CASE STUDY · 2025 · BEAUTY
7 WEEKS · DATA SCIENCE CONSULTING

Predicting the reorder before it lapses.

Taos Aer is a replenishment business, so reorders are everything. Planning was reactive, and at-risk subscribers only surfaced after they cancelled. I built a forecasting and churn-risk model, surfaced on one replenishment intelligence screen.
REORDER FORECAST HORIZON
1 CYCLE

Plan inventory a cycle ahead

SCORED PER SUBSCRIBER
CHURN RISK

At-risk routed to win-back

REPLENISHMENT SCREEN
1

Forecast, risk, subscription health

The problem.

For a replenishment brand, the whole model rests on customers coming back on schedule. But planning was reactive: inventory got ordered by looking backward, and churn showed up as a cancellation, not a warning. Stockouts and dead stock both cost money, and win-back always arrived a step too late.

The approach.

I built a model that forecasts reorders a cycle ahead and scores every subscriber for churn risk each week. Three views tie it together: a forecast view comparing predicted against actual reorders, a churn-risk view that segments members from healthy to lapsing, and a subscription health view with the retention curve and net revenue retention.

The outcome.

Reordering became proactive and win-back got ahead of the cancel button. Inventory is planned against a forecast that tracks reality closely, and at-risk subscribers are flagged and routed into a win-back flow while there is still a subscription to save, so fewer stockouts, less dead stock, and lower avoidable churn.

+
INTERACTIVE SAMPLE

Explore the dashboard.

INTERACTIVE SAMPLE · Illustrative data, not client figures. Switch tabs for reorder forecast, churn risk, and subscription health.

OPEN THE SAMPLE IN A NEW TAB
Finding out about churn after the cancel?

If your model depends on reorders but planning is reactive, I will forecast demand and flag at-risk customers before they lapse. One slot open for Q3 2026.

BOOK A 20-MIN CALL