Calculate Stripe Net Revenue Retention (NRR) from an Invoices CSV

How to export your invoice history from Stripe for NRR

To calculate trailing 12-month (T12M) NRR, your CSV export must span at least 13 completed months.

1. Open the Stripe Dashboard and go to Billing > Invoices.
2. Click Export, set a custom date range covering at least 13 to 24 months, and select All columns.
3. Download invoices.csv. Drop it above for a first chart, or import it into the xmetric app to build the NRR view.

Stripe CSV columns used for NRR and GRR

• customer (or Customer ID): unique Stripe customer ID (cus_...), used to track expansion and contraction across subscriptions.
• subscription: filters out one-time non-recurring invoices (subscription must not be empty).
• status: filtered to paid.
• created (UTC) (or Created date (UTC) / period_start (UTC)): maps revenue to the billing month.
• amount_paid: amount collected on the invoice.

Tableau LOD and Table Calculation formulas for NRR vs. GRR

• Net Revenue Retention (NRR): (Starting MRR + Expansion - Contraction - Churn) / Starting MRR for the cohort of customers who were active in the starting period. Can exceed 100%.
• Gross Revenue Retention (GRR): (Starting MRR - Contraction - Churn) / Starting MRR, capping each customer's ending MRR at their starting MRR so expansion is excluded. Cannot exceed 100%.

1. Customer MRR by Month (FIXED LOD)

// [Customer Monthly MRR]
{ FIXED [customer], DATETRUNC('month', [created (UTC)]) :
    SUM(IF [status] = "paid" AND NOT ISNULL([subscription]) THEN [Normalized MRR] ELSE 0 END)
}

Test this LOD in the formula playground, then build the monthly view on shelves in the xmetric app.

2. Trailing 12-Month NRR and GRR (SQL equivalent)

The equivalent SQL for the customer-level lookback:

WITH monthly_customer_mrr AS (
  SELECT
    customer AS customer_id,
    DATE_TRUNC('month', CAST("created (UTC)" AS TIMESTAMP)) AS bill_month,
    SUM(
      CASE
        WHEN interval = 'year' THEN amount_paid / 12.0
        ELSE amount_paid
      END
    ) AS mrr
  FROM stripe_invoices
  WHERE status = 'paid'
    AND subscription IS NOT NULL
    AND subscription <> ''
    AND amount_paid > 0
  GROUP BY 1, 2
),
paired_12m AS (
  SELECT
    start_m.bill_month + INTERVAL '12 months' AS report_month,
    start_m.customer_id,
    start_m.mrr AS starting_mrr,
    COALESCE(end_m.mrr, 0) AS ending_mrr,
    LEAST(start_m.mrr, COALESCE(end_m.mrr, 0)) AS retained_gross_mrr
  FROM monthly_customer_mrr start_m
  LEFT JOIN monthly_customer_mrr end_m
    ON start_m.customer_id = end_m.customer_id
   AND end_m.bill_month = start_m.bill_month + INTERVAL '12 months'
)
SELECT
  report_month,
  COUNT(DISTINCT customer_id) AS base_cohort_customers,
  ROUND(SUM(starting_mrr), 2) AS starting_mrr_12m_ago,
  ROUND(SUM(ending_mrr), 2) AS ending_mrr_today,
  ROUND(100.0 * SUM(ending_mrr) / NULLIF(SUM(starting_mrr), 0), 1) AS annual_nrr_pct,
  ROUND(100.0 * SUM(retained_gross_mrr) / NULLIF(SUM(starting_mrr), 0), 1) AS annual_grr_pct
FROM paired_12m
WHERE report_month <= (SELECT MAX(bill_month) FROM monthly_customer_mrr)
GROUP BY 1
ORDER BY 1;

Common NRR calculation errors on Stripe exports

• Including new customers acquired during the window: NRR strictly measures the cohort of customers who existed at the start of the period (start_m.mrr > 0). If a new customer signs up in Month 6, their revenue is excluded from the denominator and numerator until they have 12 months of history.
• Letting one customer's expansion hide another customer's churn in GRR: In Gross Revenue Retention (GRR), you must apply LEAST(starting_mrr, ending_mrr) at the individual customer level before summing across the cohort. Taking LEAST at the aggregate level produces a wrong number.
• Duplicate Customer IDs when a company changes billing email: If an existing customer creates a second Stripe cus_... record when upgrading to an annual contract, Stripe treats them as one churned customer and one new customer. If you sell B2B, create a calculated field in the xmetric app that groups by email domain (SPLIT([Customer Email], '@', 2)) to merge duplicate Stripe customer records.

Run NRR audits from Claude or ChatGPT

Every calculation in xmetric is a deterministic command. An MCP server for Claude and ChatGPT is in development, so you will be able to ask your AI assistant for NRR and GRR numbers from a Stripe CSV without relying on LLM arithmetic. Live connections to Postgres and BigQuery are also coming.