The following post was written by Jonathan Hsu, a Partner at Tribe Capital. Tribe Capital is an investor in Teampay.
When Tribe Capital made the decision to invest in Teampay, one of the things that excited us most about the product was its potential to completely transform the finance function within organizations.
If the finance team is no longer wasting valuable cycles on high-effort, low-value tasks like manually inputting purchase data, that’s time they get back to invest in other high-value, high-impact tasks that advance the company toward its goals.
One of the biggest ways that a freed-up finance team can increase their impact their company is by working cross-functionally with other segments of the business, putting their analytic skills to work to answer big-picture questions that the team doesn’t have the time or the skill set to answer themselves.
The vision and product that Teampay has created wasn’t all that impressed the Tribe Capital team: from an investor perspective, we were also very impressed with the pattern of product-market fit in the business. One of the techniques that we use to evaluate that product-market fit is called cohort analysis.
Cohort analysis is also a prime example of the kind of high-level strategic analysis finance teams can take on. In this post, we’ll look at what makes cohort analysis a powerful method for gaining a clearer window into how people are spending with your business. We’ll look at where typical lifetime value (LTV) analysis falls short, and then explore three different visualizations that provide a higher-definition picture of customer LTV.
Cohort analysis: A higher-definition approach to modeling LTV
Customer lifecycle is a big black box for most companies. There’s a loose formula that most people are familiar with, but its usefulness is extremely limited, and it doesn’t go nearly far enough in helping business leaders really understand how their customers engage with their products.
Typical lifetime value (LTV) descriptions tend to look like this:
LTV = m * r / (1+ d-r )
In other words, lifetime value is the product of the contribution margin (m) and the retention rate (r), divided by one plus the difference between a discount rate (d) and the retention rate.
There are a few problems with this model — particularly for earlier stage companies like the ones we work with at Tribe Capital:
- It assumes that retention is constant, both across cohorts and across the lifetime of an individual customer.
- It assumes that unit economics are constant across cohorts and customer lifetime
- It assumes that both retention and unit economics are constant enough over long time periods that the discount rate makes sense.
For younger companies in particular, these assumptions seldom hold up. Retention varies wildly as the product evolves; cohort sizes are tiny and inconsistent, frequently measured in weeks or months. The unit economics of the business aren’t even fully settled.
Small companies need a more flexible, more nuanced model to get to the bottom of how their customers are spending money — and that’s exactly what the finance team can help to uncover with cohort analysis.
Three methods for tracking how LTV is trending in your business
There are three main ways to visualize cohorts — each of which provides a different vantage point on customer spending behavior:
- Plot the cohorts over time, which allows you to understand whether customer spending increases, decreases, or stays constant.
- Plot the cohorts against each other as they age, which allows you to compare behavior between cohorts and notice whether newer cohorts are spending more, less, or the same as older ones.
- Plot the cohorts as a heat map, which combines both approaches and provides a “multi-dimensional” picture of how customers are spending.
Method 1: Plot of the cohort over time
The first option for modeling cohort analysis visually is to plot the LTV of various customer cohorts as a function of time.
In this graph, customers are grouped together by the month they first paid revenue to the company. The x-axis shows the number of months elapsed since the customers first paid (i.e. how old the cohort is.) The y-axis shows the cumulative revenue per customer (i.e. LTV) with each line (i.e. the legend) representing a different monthly cohort. The dashed line delineates the weighted average of all cohorts at a given cohort age, weighted by cohort size.
To interpret the graph, we simply look at the shape of the lines. There are four buckets that the behavior of each cohort can fall into:
- Flat LTV, where the cohort spends once up front and then never again. This is not necessarily a bad thing, provided the flat LTV is a value high enough to be very profitable. Luxury purchases, such as boats and sportscars, likely exhibit this type of behavior.
