Stripe Churn Rate SQL: Customer and Revenue Churn

Quick answer

Customer (logo) churn for last month. A customer counts as active when their MRR is above zero, which is Stripe’s own definition, so trials and unpaid sign-ups stay out.

-- Stripe Sigma (Trino)
-- Customer churn rate for last calendar month
with per_customer as (
  select
    customer_id,
    sum(case when local_event_timestamp < date_trunc('month', current_date) - interval '1' month then mrr_change else 0 end) as mrr_start,
    sum(case when local_event_timestamp < date_trunc('month', current_date) then mrr_change else 0 end) as mrr_end
  from subscription_item_change_events_v2_beta
  group by customer_id
)
select
  count_if(mrr_start > 0) as customers_at_start,
  count_if(mrr_start > 0 and mrr_end = 0) as churned_customers,
  round(100.0 * count_if(mrr_start > 0 and mrr_end = 0) / nullif(count_if(mrr_start > 0), 0), 2) as customer_churn_pct
from per_customer

Schema assumptions

Source	Tables used	Notes
Stripe Sigma	subscription_item_change_events_v2_beta (customer_id, currency, mrr_change, local_event_timestamp), subscriptions (id, customer_id, status, canceled_at, ended_at, cancel_at_period_end, cancel_at), customers (id, email)	mrr_change in minor units.
Stripe Data Pipeline	Same tables under schema STRIPE	Add a merchant_id filter if several accounts share the warehouse.
Fivetran	subscription_history (customer_id, status, start_date, trial_end, ended_at, _fivetran_active)	Keep _fivetran_active rows only (the current version of each subscription).
Airbyte	subscriptions (customer, status, created, canceled_at, ended_at, trial_end)	Timestamps are Unix epoch seconds; wrap them in to_timestamp() on Postgres.

Copy-ready SQL

1. Gross MRR churn and net revenue retention

How much recurring revenue last month’s existing customers took with them, and what they are worth now. Change: nothing. Rows are per currency.

-- Stripe Sigma (Trino)
with per_customer as (
  select
    customer_id,
    currency,
    sum(case when local_event_timestamp < date_trunc('month', current_date) - interval '1' month then mrr_change else 0 end) as mrr_start,
    sum(case when local_event_timestamp < date_trunc('month', current_date) then mrr_change else 0 end) as mrr_end
  from subscription_item_change_events_v2_beta
  group by customer_id, currency
)
select
  currency,
  sum(mrr_start) / 100.0 as mrr_at_start,
  sum(greatest(mrr_start - mrr_end, 0)) / 100.0 as lost_mrr,
  round(100.0 * sum(greatest(mrr_start - mrr_end, 0)) / nullif(sum(mrr_start), 0), 2) as gross_mrr_churn_pct,
  round(100.0 * sum(mrr_end) / nullif(sum(mrr_start), 0), 2) as net_revenue_retention_pct
from per_customer
where mrr_start > 0
group by currency

Example output (made-up data):

currency	mrr_at_start	lost_mrr	gross_mrr_churn_pct	net_revenue_retention_pct
usd	17150.00	610.00	3.56	101.20

2. Customer churn rate for the last 12 months

The trend, one row per month. Change: interval '12' month for a longer window.

-- Stripe Sigma (Trino)
with months as (
  select month
  from unnest(sequence(
    date_trunc('month', current_date) - interval '12' month,
    date_trunc('month', current_date) - interval '1' month,
    interval '1' month
  )) as t (month)
),
per_customer_month as (
  select
    m.month,
    e.customer_id,
    sum(case when e.local_event_timestamp < m.month then e.mrr_change else 0 end) as mrr_start,
    sum(case when e.local_event_timestamp < m.month + interval '1' month then e.mrr_change else 0 end) as mrr_end
  from months m
  cross join subscription_item_change_events_v2_beta e
  group by 1, 2
)
select
  month,
  count_if(mrr_start > 0) as customers_at_start,
  count_if(mrr_start > 0 and mrr_end = 0) as churned_customers,
  round(100.0 * count_if(mrr_start > 0 and mrr_end = 0) / nullif(count_if(mrr_start > 0), 0), 2) as churn_pct
from per_customer_month
group by month
order by month desc

Example output (made-up data):

month	customers_at_start	churned_customers	churn_pct
2026-09-01	412	13	3.16
2026-08-01	398	15	3.77

3. Customers who cancelled in the last 30 days

Who left, so you can email them. Change: the 30-day window.

-- Stripe Sigma (Trino)
select
  c.email,
  s.id as subscription_id,
  s.canceled_at,
  s.ended_at
from subscriptions s
join customers c on c.id = s.customer_id
where s.canceled_at >= current_date - interval '30' day
order by s.canceled_at desc

Example output (made-up data):

email	subscription_id	canceled_at	ended_at
jo@startup.example	sub_1Q…	2026-10-09 14:02	2026-10-31 00:00

4. Customers scheduled to cancel

Already cancelled, still paying until the period ends. This is your save list. Change: nothing.

-- Stripe Sigma (Trino)
select
  c.email,
  s.id as subscription_id,
  s.cancel_at as ends_on
from subscriptions s
join customers c on c.id = s.customer_id
where s.status = 'active'
  and s.cancel_at_period_end
order by s.cancel_at

5. Customer churn rate from Fivetran tables (Postgres)

Same definition, built from subscription dates because Fivetran has no MRR change log. A customer is active on a date if any of their subscriptions had started, finished its trial and not ended. Change: the stripe schema name.

-- PostgreSQL (Fivetran Stripe schema "stripe")
with bounds as (
  select
    date_trunc('month', current_date) - interval '1 month' as month_start,
    date_trunc('month', current_date) as month_end
),
per_customer as (
  select
    s.customer_id,
    bool_or(coalesce(s.trial_end, s.start_date) < b.month_start
            and (s.ended_at is null or s.ended_at >= b.month_start)) as active_at_start,
    bool_or(coalesce(s.trial_end, s.start_date) < b.month_end
            and (s.ended_at is null or s.ended_at >= b.month_end)) as active_at_end
  from stripe.subscription_history s
  cross join bounds b
  where s._fivetran_active
  group by s.customer_id
)
select
  count(*) filter (where active_at_start) as customers_at_start,
  count(*) filter (where active_at_start and not active_at_end) as churned_customers,
  round(100.0 * count(*) filter (where active_at_start and not active_at_end)
        / nullif(count(*) filter (where active_at_start), 0), 2) as customer_churn_pct
from per_customer;

6. Cancellations per month from Airbyte tables (Postgres)

Airbyte keeps Stripe’s Unix timestamps, so convert them first. Change: the airbyte_stripe schema name.

-- PostgreSQL (Airbyte Stripe schema "airbyte_stripe")
select
  date_trunc('month', to_timestamp(canceled_at)) as month,
  count(distinct customer) as customers_cancelled
from airbyte_stripe.subscriptions
where canceled_at is not null
  and livemode
group by 1
order by 1 desc;

Example output (made-up data):

month	customers_cancelled
2026-09-01	14
2026-08-01	17