Netflix flags a leaving subscriber weeks early. Your loyalty dashboard finds out when the sales drop.

Open your loyalty dashboard. It shows member count, points issued, redemption rate, and last month’s sales. All green.

Now find the customer who refilled every 30 days for two years and is now 50 days late. You cannot. The dashboard was not built to surface her.

She has been drifting for weeks. Smaller basket, longer gaps, a refill missed. By the time she shows up in your numbers, she is already shopping somewhere else.

Netflix predicts churn from behavior, not from who the customer is

Netflix watches the signals that move before a cancellation: watch hours falling, series started and never finished, the gap between sessions stretching out.

When those signals tip, a model flags the account and triggers a re-engagement nudge before the customer decides to leave. Netflix keeps over 90% of its subscribers this way (Netflix churn analysis, vsenk.com, Sep 2025).

A 2026 study of Netflix’s user data found the sharpest lesson for any loyalty team: demographics and plan tier barely predict churn, while behavior predicts it strongly (Bouton, Netflix churn study, Jan 2026).

Who your customer is tells you almost nothing. What she just did tells you whether she is leaving.

Your loyalty reporting scores status, not movement

Most loyalty dashboards in retail count members and report points and total sales. They describe the base as it stands today.

None of that catches a customer mid-slide. A member who moved from weekly to monthly still counts as a member. A diabetic who missed a refill still sits in the active tier until the data ages out.

The waste compounds. Antavo’s 2026 loyalty report found 27% of points go unspent and 12% expire, roughly $10 billion a year, while 51.5% of marketing budgets now go to loyalty and customer relationship management, the system that holds your member data (Antavo 2026 Global Customer Loyalty Report, Business Wire).

You are spending more on the program and still cannot name the customer about to walk.

Three questions to ask your own data this week

If your team cannot answer these from the current system, that gap is the leak.

01 · Drift detection

How many of last quarter’s active members have quietly slowed down this quarter?

If the answer takes a manual export and a week, you are reading status, not movement. Netflix reads this daily and acts inside the first 90 days, where most churn happens (Bouton, Jan 2026).

02 · Segment triggers

Does a slowing customer get a different message from an active one, or the same blast?

One promo to every member trains your best customers to ignore you. Watsons Club passed 8 million members in 2025, scale that only pays if the message changes by behavior (BERNAMA, The Star, 2025).

03 · Win-back timing

When you reach a lapsing customer, has she already gone, or is there still time?

A win-back sent after she has switched stores is a discount to someone who left. The signal you needed was visible weeks earlier, in your own till and refill data.

One thing you can do ASAP

Pick one chronic category, hypertension or diabetes refills, and pull every member whose last refill is more than 15 days past their usual cycle.

That list is your drifting base. It already exists in your till and loyalty data. The fact that nobody runs it weekly is the whole problem.

Reach that list before they run out, not after. That single move is the difference between a refill and a lost customer.

Thanks for reading this far. The customer leaving you this month is not a stranger. She is a member you already have the data on.