Evaluating On-Demand Service Businesses

This post originally appeared on my blog.

A couple months ago, I was attempting to understand good retention for an on-demand service business. I was trying to get benchmarks for what good cohort retention rates would look like for businesses such as food delivery, groceries, laundry, and home cleaning. The goal of this post is to summarize my key learnings about how to evaluate on-demand, service based businesses

In my role at Naspers, we help a variety of companies with their core metrics. Having had less experience with on-demand service businesses, my goal was to get benchmarks for what was good and great in the space to help our portfolio companies evaluate their performance.

I ended up learning that the benchmarks are not all that valuable unless you are taking a number of other metrics and variables into account. My key learnings from the knowledgeable experts in the Greylock Partners Growth Community were the following:

• The cohort curve needs to flatten out at some point. If it continues to decline to 0, that’s much more challenging and a product market-fit question.
• It’s important that the LTV payback time of the cohort happens relatively quickly and predictably, preferably within the first 6 months
• The density of the business matters. Oftentimes, the way these sorts of businesses scale is largely dependent on the density of customers and the ability of a single service person to work across multiple customers in a given geography
• It varies quite widely based on the type of service based business.

Cohort curve must flatten. This was a great point made by Brian Balfour, and one that he has reiterated in many of his talks and blog posts (http://www.coelevate.com/essays/product-market-fit). His quote to me was “If the curve isn’t flat that means the users at some point don’t see the consistent core value. That is hard to fix. Shifting the curve up means they need to do a better job of getting initial users to experience and understand the core value.”

The cohort curve must flatten out at some point. Otherwise, you have more of a problem with long term retention and product/market fit, and it will be harder to build stackable cohorts over time.

Essentially, if you are trying to determine whether to focus on acquisition of new users and driving rapid growth and you see that your cohorts drop to zero in the near term, that’s the biggest problem you can fix for your business and should likely be at the top of both your product and growth roadmaps.

Further, in the case of evaluating a business, if you see a cohort curve that drops to zero, it’s a good indication that the product doesn’t have long term product market fit and that the company will always have to spend on acquisition or leverage other tactics to bring new customers in, which makes it a business that’s difficult to sustain in the long term.

LTV payback time of the cohort should happen relatively quickly and predictably. This was a great point made by Casey Winters of Pinterest. Essentially, looking at the curve in a vacuum doesn’t provide the context to understand whether or not the business has long term profitability. It’s important to understand the Cost Per Acquisition (CPA) as an essential element of any analysis of this sort.

For example, if you spend 100 dollars for a customer who retains for 6 months, but that customer only spend 10 dollars a month, while your retention might be decent, the LTV (Life Time Value) of a customer doesn’t justify the initial investment. Therefore, the business will either need to drive up the LTV of the customer or reduce the CPA before you could conceivably make deep investments in further growth of the business.

Alternatively, a business that spends $100 to acquire a customer with an LTV of $150 enables the business to make $50 for every customer and leverage that additional money to go out and find more customers (this is intentionally simplified to not include other costs in the model, but any model should look only at the profitability of the customer minus costs).

The density of the business matters. Ivan Kirigin from YesGraph shared some fantastic points here, such as “unit economics change with density. For example, the overhead of someone driving to pick up laundry is smaller per transaction if their route has more stops.” Essentially, the more users these types of businesses have within a geography or city is very important to driving down the cost of operating the business over time. If customers are highly distributed, then the per unit cost of running the business will be much higher than a business that has many customers in a smaller, serviceable area.

This also matters to the profitability point from before, as Ivan points out “it could be that driving acquisition helps each acquired user to be more profitable.” Essentially, getting more users in a specific geography may drive down the cost to service all other users in that service area, reducing the operational cost and increasing profitability for the business.

The type of business/industry will require different benchmarks and different retention curves. Everyone I talked with re-iterated this point, but I also got a lot of great insight from this blog post by Mahesh Vellanki at Redpoint: http://mahesh-vc.com/10-step-framework-for-evaluating-on-demand-startups/. Mahesh has a great outline in the frequency per week section that shows how much it may differ by industry, which may make a large impact on profitability and retention of customer longer term.

In general, I also found this blog post to be a pretty authoritative guide on how to think about on-demand startups in general, and highly recommend it as a read for anyone considering joining a company in the space, investing in one, or starting one.

Hope this was helpful. It was a great learning experience from me and provided me with a lot of insights to help evaluate the success of on-demand businesses.