An early startup rarely has a complete view of a customer’s lifetime. A lifetime value estimate can still help planning, but it should expose uncertainty rather than hide it inside a formula.
The most dangerous version is a precise number built from a short period of favorable retention and an assumption that the pattern continues indefinitely.
Start with observed contribution
Show what customers have actually produced so far: revenue, relevant costs, and retention over an identified period. Keep the observation window visible. Separate that evidence from the future contribution the model assumes.
For a hypothetical subscription cohort observed for three months, strong early retention does not establish what happens over three years. The model needs an explicit assumption to cross that gap.
Build a small scenario table
| Input | Conservative case | Working case | Upside case |
|---|---|---|---|
| Retention | Earlier loss of customers | Current best estimate | Longer sustained use |
| Contribution | Higher cost to serve | Expected margin | Better operating efficiency |
| Expansion | Limited additional value | Evidence-based growth | Additional adoption succeeds |
Populate the table with company-specific assumptions. These labels are not benchmark values.
Keep segments separate when needed
A blended average can mix customers with different buying reasons, usage, and support costs. Segment where it changes the acquisition decision. Avoid slicing so finely that every estimate rests on one or two accounts.
Examine the acquisition implication
Ask how much the company could reasonably spend if the conservative case occurs. Compare that with cash availability and the time needed to recover the investment. A high modeled lifetime value does not remove a near-term funding constraint.
Update with new evidence
Track how estimates change as cohorts mature. Preserve earlier versions so the team can see whether it consistently overestimates retention or expansion. That calibration is more valuable than repeatedly presenting the newest optimistic number.
Use shorter-horizon measures when they are more credible
Observed contribution, activation, retention at a defined age, and CAC payback can provide useful decisions while the lifetime model remains uncertain. They answer narrower questions with clearer evidence.
A useful LTV model should tell you which assumption matters most and what would change the acquisition decision. If it only provides a large number to justify spending, it needs another review.
Co-founder and CEO of Stackmatix, startup advisor, and former Head of Sales at MightyHive. Connect with Matt on LinkedIn ↗ · More about Matt →