Retention Is Not One Number — Every Vertical Defines It Differently

Retention is not a number you optimise. It is the sum of every experience the user had with you.

The product. The listing quality. The menu photo. The delivery. The driver. The support agent. The return policy.

When retention drops, most teams open the CRM. They design a re-engagement campaign. They offer a discount. They A/B test the push notification copy. But the user stopped coming back because something broke. The campaign does not fix what broke. It just reaches someone who has already made up their mind.

Why retention means something different in every vertical

In classifieds, the supply side is everything. Retention is a seller who comes back to list more. Most sellers leave because their first ad got no replies — not because you failed to send them a push notification. The fix is engineering the conditions that make the first sale happen before they give up.

In ecommerce, retention is a buyer who places a second, third, and fourth order. If buyers do not come back, it is mostly because the product looked nothing like the photo, the delivery was late, or support took a week to respond. A discount code fixes none of that. The problem is upstream — in the seller quality, the fulfilment operation, the after-sales experience.

In food delivery, retention is not a monthly metric. It is how many times someone orders in a week. A food platform that retains well is one where the restaurant consistently sends the right items, the driver arrives on time, and support picks up the phone when something goes wrong. It breaks on any one of those three variables.

In ride-hailing, retention is the weekly active rider. It breaks when the ETA is too long, the driver rating is too low, or the price feels inconsistent. All three are operational problems — not marketing problems.

The team that owns retention does not exist

This is why retention is hard. It is not owned by marketing. It is not owned by product. It is not owned by operations.

It is the output of every function — working, or not working — together.

A bad driver experience is an operations problem. A product that looks nothing like the photo is a seller quality problem. A support ticket that takes a week is a customer service problem. A push notification that arrives three days after the user has already decided to leave is a CRM problem.

All of them show up in the same number: retention rate. None of them are fixed by working on the number directly.

What the data actually shows

When you build the right model across a business, retention signals are almost never in the place the team is looking.

The churn is happening 48 hours after a bad first experience — not 30 days later when the re-engagement campaign tries to catch the user. The customers who leave are not randomly distributed — they cluster around specific failure points. A particular category with low seller response rates. A restaurant with consistently late deliveries. A driver segment with below-threshold ratings.

The model does not just tell you that retention is low. It tells you exactly where the experience broke — and which team needs to fix it.

The teams that get retention right

The businesses with the highest retention rates are not running better CRM campaigns. They are holding every function to a higher standard of experience.

They measure whether the first transaction worked — not just whether it happened. They track the quality of supply, not just the quantity. They define retention differently for each vertical and build separate models for each definition.

Retention is the output. The experience is the input. Fix the input first.