NEW Churn intelligence for token-based SaaS

Catch churn
before Stripe does.

Retainly reads token usage and billing signals, scores every customer's churn risk with ML, and runs the retention play automatically — while the revenue is still saveable.

74%
of at-risk MRR is recoverable
15 min
to first churn score
2
integrations, zero code
Built for usage-based teams running on
◈ Stripe ▲ Supabase ◇ Postgres ✦ Resend ⬡ Cal.com
The problem

Most paying customers leave quietly.

In a token model there's no slammed door — usage just fades. By the time Stripe logs the cancellation, the moment to act is long gone. The signals were there the whole time. Nobody was watching.

01
Token usage drops 60% in a week
Nobody sees it. No alert fires.
02
Payment slips 3 days past due
No one reaches out. Card never updated.
03
Hasn't logged in for 2 weeks
No trigger. No re-engagement.
04
Subscription gets cancelled
You find out in the dashboard. Too late.
Avg monthly churn · AI SaaS
5–8%
On $500k ARR, that's
$25–40k
MRR lost — every month
most of it saveable
How it works

Three steps, running in the background.

Connect your stack once. Retainly handles collection, prediction, and action on a loop — so retention happens without anyone remembering to do it.

STEP 01

Collect the signals

Stripe gives billing behavior — payments, dunning, plan changes. Supabase gives product activity — token burn, logins, feature use. Synced continuously, normalized in one place.

STEP 02

Predict the risk

A gradient-boosting model scores every customer 0–100, with the top reasons and a predicted churn window. It learns your data over time — not a generic benchmark.

STEP 03

Run the retention play

Cross your threshold and the AI agent acts — a personalized email, a Stripe coupon, a call invite. Fully automatic, or drafted for one-click human approval.

Churn Risk Score

A number you can act on — with the why.

Every customer gets a live 0–100 score, the three factors driving it, and the window they're likely to leave in. No black box — the reasons are the whole point.

Token activity 0.92
Billing status 0.78
Engagement 0.64
Plan context 0.41

FEATURE WEIGHTS · per-customer retrained · XGBoost

NV
Northwind Vectors
growth · $390/mo · 14mo
FLAGGED
82risk
● High risk
74% likely to churn
within 14 days
Top factors
1 Token usage down 71% over 14 days
2 Last login 11 days ago
3 Payment failed once this cycle
AI retention agent

It doesn't just flag. It acts.

Cross your risk threshold and the agent runs the right play — written from the customer's actual profile and the reason they're leaving. You stay in control of when and how.

Personalized email — context-aware copy tied to the risk reason, not a generic blast.
Auto discount — applies a Stripe coupon to the next invoice when it'll save the account.
Call invite — drops a Cal.com link for high-value accounts worth a human touch.
Approval mode — agent drafts, you confirm. Review every send before it goes out.
Integrations

Two connections. No SDK.

OAuth into Stripe, map a couple of Supabase tables, and Retainly starts scoring your book in minutes. Map fields by hand — you stay in control of what's read.

Stripe
Billing & payment behavior

Payment history, dunning, refunds, subscription status and MRR per customer — read via OAuth, in real time.

past_duemrrcanceledtrialingrefunds
Connected · syncing
Supabase
Product & token activity

Token consumption by day, login frequency, feature usage. Map users and usage_events tables by hand at connect.

tokens_usedlast_seensessionsfeature_hits
Connected · syncing

Not on Supabase? Direct Postgres, PlanetScale & Firebase support on the roadmap.

FAQ

Questions, answered.

How fast do I get my first churn scores?
After you connect Stripe and map your Supabase tables, the first sync and scoring pass typically completes in about 15 minutes. You'll land on a dashboard ranking your current customers by churn risk, each with its top reasons.
Do I need enough customers to train a model?
No. Retainly starts on a shared baseline model that works from day one, then fine-tunes to your data as history accumulates. Smaller books still get useful, explainable scores immediately.
Will the AI agent email customers without my say-so?
Only if you want it to. You set the rules — which action fires at which score. Approval mode keeps a human in the loop: the agent drafts every email and offer, and nothing sends until you confirm.
We don't use Supabase. Can we still use Retainly?
Supabase is supported today. Direct Postgres, PlanetScale and Firebase connectors are on the near-term roadmap — if you're on a standard Postgres database, reach out and we'll prioritize it.
How is pricing structured?
You pay for the number of customers you track, not customers you save — predictable and scaling with your business. Plans run from Starter (200 customers) to Scale (5,000), with Enterprise for unlimited and white-label.
What about data privacy and GDPR?
Data is processed in-region with DPA agreements and data minimization by default. Emails can send from your own domain with proper SPF/DMARC so deliverability and trust stay yours.
Early access

Keep the revenue
you already earned.

Join the waitlist for early access. Be first to connect your stack and turn passive churn into recovered MRR.

No credit card 15-min setup Cancel anytime