- Sublinear LTV, where the cohort keeps spending, but the amount of spend decreases over time. This model describes most businesses. Say you have a favorite cookware brand: you load up on all of your cookware needs during your first trip, and you may buy one or two things here and there after that. But you’re most likely not going to repeat that initial upfront spending spree again.
- Linear LTV, where the cohort spends consistently over time. This behavior is typical of most subscription businesses, like Spotify and Netflix, where customers sign up for the service and then pay the set subscription amount on a regular basis until cancellation.
- Super-linear LTV, where the cohort spends more as their time with the company goes on. The prototypical super-linear example is Amazon. Customers order one item and have a positive experience, so they go on to by more. Eventually they decide their frequency of use justifies the cost of Prime, so they sign up. Then they buy more, and maybe add an Amazon Echo or a Ring.
At Tribe Capital, we tend to look for our businesses to exhibit LTV behavior that is either linear or super-linear in at least some cohorts of customer. If a business has flat or sublinear LTV, it tells us that customer acquisition costs associated with that business are likely to be higher, and it doesn’t have the kind of margins that make for an attractive investment opportunity.
For businesses that are using cohort analysis for their own internal purposes rather than as an investment evaluation tool, the considerations may be different:
If customer spend is declining over time, it may be a sign that something is broken with your product and needs to be fixed.
If newer cohorts are spending less over time than older cohorts did, it may be a sign that a more recently introduced feature is not connecting with customers
The important part is to get the higher-definition visibility into customer behavior, so that you can then decide where to go from there.
Method 2: Plot of the trend across cohorts at a fixed age
The second way to visualize cohort data is to focus on the trend at a fixed age. This effectively reverses the x-axis and the legend of the first graph: now cohort month is the x-axis, and each line represents a cohort age (in months).
This allows us to compare spending between cohorts and see whether one group spends more at a similar point in their lifetime. The cohort size (represented in grey) is included because it allows us to see how the cohort size maps against the buying behavior within the cohort.
Looking at the graph, we can see that from July 2017 to July 2018, the average customer LTV after 3 months (LTV Period = 3) increased from $1.2K to $1.8K; however, since July 2018 it has steadily decreased. Meanwhile, cohort size has increased, which signals a problem within the business: acquiring a lot more customers isn’t necessarily great if you’re making so much less money on them that you end up in the same place or worse.
This suggests that the business is likely suffering from a dropoff in product quality or customer experience as the company scales. Equipped with this information, businesses will likely want to conduct an audit of the company’s core functions — in particular product, marketing, and customer support — in order to pinpoint the source of the problem, correct it, and get the business back on track to more consistent and profitable customer behavior.
Method 3: Heat map
The third method teams can use to visualize cohort behavior is a heatmap. This method makes it easy to look across both cohorts and cohort age, while obscuring the shape of the curve.
Heatmaps represent age effects (e.g. contract renewal) vertically; single cohort effects (e.g. a big marketing spend for one month) horizontally; and fixed calendar time effects (e.g. seasonality) diagonally.
Here is the same revenue LTV data from above, presented as a heatmap:
Reading the graph from left to right, we can see clearly which cohorts became outliers in either direction. Reading the graph from top to bottom, you can see the decrease in revenue retention and revenue LTV per customer in newer cohorts (as well as a slowdown in overall cohort size growth).
This could be a signal of the company raising prices which results in the super-linear LTV growth among older cohorts while also simultaneously causing customer churn and lower LTVs in newer cohorts.
Embrace a growth-oriented approach to finance
A core component of the Teampay philosophy is that companies and their finance teams should be spending less time on overhead and more time on growth. In this post, we’ve looked at one way that finance teams can do that.
By applying strategic analyses like cohort analysis, companies can gain valuable insights into the behavior of their customers, and what they need to do to unlock the next level of growth. These are insights you do not get without a team with the bandwidth to chase the data down, visualize it, and interpret it. So get the low-value tasks off your finance team’s place and set them loose on your company’s data.
This post is adapted from a longer article from the Tribe Capital blog. You can view that article here.