Perhaps the most contentious aspect of user acquisition strategy is the payback window -- and the decision to extend it. Often, extending the payback window is seen as an easy way to scale spend, but doing that is fraught. I'll explain why (1/X)
2/ First, I've written at length about payback windows and why recoup timelines should be dictated by cash flow concerns. In this post I propose a framework for marketing
P&L management using LTV and ROAS windows mobiledevmemo.com/ltv-roas-marke…
P&L management using LTV and ROAS windows mobiledevmemo.com/ltv-roas-marke…
3/ When a marketing team reaches a point of stasis with spend -- or, more commonly, when spend begins to decline at a CAC that is acceptable to it -- it is tempting to simply extend the payback window to allow for an increase in budget quantmar.com/255/What-point…
4/ What's the benefit of this? If the 15-month LTV for a given segment is 10% higher than its 12-month LTV, then it allows the advertiser to bid 10% more on traffic mobiledevmemo.com/ltv-cpi-vs-cas…
5/ But doing this introduces three problems. First, by definition, 3 fewer monthly cohorts have progressed through 15 months in the product than 12 months. The product's retention curve might reduce the number of users that have aged to that point further mobiledevmemo.com/much-data-need…
6/ Second, extending the payback window puts stress on cash flow: more money is expended upfront and it is returned over a longer period of time. Not all companies have the cash reserves to fund this. mobiledevmemo.com/optimizing-cam…
7/ But perhaps most importantly: the less data the advertiser has for the longer frontier in the timeline, the more volatile the projections are going to be. Often an advertiser trades extreme volatility for some single-digit increase to bid prices.
8/ This could actually ultimately reduce the amount of spend: if the volatility in the predictions increases meaningfully and the absolute error rate changes (when most advertisers are working against thin margins -- 10%ish), then wins could start looking like losses more often
9/ One reason that it's difficult to build LTV models is that doing so requires a lot of data to produce reliable predictions, and data availability for later-stage windows is diminished by the retention curve mobiledevmemo.com/much-data-need…
10/ This is one of the reasons that I dont like using the LTV metric for marketing budgeting / performance optimization. I prefer starting from a short payback window and increasing that frontier as doing so is supported by the data mobiledevmemo.com/retire-the-lif…
• • •
Missing some Tweet in this thread? You can try to
force a refresh
