What is a good retention curve for a startup?

A good curve is one that flattens to a stable plateau rather than decaying to zero. The exact plateau depends on your category — daily-use consumer products hold a higher percentage than infrequent tools — but the shape matters more than the number. A flat, non-zero plateau means a core of users keeps coming back, which is the foundation of durable growth.

Why do investors care about retention more than growth?

Because growth can be bought and retention cannot. A big acquisition budget can make almost any chart go up and to the right for a while. Retention shows whether users stay once they arrive — and only retained users compound. Investors have watched too many fast-growing companies with leaky retention stall, so they check the curve first.

How should I measure retention — blended or by cohort?

By cohort. Group users by when they signed up and track each group's return behavior over time. A blended average mixes old loyal users with new ones and can trend up simply because your oldest cohort survives, hiding a worsening product. Cohorts show whether newer users retain better than older ones, which is what you actually want to know.

The Retention Curve That Predicts Your Next Raise

If an investor could see only one chart before deciding whether to fund you, most would pick the retention curve. Not growth, not revenue, not your traffic. Retention, because it is the one number that is almost impossible to fake and that predicts everything else. This is how to read your own curve the way a good investor reads it — so you fix what it tells you before someone else points it out.

What the curve actually is

A retention curve plots the percentage of a group of users who are still active some time after they first showed up. Day 1, day 7, day 30, and onward — or week 1 through week 12, depending on how often your product is meant to be used. It answers the only question that matters for durability: when new users arrive, do they stay?

Everything good about a business compounds off that answer. Word of mouth, expansion revenue, payback on acquisition spend — all of it assumes users stick around. If they do not, you are not building a compounding asset. You are running a treadmill that gets more expensive the faster you go.

The shape that matters: does it flatten?

There are really only two outcomes, and you can tell them apart at a glance.

A decaying curve falls and keeps falling toward zero. Every cohort eventually churns out completely. This means you do not have retention — you have a series of one-time visits. Growth here is renting users: the moment you stop paying to acquire, the business shrinks.
A flattening curve drops at first — some people were never a fit, and that is normal — and then holds at a stable, non-zero plateau. That plateau is your real business. It is the fraction of users for whom the product genuinely stuck. Everything you build stacks on top of it.

The height of the plateau depends on your category. A product people are meant to use every day should hold a higher percentage than one they need a few times a year, so compare yourself to the right shape, not to someone else’s number. But the existence of a plateau is non-negotiable. No flattening, no durable business.

Acquisition sets how fast users pour in. Retention sets how many stay in the bucket. You cannot fill a bucket with holes by pouring faster.

Measure by cohort or fool yourself

The most common way founders lie to themselves is by looking at a single blended retention number across all users. Blended numbers drift upward for a bad reason: your oldest, most loyal users survive and come to dominate the average, so the headline improves even as the product gets worse for everyone new.

Plot it by cohort instead. Take everyone who joined in a given week or month, follow that group over time, and lay the cohorts side by side. Now you can see the thing that actually predicts your future: are newer cohorts retaining better than older ones?

If newer cohorts flatten higher, the product is getting stickier. That is the strongest possible signal — it means the work you are doing is compounding.
If every cohort decays the same way, your improvements are cosmetic. You are shipping features that do not change whether people stay.
If newer cohorts are worse, growth is outrunning quality — you are acquiring less-qualified users faster than the product can hold them.

Fix the flattening before you spend on growth

The instinct when growth stalls is to spend more on acquisition. If your curve does not flatten, that is the worst thing you can do — you are buying users into a leaky bucket and paying for the privilege of watching them leave.

The order of operations is:

Find the plateau. Where does the best cohort settle, and how high?
Study the survivors. What is different about the users who stayed? What did they do in their first session that churned users never did? That behavior is your activation target.
Get more new users to that behavior faster. Most retention gains come from the first session, not from win-back campaigns. Redesign onboarding so more people reach the sticky moment before they lose interest.
Only then, pour on acquisition. Once newer cohorts flatten higher than older ones, every marketing dollar compounds instead of leaking.

What this predicts about your raise

Bring cohort retention curves to a fundraise and you change the conversation. A flattening curve, with newer cohorts improving, tells an investor that your growth is real and that their money buys compounding, not a temporary bump. A decaying curve tells them the opposite, no matter how impressive your top-line growth looks — and experienced investors have been burned enough times to check.

So read your own curve first. It is the most honest feedback your product will ever give you, and unlike an investor, it will tell you months before the round.