Diagnose churn by cohort

1. The constraint and what graduates it

Item	At this stage
The constraint	Churn rises with the volume and the new product: there is a second product, and more customers coming through both it and the existing one
To graduate	Segment customers by cohort, and create activation points

Activation points are in section 6 of SOP 157. This page is the segmenting.

2. Segment by when they signed up

An easy first cut is by when customers signed up, or by the product they bought. By cohort means by sign-up time: the customers who joined in July, in August, in September, in October, in November.

It matters because, most of the time, when something goes wrong and churn rises, you can find what you changed, stopped or started doing at the time it rose. The practice described segments customers by when they signed up, to see what changed, or did not, or stopped or started, at the time those customers came in. That is one of the first things to do. A lot of times the first reaction to churn is that everything has to be fixed; first find what you stopped or started doing when those customers arrived.

What the cut can show, in the examples given:

• Something you changed. You might have added a second product, and maybe the customer service team was not on its game because it was not fully trained on it, and that happened to fall in July.
• The season. In one business, the cut could show that customers who signed up in January churned more than those who signed up in October. Maybe in January everyone is excited and signs up on the new year's energy, while the October sign-ups are simply committed to the result and did not need a new year to start.

Segmenting teaches you more about why customers left, based on when they signed up.

3. Read the grid

Lay the cohorts out by month of their life as customers: month one, two, three, four, five. The newest cohort has only reached month one, the one before it two months, then three, then four. The grid shows where customers fall away: in the example grid, a massive drop between month four and month five. Then ask what happens there, or fails to happen that should, and whether there is something you can do to get people over that hump. Once you have the data, these points become much easier to find.

4. Act before the drop

Look for something you can put in place one to two months before churn, as an incentive to stay. If churn happens at month six, do not give it at month six; give it at month five or month four. Month six is the average, which means some people are already leaving before it. Push people over the hump.

Look at both of these in each customer's life:

• When they signed up. That can set them up to succeed or fail at the point of churn.
• The point of churn. It can tell you whether something in the four to eight weeks before it led up to it.

People often decide to leave one or two months beforehand anyway, so you want to be a little early.

Flag: before the churn point, or after it. SOP 140, section 3, puts incentives just after major churn points: if most customers leave after the third month, something unlocks in month 4. Here, the incentive comes one to two months before the churn point: for churn at month six, give it at month five or month four. What each costs: an unlock after the point misses the customers who go before it or make up their minds early; an incentive before the point is handed over before the customer has stayed past it. This page does not settle which reading is right.

5. Segment by product

Segment the same customers by product as well. You might find that the highest point of churn for your level-one product is month four, while for your second-level product it is month 10. Look at churn without segmenting by product and you will solve the wrong problems, for the wrong people, for the wrong product. So always segment by product too, alongside when they signed up.

6. Segment by rep

Sometimes you even want to segment churn by rep:

• By sales rep, to see whether the churn comes from improper expectations set in the sale.
• By the assigned customer success manager, where your business gives each customer an assigned rep, account manager or account executive.

Each rep scoring every account red, yellow or green weekly, and a director coaching any rep whose churn runs higher, is SOP 273, sections 3 and 8.

7. Get specific

The more specific you can get with your customer data, the more likely you are to solve the actual reason someone is leaving. Do not change the entire product or service over one sales rep, one customer service rep, one bad month, or one poor month of onboarding. Get as specific as possible, so you do not make a sweeping change for a one-time event.

8. Where this sits

Whether churn is counted in customers or in revenue is SOP 130 — Separate revenue churn from logo churn. Once the segments show where customers drop, finding and building the activation point is the next step, on SOP 157 section 6.

9. What this page does not decide for you

• Whether an incentive goes before or after the churn point (section 4). This page does not settle which reading is right.
• What incentive to put in place before the drop. Not established on this page.
• How many months of data a grid needs. This page gives no figure for it.

10. What this page does not cover

Cancellation calls are SOP 132 — Run the cancellation call, and a survey of what customers value is SOP 131 — Run the keep-one and remove-one survey.