Segmentation and analytics

Predictive LTV: player value from the first days

Predictive LTV (pLTV) is an estimate of a player's future NGR based on their first days: the first deposit amount, deposits over 7 days, games and payment method. It lets you decide on CPA rates, early VIP contact and bonuses without waiting for a year of history.

Yevhen SychovOctober 2026

Formula and example

Predictive LTV = NGR for the first 7 days x multiplier from past cohortsFor example, in past cohorts 12-month NGR is on average 4 times the NGR of the first 7 days. A new cohort brought in €50 per player in its first week: the forecast is €200 per player for the year.

The multiplier is rough for a single player, but it works for a cohort or a source. A model that learns from past cohorts and looks at signals from the first days is more accurate:

  • the first deposit amount and the number of deposits in 7 days;
  • games and average bet: slots, live casino, sports betting;
  • payment method, traffic source and country.

How to use it

  • CPA rates: a source with a forecast of €300 per player can carry a higher CPA than a source with a forecast of €120. You decide after two weeks, not six months.
  • Early VIP: a VIP manager looks after future high rollers with a top forecast from the first week.
  • Bonuses: the offer size follows the player's expected value, not the amount of a single deposit.

Checking against actuals

The forecast is checked against the actual LTV of every new cohort and for every source: the 90-day NGR forecast against the actual figure on day 90. For example, the forecast was €80 per player and the actual figure is €60: the model overstates by €20, and the CPA rates calculated from it are overstated too. It is handy to collect the actuals in a cohort table, as for cohort retention. When the product, traffic or bonuses change, the model is retrained.

Common mistakes

  • Forecasting deposits. Large deposits made for bonuses and followed by quick withdrawals give a high forecast and low NGR. It is NGR that needs forecasting.
  • One model for everything. Sources, countries and products have different revenue curves: a slots multiplier does not fit sports betting.
  • Treating a high forecast as a reason for big offers. A sharp rise in spending can be a sign of a problem, not of value. Check for risk signs first, and for big players, the source of funds too.
All glossary terms