This is a very important question, and I think companies don't take it seriously enough. Changing payback windows isn't a trivial decision: it fundamentally changes the pace of cashflow for the business and can materially impact monthly P&L. But beyond those mechanical changes, which can be modeled, it also signals to the product team that there are no more levers for growth available beyond increasing acquisition costs (which is what extending a payback window is), which, from my perspective, effectively makes the case that the product has started a long-term decline.
Of course, the above doesn't have to be true: if a company was under-spending on marketing for a product by buying against an LTV that was way too conservative, then extending the payback window is a sensible thing to do. But my experience has been that companies generally run marketing against aggressive payback windows to begin with (that is, they operate rationally), and that extending the payback window has signaled "the beginning of the end" of the product. There are a few reasons for this.
- It's fairly common for companies to miscalculate / overestimate the degree to which a cohort continues to monetize beyond its break-even point -- especially very late in the product's lifecycle. At release, it might be realistic that cohorts into a product break even at 90 days and continue to contribute another 50% of CAC over the next 3 months (that is, 50% profit margin after 6 months). But does that 50% margin hold one year or two years after release? That's an assumption that needs to be tested.
- Oftentimes marketing saturation has degraded ad effectiveness after a year or two to the point that the operational ROAS model doesn't hold for new cohorts -- the curve has changed. So extending the payback window against the curve that is being executed against (which is no longer valid) simply puts new cohorts even further underwater at the assumed break-even point.
My general take on payback windows is that it's not sensible to declare some inflexible strategic payback window in the first place (eg. "We buy against a 90-day payback window"). I believe the user acquisition team should constantly be measuring ROAS and only incrementally move payback windows out when they are confident that they have aggregated enough data to feel confident that the ROAS model holds at that point. I discussed this approach in an article called It's time to retire the LTV metric. In the article, I outlined an approach to incrementally testing break-even points as spend scales:
So to directly answer the question: I don't like the approach of declaring a payback window and buying against that until it needs to be changed. I prefer to constantly shift the payback window up or down as cohorts prove out the ROAS curve and market conditions change (eg. does your payback window / ROAS model hold for cohorts bought around Christmas? Probably not). But even within the set payback window paradigm, I think teams for the most part take the decision to extend the payback window way too lightly: it's a big deal, and it says a lot about how the product is expected to grow going forward.