Predictive LTV for acquisition and retention
Kohepets sees customer value beyond the first order.
A predictive model gives acquisition and retention the same view of likely customer value, bringing repeat purchasing into decisions that used to stop at the first order.
The impact
28%
higher predicted acquisition value
Predictive LTV for acquisition and retention
Case study facts
- Industry
- Pet retail & pharmacy
- Team
- Marketing
- Systems connected
- 4
- Time to live
- 3 weeks
- Engagement
- AI implementation
The impact
Results for the team
- higher predicted acquisition value
- 28%
- higher predicted acquisition value
- Marketing compares acquisition audiences using predicted lifetime value, prioritizing cohorts with higher expected value.
- more repeat purchases
- 21%
- more repeat purchases
- Repeat purchases track whether customers return to buy again, alongside the model's prediction of their future value.
- faster audience prioritization
- 4×
- faster audience prioritization
- Teams spend less time ranking acquisition and retention audiences because both use the same predicted-value signals.
Situation
One customer. Two views of value.
A first purchase says little about a customer's longer-term relationship with a pet retailer. Campaign reporting and retention activity can optimize separate short-term goals without a shared view of repeat demand.
Challenge
The first order hid the longer relationship.
Acquisition needed to compare the quality of new customers, while retention needed to prioritize existing relationships. Both required a consistent forecast horizon and a way to tell a reliable prediction from an uncertain one.
What we built
A shared prediction of customer value.
We built a pipeline that assembles purchase history, recency, frequency and product mix into customer-level features. A predictive model estimates future value over a defined horizon, with uncertainty kept alongside every score.
Those estimates power acquisition cohort comparisons and retention priorities, and they are tested against realized value, not treated as guaranteed revenue.
- 01
Unify purchase history
Resolve customer identities and net out cancellations and refunds.
- 02
Engineer value signals
- Feature engineering
Derive repeat cadence, recency, frequency and category patterns for every customer.
- 03
Predict future value
- Predictive model
The model scores each customer over a consistent horizon, with uncertainty attached.
- 04
Validate calibration
- Backtesting
Predictions are checked against later purchases using time-based holdouts.
- 05
Activate audiences
Acquisition and retention teams receive ready-to-use value bands.
- 06
Refresh continuously
Estimates update as customers buy again or their behavior shifts.
Evals and guardrails
Test predicted value. Keep spending under review.
Forecast evals
Test predictions on later purchases the model has not seen. Measure errors and how often actual customer value falls within the predicted range.
Reliability evals
Compare predicted and actual value across customer groups. Test limited purchase histories and exclude information unavailable at prediction time.
Guardrails
Check that customer data is current and audiences are eligible. Use a fallback for limited histories; require approval for budget changes.
Results
Growth decisions look beyond the first sale.
Marketing compares acquisition audiences using predicted lifetime value, prioritizing cohorts with higher expected value. Repeat purchases track whether customers return to buy again, alongside the model's prediction of their future value.
Teams spend less time ranking acquisition and retention audiences because both use the same predicted-value signals.
- higher predicted acquisition value
- 28%higher predicted acquisition value
- more repeat purchases
- 21%more repeat purchases
- faster audience prioritization
- 4×faster audience prioritization
From Kohepets
“We have worked with Revensi for many years. In that time they have made vast improvements to how our business runs, and they are great people to work with. We highly recommend them.”
Build Capacity To Grow. Own Your Intelligence.
Bring the workflow slowing your sales, orders or delivery. We'll map the bottleneck and show you what the AI system could look like.