window is an episode with a start and an end. It is absolute rather than relative: whether a person is in a window does not depend on what everyone else is doing. It closes on its own once the research stops, without anyone removing the person. A buying-window score, like every Gravity score, is still a rank from 0 to 1; see Reading a score.

Each model card states the span of behavior the model reads and how often it is re-scored. How those spans and the cuts for each model are set is not published.

Figure 1. A habit is read against everyone else; a buying window opens and closes. Invented shapes. Left: a habit score compares one person's repeated behavior with the population over weeks, so it is relative and slow to change. Right: a buying window is an episode. It opens when research starts and closes on its own when the research stops, without anyone removing the person. The spans and cuts Moonbrush uses are not published.

Schematic, not data

4. Why the two are kept apart

Table 1. How the two kinds of model differ, and the evidence behind each difference.

Property	Habit and shopping manner	Buying window
What it reads	Repeated behavior over weeks	A recent, still-warm episode
Reference point	Relative to the population	Absolute: in or out
How it changes	Slowly	Opens with an episode of research and closes when it ends
Evidence	Past behavior predicts behavior in stable contexts (Ouellette and Wood, 1998; Wood et al., 2002)	Life changes and decision stage change response (Verplanken and Roy, 2016; Lambrecht and Tucker, 2013)
Failure if mixed up	A passing episode is mistaken for a lasting trait	A lasting habit is treated as if it were about to end

A model that blended the two would rank a person who researched a roof once, two years ago, alongside one who is researching it today, and it would treat a stable preference as if it were about to expire.

5. Limits and threats to validity

• Window lengths are not given by the literature. The research shows that windows exist and that effects fall quickly. It does not fix a length for any product, and the three-month figure above is specific to one setting and described by its authors as tentative.
• Lab and field settings differ. Several of the habit findings come from diary and survey studies, and the timing findings from single firms’ field experiments. Each generalizes with care.
• Seen and unseen behavior. Habits that happen offline are visible only through their online traces.

6. Questions an auditor should ask

1. For a buying-window model, what span of behavior does it read, and how was that span chosen?
2. How often is each kind of model re-scored, and how does a stale score show in a delivered file?
3. Does any habit model’s ranking change sharply between releases, and if so, why?

7. Conclusion

Habit and episode are different things in the research, and they are different kinds of model in Gravity. Habits are read from repetition, relative to everyone else, because that is where habit research finds them. Buying windows are read as episodes that close on their own, because life changes and decision stages open windows that do not stay open.

References

1. Andreasen, A. R. (1984). Life status changes and changes in consumer preferences and satisfaction. Journal of Consumer Research, 11(3), 784–794. doi.org/10.1086/209014
2. Fader, P. S., Hardie, B. G. S., & Lee, K. L. (2005). “Counting your customers” the easy way: An alternative to the Pareto/NBD model. Marketing Science, 24(2), 275–284. doi.org/10.1287/mksc.1040.0098
3. Lambrecht, A., & Tucker, C. (2013). When does retargeting work? Information specificity in online advertising. Journal of Marketing Research, 50(5), 561–576. doi.org/10.1509/jmr.11.0503
4. Ouellette, J. A., & Wood, W. (1998). Habit and intention in everyday life: The multiple processes by which past behavior predicts future behavior. Psychological Bulletin, 124(1), 54–74. doi.org/10.1037/0033-2909.124.1.54
5. Sahni, N. S., Narayanan, S., & Kalyanam, K. (2019). An experimental investigation of the effects of retargeted advertising: The role of frequency and timing. Journal of Marketing Research, 56(3), 401–418. doi.org/10.1177/0022243718813987
6. Schmittlein, D. C., Morrison, D. G., & Colombo, R. (1987). Counting your customers: Who are they and what will they do next?. Management Science, 33(1), 1–24. doi.org/10.1287/mnsc.33.1.1
7. Verplanken, B., & Roy, D. (2016). Empowering interventions to promote sustainable lifestyles: Testing the habit discontinuity hypothesis in a field experiment. Journal of Environmental Psychology, 45, 127–134. doi.org/10.1016/j.jenvp.2015.11.008
8. Wood, W., Quinn, J. M., & Kashy, D. A. (2002). Habits in everyday life: Thought, emotion, and action. Journal of Personality and Social Psychology, 83(6), 1281–1297. doi.org/10.1037/0022-3514.83.6.1281
9. Wood, W., & Neal, D. T. (2009). The habitual consumer. Journal of Consumer Psychology, 19(4), 579–592. doi.org/10.1016/j.jcps.2009.08.003
10. Wood, W., & Rünger, D. (2016). Psychology of habit. Annual Review of Psychology, 67, 289–314. doi.org/10.1146/annurev-psych-122414-033417

How to cite this note

Moonbrush. (2026). Habits and buying windows (Research note 2, version 1.0). https://docs.moonbrush.com/research/habits-and-buying-windows/