What Is a Lapsed Customer? How to Define Inactivity Without Guessing
A customer is not lapsed because 90 days passed.
If I buy a 30-day supplement from you and disappear for three months, something probably changed.
If I buy a dining table from you and disappear for three months, nothing unusual has happened at all.
Same customer status in the CRM.
Same number of days.
Completely different commercial reality.
And that is the problem with the way “lapsed customer” is often defined.
Most businesses need a rule because software needs rules.
So someone chooses:
60 days.
90 days.
180 days.
Then every customer who crosses the line becomes “inactive.”
It is operationally neat.
But the calendar is not the customer.
The more useful question is:
Should this person reasonably have bought again by now?
That is where lapse begins to become commercially meaningful.
Even the major platforms do not agree on one lapse definition
This alone should make us suspicious of universal inactivity rules.
Ometria’s default lifecycle defines a customer as:
- active if they purchased within the past 182 days,
- at risk between 182 and 385 days,
- lapsed after 385 days.
Ometria explicitly says those periods should be customized to the business and offers order-gap analysis to understand how long customers normally take between purchases. (success.ometria.com)
Bluecore takes a different approach in its retail customer-movement framework. It classifies customers who have not purchased in the previous year as inactive and then divides them into:
- 13–24 months,
- 25–36 months,
- 37+ months. (support.bluecore.com)
Neither company is necessarily wrong.
They are solving different problems across different retail bases.
The lesson is that “lapsed” is a business definition before it is a software field.
The Lapse Window
At Altimax, we use the term Lapse Window for:
The point at which a customer’s current absence becomes meaningfully unusual relative to expected buying behaviour.
There are several ways to estimate it.
The strongest brands will often use more than one.
Method 1: Look at normal time between orders
Suppose customers who buy Product A typically return in:
31 days
33 days
36 days
38 days
41 days
That gives us a fairly clear rhythm.
A customer at day 35 is not necessarily late.
A customer at day 80 probably is.
RetentionX’s Time Between Orders reporting makes an important distinction here by showing both average and median days between purchases, broken down by the step from one order to the next. (help.retentionx.com)
That distinction matters.
Consider these order gaps:
30
31
32
34
190
The average is:
63.4 days.
The median is:
32 days.
If you build your reactivation timing around the average, one outlier has nearly doubled your view of normal customer behaviour.
This is why I would almost always want to see the distribution and median, not just the average.
Method 2: Use the customer’s own history
Brand-level behaviour is useful.
Individual customer behaviour can be even better.
Suppose one customer ordered:
Day 1
Day 33
Day 65
Day 100
Day 134
Then nothing.
They have effectively told you:
“I buy about once a month.”
When they reach day 90 since their last order, we do not need a generic ecommerce benchmark to tell us something changed.
Their own history is enough.
Now compare that with a customer who bought twice:
Day 1
Day 210
Their buying rhythm is completely different.
The brand should not treat both customers as though they are operating on the same clock.
Method 3: Adjust for what they actually bought
This is where lifecycle automation often becomes less intelligent than it appears.
Two people buy the same supplement on the same date.
Customer A buys:
1 unit
Customer B buys:
4 units
Each unit is roughly a 30-day supply.
A day-40 replenishment campaign makes reasonable sense for Customer A.
For Customer B, it makes no sense at all.
They may have almost three months of product remaining.
A basic replenishment estimate could be:
Expected product duration
=
Units purchased × estimated usage period per unit
If the customer bought four 30-day units:
4 × 30
= 120 expected days of supply
That is not perfect.
People use products differently.
But it is far more commercially sensible than pretending both customers are due at the same time.
Method 4: Recognize that order number changes the meaning of lapse
A customer who bought once gives you very little individual purchase-history data.
A customer who bought eight times gives you much more.
That means the lapse logic should often change depending on where the customer sits in the relationship.
For a one-time buyer, we may rely on:
- cohort behaviour,
- first-product behaviour,
- average time to second purchase.
For a repeat customer, we can increasingly rely on:
- their own purchase cadence,
- product preferences,
- frequency changes,
- spend changes.
RetentionX explicitly breaks time-between-order analysis down by purchase step because customers can behave differently between order one and two than they do later in the relationship. (help.retentionx.com)
That matters.
The first-to-second-purchase problem is not necessarily the same as the fourth-to-fifth-purchase problem.
A customer can be “at risk” before they are lapsed
This distinction is useful.
Ometria separates at risk from lapsed, and its platform tracks the dates customers are expected to enter each state. (support.ometria.com)
That is a better mental model than waiting until a customer has disappeared completely.
Imagine someone normally buys every 45 days.
At day 35:
Probably active.
At day 50:
Maybe slightly late.
At day 70:
Something may be changing.
At day 120:
The relationship is meaningfully different.
Those are not four identical moments.
A business should not wait until the last one before noticing.
The Altimax lapse ladder
A practical customer model might look like this:
| Stage | Meaning |
|---|---|
| Active | Still inside normal expected behaviour |
| Watch | Approaching the outer edge of normal behaviour |
| At Risk | Behaviour is beginning to deviate |
| Early Lapse | Customer has moved beyond expected timing |
| Established Lapse | Absence is now materially unusual |
| Deep Lapse | Relationship has been inactive long enough that both recovery probability and contact risk need reconsideration |
The exact number of days should be different by:
- product,
- order count,
- category,
- customer history,
- seasonality.
The seasonal customer is another reason fixed rules fail
Imagine a customer who buys Christmas gifts every November.
They purchased:
November 2023
November 2024
November 2025
It is August 2026.
Are they lapsed?
Not necessarily.
Their buying behaviour may be perfectly intact.
Now imagine the system sends them:
“We miss you.”
in February.
The automation has technically identified inactivity.
