How to manage marketing spend during a recession

Last Thursday, the Dow Jones Industrial Average index dropped 10%, representing the single largest one-day decrease since 1987 (it gained most of that back the next day, but it remains down nearly 20% from the beginning of the year). And on Sunday, Federal Reserve chairman Jay Powell announced that the Fed would cut its interest rate target by 1% to between 0 – 0.25%, matching the interest rate target during the global financial crisis. The economic impact of the coronavirus has arrived, and a worldwide economic recession seems plausible if not likely.

How should consumer technology companies prepare for the possibility of a recession? The global financial crisis of 2008-2010 largely predated the rise of mobile — the App Store launched in 2008, and the freemium model only really became dominant in Western markets in the early 2010s. There is no playbook for freemium app developers to draw from when considering how to adjust their operations, especially their user acquisition marketing spending, during an economic downturn.

The first thing any consumer technology company tends to think about with a looming recession is how the acquisition of new customers will be impacted, but in reality, there are two key considerations that affect a company’s future cash flows against the assumption of decreased consumer spending:

  1. How should the company plan for reduced future income as a result of reduced marketing spending on new users;
  2. How should the company plan for reduced future income as a result of current users spending less?

The first point is deceptively challenging to model: if customer LTVs are lower as a result of reduced spending, then marketing budgets adjust lower, and (assuming acquisition costs remain constant) new user volumes decrease.

Consider a hypothetical product with a 90-day LTV of $2:

What happens if user behavior suddenly changes and early-stage monetization for cohorts decreases by 30% within the first five days of install?

A marketing team might decide to simply use its existing underlying LTV model to project cohort monetization out to Day 90 to $1.40, using a 30% decrease across the timeline, as below:

But is that a valid assumption? If early-stage monetization decreases by such a significant amount, does it make sense that the pattern of user monetization remains the same? What if the LTV model — the shape of the progression of cumulative revenue per cohort — has fundamentally changed, as below?

In this case, not only are the early-stage monetization values lower for the post-recession curve, but the curve’s shape is actually different: it stops increasing altogether at around Day 50 and reaches an ultimate Day 90 LTV value of $1.10. Compare this to the projected post-recession values, which were generated using the same curve coefficients as the pre-recession curve:

The difference — $1.10 measured Day 90 LTV versus a projected $1.40 projected LTV, for a roughly 21% delta — is meaningful. By using the pre-recession LTV curve to project post-recession monetization values to Day 90, the marketing team overbid for traffic by 21% because it assumed that, while monetization values for users decreased, overall monetization behavior and timing didn’t. That’s a dangerous assumption to make.

The reality is that any exogenous factor can change user monetization behavior, and so a disciplined, consistent re-evaluation of the LTV model used to derive marketing bids is a necessity. But it’s imperative to not rely on historical models of user monetization as the broader economic scenery transforms. The approach that I outlined in It’s time to retire the LTV metric is relevant here: using a constantly-evolving ROAS frontier based on observed cumulative monetization can help advertisers reset their assumptions around user behavior and adapt to changing user monetization. Users will spend money differently in a recession than they do otherwise; LTV models must be challenged and revised to reflect that.

The second consideration — that of existing user monetization behavior — is more straightforward to model in this context but potentially more difficult to accommodate in a cash flow projection. If an existing cohort suddenly begins to monetize differently — say, it stops tracking to the model after 60 or 90 or 180 days after joining the product — then all past projections of cash flow are invalidated. The losses in this case have already taken place, they just haven’t been appropriately booked because past spend was deployed on the basis of a model that no longer fits user monetization.

In these cases, assessing the damage requires breaking out cohort ages and “marking to market” the differences in previously projected monetization to the new observed behavior. Using a tool like Theseus to break cohorts apart can help; the advertiser needs to know how many cohorts exist of age X and then adjust those cohorts’ assumed monetization down to accommodate their changed behavior. The term I defined for assumed future monetization based on an LTV curve in Building a marketing P&L using LTV and ROAS is projected receivable: when users stop spending to the model, the advertiser’s projected receivables change, and it must not only adjust its bids for future cohorts but also book the losses for those cohorts in the months they were acquired.

It’s always import to diligently test and revise LTV models, but all advertisers need to consider how a potential recession will impact the monetization they’ll see both from new cohorts going forward as well as existing cohorts that they acquired against assumptions that may prove to be invalid. Uncertainty is a constant, but the magnitude of uncertainty that many consumer technology companies will face in the coming weeks and months is likely unprecedented in the mobile era. Assiduous assumptions testing, cash flow management, and revenue modeling can help protect companies from careless advertising spend.

Photo by Dan Burton on Unsplash