How to prioritize work with KPIs
Part 2 of 2 (first part)
In my recent post, I’ve talked about KPI trees and how product teams can use them. The most obvious application is prioritization between different projects. For example, KPI tree can help answer the question: “We have two potential projects of the same difficulty. One will improve conversion by 0.5% another will decrease customer contact rate by 10%? Which one should we do first?”.
This time, I’ll dive deeper into project impact comparison in wholly different areas of the product, share product growth formula that product teams should maximize and answer the above question. But let’s start with the most common KPI tree template.
KPI tree template
After building KPI trees with many different teams, I noticed that most of them have a similar structure. Root has two branches representing a total value of provided services and profit per value unit. The type of the product defines value, for example, it can be a volume of transferred money for TransferWise, a total number of the search sessions for Google, total number or duration of driving session for the team working on a product for drivers at Uber. The value branch then splits into branches representing a number of active users and value per one active user. Profit branch in its term breaks into revenues and costs per unit of value. In result, most common KPI tree could look something like the one shown in the image below.
We can write down this KPI template with the set of formulas:
Product profit =
= Total value * Profit per value unit=
= (Active users * Value per user) * (Revenue per value unit — Cost per value unit).
Value per user = interaction per user * average value per interaction.
By value I mean value that the product brings to the customer, not the value customer bring to the company. They are related, and the more valuable product for the customer the higher profit user can generate for the company: Monthly profit per user = Monthly Value per user * (Revenue per value unit — Cost per value unit)
Active Users (AU) are customers, who will make at least one interaction with the product in the future. This metric is challenging to measure, and I will tell more about how to work with this metric it in future. Meanwhile, there is a widespread proxy for Active Users that is much easier to measure Monthly Active Users (MAU), users who made at least one interaction with the product in last month. MAU tend to have a linear correlation with AU, i.e., Active users equal to MAU multiplied on some constant coefficient.
Let me introduce some abbreviations for metrics that I will use a lot further:
Monthly new users = MNU (Customers who do their first interaction with the product during the month)
Monthly churned users = MCU (Customers who stop using the product during the following month)
The main prioritization formula
Comparing the impact of acquisition (MNU) and retention (MCU) might be tricky, and if we compare them directly, we might come to the wrong conclusions. Let’s say we have two projects of the same size; the first one will bring 100 additional new users per month, while another will save 100 customers from the churn. Which one should we prioritize? 100 new customers are the same as 100 saved from churn so should not have the difference which project we choose? Not really.
We might be in the first days of the product with no active user base and having 100 customers churning per month could mean that customers are not sticking to the product and in this situation bringing additional 100 customers will not change much, as they will leave product at the same month. Alternatively, we can talk about the mature product with the substantial loyal user base but small acquisition channels; and adding 100 new customers per month will make them stay for years with the product and bring a lot of value. In the first case, we clearly should focus on retention while in the second on the acquisition. But how to understand what is the best option in each particular case?
The goal is to maximize the value in the long term which is proportional to the number of Active Users product will have at that time. But how we can know what will be the number of active users in the long term? There is one trick that can help us estimate it. In the long term product will get to the balanced state, equilibrium point when new customers number equal to churning ones, MNU = MCU. Monthly Churned users connected to Active users thought the monthly Churn Rate (CR) and the relationship for most products could be well approximated by the simple formula MCU = AU * CR. Which gives us:
MNU = MCU = future AU * CR =>
future AU = MNU/CR
Product team’s job is to maximize product value in the long term which is equal to the number of active users in the long term multiplied by the profit per user. If we write it down we will get the main prioritization formula:
Max (VU*AU) = Max (VU*MNU/CR)
Product team’s job is to optimize future value or maximize New Users / Churn rate * Value per User
Note: I’m not touching network effects and virality in this article. In few words, they set a positive correlation between Active Users and multipliers in the prioritization formula, first of all, Monthly New Users. In other words, the more Active users we have, the more New users and potentially less churn and higher value per users we are getting. That not only could increase the speed at which system gets to the balanced state but also make it higher.
Example
Let’s use these formulas to answer the question we set in the beginning. What’s better: improving conversion by 0.5% or decreasing customer contact rate by 10%?
Conversion rate has a direct correlation with Monthly New Users so improving conversion rate will rise MNU by 0.5%. From another side, let’s say customer support cost state for 10% of our total costs. Also, let’s assume our profit margin is 30% of our costs. In this case drop of contact rate by 10% will lead to 1% cost reduction. This 1% of costs will become part of margin, or approximately 30% will change to 31%, which is equal to 3.3% growth of margin. An increase of margins proportional to the growth of Monthly Value per User (MVU). Rising MVU by 3.3% will have an almost 7x higher impact on MVU*MNU/CR compared to rising of MNU by 0.5%, so we should work on contact rate.
Monthly Value per User and Monthly New Users are quite simple to measure. Monthly user value composed of Cost and Revenue per interaction and Number of interaction per active user. All this metrics are available in almost any analytical system. Churn rate is the most tricky to measure, and I’ll focus on it in one of my next posts.