Commercially, it has misunderstood the customer.
Seasonality belongs inside lapse logic.
Durable goods need a completely different model
Bluecore’s benchmarks show that health and beauty has much higher purchase frequency than categories such as home goods, precisely because the natural purchase occasion is different. (support.bluecore.com)
This is why a furniture customer might look inactive by a consumables standard while still having substantial future value.
The next purchase may simply be:
- another room,
- an accessory,
- replacement furniture,
- an occasion-driven purchase.
The customer is not late buying the same sofa.
The business needs to think about category progression, not replenishment.
Lapse can also be a behaviour change rather than complete disappearance
Imagine a customer used to order every month.
Then every six weeks.
Then every three months.
Then nothing.
The final purchase date matters.
But the deterioration began much earlier.
Signals can include:
- increasing gaps between orders,
- falling AOV,
- fewer categories purchased,
- cancellation,
- declining engagement,
- increasing discount dependence.
Revenue Recapture becomes much stronger when it watches movement, not only the final state.
Bluecore’s customer-movement framework is built around this idea: customers move between new, active, inactive and reactivated states, and the point is to understand how customer value shifts over time rather than watching only channel metrics. (bluecore.com)
What about a customer who never bought twice?
They deserve their own lapse definition.
Suppose the median customer who eventually places order #2 does so in 48 days.
Then the probability curve might look something like:
- 30% of second orders by day 30
- 60% by day 50
- 85% by day 80
- 95% by day 110
A customer at day 20 is not lapsed.
A customer at day 120 may now be meaningfully outside the normal Second-Purchase Window.
This is why a first-time buyer should not simply be dropped into the same 90-day inactivity segment as a previously loyal customer.
They represent different problems.
A useful lapse calculation
A simple starting model might be:
Lapse threshold
=
Expected reorder interval + tolerance buffer
Suppose:
Median reorder interval:
45 days
Tolerance:
15 days
Potential lapse threshold:
60 days
But then we can improve it.
For example:
Adjusted lapse threshold
=
Product duration × quantity purchased
adjusted for:
- observed cohort behaviour,
- customer history,
- seasonality.
The formula does not need to be perfect.
It needs to be commercially better than:
90 days because someone once chose 90 days.
Do not confuse email inactivity with customer lapse
Another common mistake:
Someone has not opened an email for six months.
That does not automatically mean they have stopped buying.
The reverse is also true.
Someone can open every newsletter and never purchase.
Purchase behaviour and marketing engagement are different signals.
Ometria, for example, stores lifecycle dates, purchase dates, web activity and email activity separately. (support.ometria.com)
That separation is useful.
A customer can be:
commercially active but marketing-unengaged
or:
marketing-engaged but commercially lapsed.
Do not collapse those into one definition.
Why defining lapse correctly matters financially
Timing changes economics.
Imagine a customer who normally buys every 60 days.
You decide the customer is lapsed at day 45 and send:
20% off.
They buy.
The campaign gets credit.
But the customer may simply have bought early because you paid them to.
Now imagine you wait until day 150.
By then:
- brand memory may be weaker,
- alternative habits may be established,
- engagement may be lower,
- deliverability may be worse.
Both mistakes cost money.
The objective is not:
earliest possible send.
Or:
longest possible wait.
It is finding the point where absence becomes meaningful enough to justify intervention.
That is the Lapse Window.
A worked example
Consider a skincare brand selling a serum.
Historical analysis shows:
Median time between first and second order:
54 days.
Median between second and third:
46 days.
Median after order three:
39 days.
That already tells us something interesting.
The relationship becomes more frequent as customers mature.
Now imagine three customers.
Customer A
1 order
65 days since purchase
Potentially approaching second-purchase lapse.
Customer B
5 orders
Normally buys every 38–42 days
70 days since last purchase
Much stronger lapse signal.
Customer C
2 orders
Bought three units last time
70 days since purchase
Potentially still active because inventory should last much longer.
Same 70 days.
Three different commercial interpretations.
The real goal is not defining “lapsed” perfectly
There is a temptation to build an impossibly sophisticated model.
You do not need that to improve on a global 90-day rule.
The objective is simply to avoid obvious mistakes.
If you know:
- what they bought,
- how much they bought,
- how often similar customers reorder,
- how often this customer used to reorder,
you can already make a materially better decision.
Frequently Asked Questions
Is 90 days a good lapsed-customer definition?
It can be.
But only if 90 days makes sense for the customer and product.
How do I define lapse for a one-time buyer?
Use cohort-level time-to-second-purchase behaviour rather than personal purchase cadence.
Should every product have a different lapse window?
Not necessarily every SKU, but materially different product consumption cycles should often be separated.
Should I use average or median time between purchases?
Look at both. Median is often less distorted by unusually long gaps.
Is an unsubscribed customer still lapsed?
They may be behaviourally lapsed, but unsubscribe status determines whether marketing contact is appropriate.
Can customers become active again?
Yes. Ometria explicitly tracks customers transitioning from lapsed back to active as “won back.” (support.ometria.com)
The simplest definition
A lapsed customer is not someone who crossed a calendar threshold.
A lapsed customer is someone whose absence has become unusual enough that the business should notice.
The customer who buys every 30 days and disappears for 100 is telling you something.
The customer who buys a sofa and waits 100 days is not.
If your system treats them the same, the problem is not the customer.
It is the definition.
That is why Revenue Recapture starts by understanding the Lapse Window before we build the win-back campaign.
Related reading
Customer Reactivation: How DTC Brands Can Recover Revenue From Lapsed Customers
How to Identify Lapsed Customers in Ecommerce
Want to know when your customers actually become lapsed?
The Revenue Recapture Analysis looks at real purchase behaviour rather than assuming every customer runs on the same clock.