What is Predicted Lifetime Value (pLTV)?

Predicted lifetime value (pLTV) is a forward-looking estimate of the revenue or gross margin a customer will generate in the future, usually over a fixed horizon such as 12 or 24 months. It is predicted from purchase history and other signals, often right after the first order, so that acquisition and retention decisions can use it early.

pLTV vs. historical LTV

Both terms describe customer value, but they look in different directions:

• Historical LTV is backward-looking: it sums the revenue or margin a customer, or a cohort of customers, has already generated up to today.
• Predicted LTV (pLTV) is forward-looking: it estimates the revenue or margin a customer will generate in the future, usually over a fixed horizon such as 12 or 24 months.

Historical LTV is known only after the fact, which is too late for decisions about how much to pay for a customer acquired today. pLTV fills that gap. The forward-looking definition matches how marketing research defines customer lifetime value: the expected sum of discounted future margins (Gupta et al., 2006). The two are connected: historical data train the pLTV model, and historical LTV of older cohorts is the benchmark for checking its accuracy.

How pLTV is predicted

Probabilistic customer-base models

Ecommerce and retail relationships are non-contractual: customers do not announce when they stop buying. Probabilistic models handle this by estimating, from each customer's recency, frequency and time since first purchase, how likely they are to still be active and how many purchases they will make. The BG/NBD model is a widely used example and a simpler alternative to the earlier Pareto/NBD model (Fader, Hardie & Lee, 2005). A separate spend model then estimates the value per purchase. Open-source implementations of these models are available, for example in PyMC-Marketing.

Machine learning models

Regression and machine learning models predict future value from a wider set of features: first-order value and basket contents, product categories, discount use, acquisition channel, returns and browsing behavior. These models can predict pLTV right after the first order, when a purchase-history model has little to go on. Their accuracy depends on having enough mature cohorts to learn from.

Key modeling choices

• Horizon: a defined window (for example 12 or 24 months) is easier to validate than an open-ended "lifetime".
• Revenue or margin: margin-based pLTV, net of returns, is the better basis for acquisition decisions.
• Discounting: for long horizons, future margins should be discounted to present value (Gupta et al., 2006).

How pLTV is used in ecommerce and retail

• Acquisition targets: pLTV sets an upper bound on how much it is worth paying for a new customer. Comparing it with customer acquisition cost (CAC) shows whether acquisition is profitable over the chosen horizon.
• Channel and campaign comparison: channels, campaigns and first-order offers can attract customers with very different future value. A discount-led campaign may have a low CAC but a low pLTV.
• Retention and CRM: segments with high pLTV can be prioritized for retention activities and service. Customer value runs both ways: the value customers deliver to the firm depends on the value the firm delivers to them (Kumar & Reinartz, 2016).
• Value-based bidding: pLTV can be passed to ad platforms as a conversion value, so that bidding optimizes toward valuable customers rather than order count.

pLTV as a KPI in marketing mix modeling

A standard marketing mix model (MMM) estimates the incremental sales each channel causes in the measured period. It does not capture the future value of the new customers a channel brings in. pLTV can add that longer-term view:

• Model new customers separately: an MMM built on new-customer sales or counts gives the incremental new customers per channel. See how a new vs. returning customer split in MMM results helps to drive both growth and efficiency.
• Value them with pLTV: multiplying incremental new customers by their pLTV gives an estimate of the long-term value each channel creates, which can be compared with spend as a pLTV-based incremental ROAS (iROAS).

The key requirement is that the customers are counted incrementally. Applying pLTV to platform-attributed customers credits channels for customers who would have bought anyway. For how platform metrics differ from incremental ones, see platform ROAS vs. incremental ROAS.

Common challenges

• Validation: compare predictions with realized value of older cohorts at the same horizon before relying on the model.
• Changing behavior: new product ranges, pricing or economic conditions can make past purchase patterns a poor guide. Retrain models regularly.
• Cold start: predictions immediately after the first order are uncertain; report ranges, not just point estimates.
• Customer identity: guest checkouts and offline purchases split one customer into several, which biases pLTV downward.

Key takeaways

• Historical LTV is backward-looking; predicted LTV (pLTV) is forward-looking and is what acquisition decisions need.
• In ecommerce, pLTV is typically built with probabilistic purchase models such as BG/NBD or with machine learning on first-order and behavioral features.
• In MMM, pLTV is most useful for valuing incremental new customers by channel, giving a longer-term view of iROAS.

Related terms

• Lifetime value (LTV)
• Customer acquisition cost (CAC)
• Average order value (AOV)
• Return on advertising spend (ROAS)
• Bayesian statistics

Further reading on Sellforte

• How a new vs. returning customer split in MMM results helps to drive both growth and efficiency
• How can MMM help optimize ecommerce KPI challenges?
• MMM for ecommerce brands

References

1. Fader, P. S., Hardie, B. G. S., & Lee, K. L. (2005). "Counting your customers" the easy way: An alternative to the Pareto/NBD model. Marketing Science, 24(2), 275–284. https://doi.org/10.1287/mksc.1040.0098
2. Gupta, S., Hanssens, D., Hardie, B., Kahn, W., Kumar, V., Lin, N., Ravishanker, N., & Sriram, S. (2006). Modeling customer lifetime value. Journal of Service Research, 9(2), 139–155. https://doi.org/10.1177/1094670506293810
3. Kumar, V., & Reinartz, W. (2016). Creating enduring customer value. Journal of Marketing, 80(6), 36–68. https://doi.org/10.1509/jm.15.0414
4. PyMC Labs and contributors. (n.d.). PyMC-Marketing [Software documentation: customer lifetime value models]. Retrieved October 30, 2026, from https://www.pymc-marketing.io/

Authors

Lauri Potka is the Chief Operating Officer at Sellforte, with over 15 years of experience in Marketing Mix Modeling, marketing measurement, and media spend optimization. Before joining Sellforte, he worked as a management consultant at the Boston Consulting Group, advising some of the world’s largest advertisers on data-driven marketing optimization. Follow Lauri in LinkedIn, where he is one of the leading voices in MMM and marketing measurement.