It's Payback Time (again)
In a previous post on the value of modeling and how to think about payback, I shared some basic tools and frameworks to approach unit economics and sustainability.
Now I want to share a follow up that goes into a bit more detail about the tactical decisions that flow from there and start understanding what goes into basic ~~cohort modeling~~.
This is going to build on the content/concepts I covered previously. If you haven’t read Role Modeling or don’t feel like you’re comfortable with these concepts, I’d suggest pausing and checking that out first. Here’s a refresher:
my goal is usually not to determine what will happen, but rather to understand what would need to be true for something to happen. To my mind, the point of modeling is to ask and answer questions rigorously, and to be explicit about your assumptions. Putting things into numbers and breaking processes into discrete steps forces you to be specific in your thinking and with the story you’re telling, even if the numbers and steps are themselves unspecific.
Because startups are money-losing growth machines by design, lots of traditional financial modeling just doesn’t apply. Too often that means overcompensating and looking at top-line performance absent any more rigorous analysis of what I think of as “sustainability.” Is the growth healthy? People throw around all kinds of terms to asses the health and sustainability of startups. I think it’s mostly bullshit and doesn’t capture or describe anything meaningful.
I’ve found myself increasingly creating models (which again are thinking frameworks rather than predictive tools) to blend together all the various top-line figures into a more-startup oriented version of indicative health. I like to think about things in terms of payback in particular.
Once you’ve begun to understand the basic economics of a business, you’ll need to start thinking about more tactical (but no less important) questions using the same general framework. I want to focus on one in particular. What’s the potential impact of an upfront payment versus a pay as you go model? This is obviously crucial to any business with designs on subscription or repeat revenue.
Once again, I’ll use Harry’s and an example and, once again, all these numbers are totally made up and very very wrong.
Let start with some simple assumptions and say that a full year of Harry’s blades and shaving cream costs you $48 spread across four, quarterly shipments. Let’s also use the same $35 CPA and 70% gross margin we used in the previous payback analysis. The crucial output here is “periods to payback” because it answers what needs to be true. The 7 lifetime orders per customer is then a reasonable assumption that shows out where things net out. Here’s what that got us last time around:
Back to the matter at hand. If you charge people upfront, you’ll probably have fewer customers (asking for more money today is a barrier to purchase). On the other hand, your customers probably won’t churn as much because they’ve already committed to paying (even if you give them a cancellation option or risk free guarantee). Plus, maybe you can charge higher rates for pay as you go. After all, “pay for the year and get 10% off” really just means “pay as you go and I’ll charge you an extra 10%.”
This seems complex enough for now so I’ll put aside the implications on cash flow for the moment. That’s a topic for another time but suffice to say that upfront payment is favorable to you for all the reasons that pay as you go is favorable to your customers.
As I’ve said, the point of this exercise is to answer what needs to be true in order for me to meet my desired outcomes. Everything here is about being rigorous in our thinking, not trying to predict the future. I’m illustrating a general concept, not proving a specific point.
Now let’s use those same assumptions for Harry’s and add in some more info. We’ll assume that Harry’s converts 1.5% of “quality” (non-bounce) visitors to its website into customers. Seems reasonable enough. Some easy back of the envelope math tells us that that means Harry’s is paying $0.53 for each “lead” (person an ad pushes to its site). Finally, we’ll make some simple assumptions around churn/cancellations after each shipment. Here’s what we get:
You might look at this and think the numbers don’t tie. I said it would take 4.17 orders to pay back the CPA, now that only seems like it happens around order 8. What gives?
Unlike the previous Harry’s payback model, this is a time series. That means that churn/retention happens in “real time” as people attrite off with each order rather than all at once at the end. So if you sum up the cohort population percentages through shipment 7 (when net payback starts to get into the black, you’ll get ≈4 orders on average for that cohort. Orders to payback is right in line for the whole population but it takes longer to get there because so many customers churn off far in advance.
(If you couldn’t already tell, this is getting dangerously close to the cohort analysis post I’ve promised.)
Now, putting on our operator hats, we want to know “how do I make this better?” At bare minimum, we’ll want to think through the tradeoffs of an altered model. Everyone seems to offer some kind of “subscribe and save” or “pay now and save” option so there must be something to it. Let’s see what happens.
To be conservative, we’ll say that pay as you go will costs users nothing extra. Churn should go up because customers don’t feel like they’ve already spent the money and conversion should go up because pay as you go is a lower barrier to purchase. We don’t know by how much either will change but we’ll say that both churn and conversion increase by 25% . That gain on conversion decreases CPA because CPL stays the same but now more of those users are actually buying once they hit the site. Otherwise, the inputs are exactly the same. The outcomes, however, vary widely from the first case:
What we see is that even though orders per user over the two year period decreases from 4.37 (paying annually upfront) to 3.77 (pay as you go), net payback more than doubles from 5% to 13% throughout the same timespan.
So obviously this is the right answer, right?
Not necessarily. You have to remember that I’m making some pretty wild assumptions. The devil is in the details and no matter how robust your model and how much data you have, early stage operators need to have conviction behind their choices and a POV that goes beyond 20 minutes of excel. The “right” answer will vary based on factors this type of model couldn’t possibly capture, factors that are intrinsic to your customers and your product and your brand and your cash flow needs and your goals.
But this is at least a good place to start.
For anyone who’s interested, I’ve updated the payback model in Google Sheets to include this exercise. Play around with it, let me know what I got wrong, and tell me what I should be thinking about next.