Key Takeaways

• Contraction MRR counts partial losses from customers who stay, so a cancelled account belongs in churn MRR and folding it in leaves the churn line reading zero.

• Chargebee's glossary and Baremetrics both publish "Contraction MRR = Downgrade MRR + Cancellation MRR", contradicting Chargebee's own product docs, where cancellation is churn.

• On a $180,000 opening MRR book, $900 of seat cuts, an $800 downgrade, a $300 discount, and $650 of usage decline give $2,650 of contraction, a 1.47% rate. Adding one $1,200 cancellation reports $3,850, overstating contraction by 45%.

• ChartMogul's platform data puts contraction at roughly 20% to 30% of all MRR lost below $10k MRR, near 30% above $1m MRR, and as high as 45% of MRR lost at higher ARPA.

• A metered customer can contract for months without ever touching their plan, so reports keyed to plan-change events miss most contraction in usage-priced books.

What counts as contraction MRR and what counts as churn?

The customer's remaining balance settles it. If any recurring revenue survives from that account, the drop is contraction. If the last subscription ends and the balance hits zero, the whole amount is churn. Nothing sits in both places, and MRR movements is where both lines meet the rest of the waterfall.

Two of the most-read definitional pages get this wrong. Chargebee's glossary and Baremetrics both define contraction as downgrades plus cancellations, which leaves revenue churn with nothing to report. Chargebee's product docs use the correct split, so its glossary disagrees with its dashboard.

How do you calculate contraction MRR for a month?

Add up every MRR decrease from customers who kept paying, then divide by opening MRR for the rate. Monetizely puts it as revenue lost from existing customers who stayed, divided by opening total revenue, times 100. The dollar figure feeds the waterfall, the rate feeds gross revenue retention.

Here's one September on a $180,000 opening book:

That's a 1.47% contraction rate. Run the same month through the downgrade-plus-cancellation formula and you report $3,850, a 2.14% rate, with Kestrel's churn erased. One metric 45% too high, another missing.

What happens when a customer's consumption falls but their plan never changes?

They contract anyway, and no plan event fires to tell you. Metered charges recalculate every period, so a customer running 21% fewer tokens contracts by whatever that costs them, on the same plan and the same contract. I've watched teams call a month flat because nobody downgraded, while metered revenue fell 6%.

What that changes about the reporting:

• Contraction has to derive from the billed amount period over period. A subscription-change log shows zero.

• Seasonality reads as contraction. A retail-heavy book contracts every January and expands every November, and neither movement says anything about satisfaction.

• Consumption drops arrive first. Profitwell's research, cited by Monetizely, found contraction usually precedes full churn by 3 to 6 months, and usage decline is the earliest form of it.

• m3ter tracks a separate usage contraction rate, the share of accounts decreasing consumption, because a dollar total hides whether one whale slowed or forty accounts did.

• Committed-spend customers contract on paper but not in cash, since the commitment still bills. Split committed from overage revenue to keep that straight.

I keep two contraction lines: one for decisions a human made, one for consumption that moved on its own. Only the first belongs in a retention review.

Related terms

These are the neighbours worth reading next, in the order they connect:

• MRR movements places contraction alongside new business, expansion, churn, and reactivation, and reconciles the five to closing MRR.

• Revenue churn is the line contraction gets confused with, covering only accounts that left entirely.

• Gross revenue retention subtracts contraction and churn while ignoring every inflow, so it's the metric contraction hits hardest.

• Net revenue retention nets contraction against expansion, which is how a contracting book can still post a number above 100%.

• MRR vs ARR matters when you annualise contraction, because one month's drop times twelve overstates the year.

• Per-seat pricing produces the most common contraction event there is, a seat count coming down at renewal.

FAQ

What is a good contraction rate for SaaS?

Under 1% a month is elite. KeyBanc Capital's SaaS survey bands 1% to 2% as good and anything above 3% as concerning. Price point moves the bar, since ChartMogul's data shows contraction reaching 45% of all MRR lost at higher ARPA, where accounts are large enough to trim instead of leave.

Does a discount count as contraction MRR?

Yes. A new discount lowers the net recurring amount the customer pays, so the reduction posts as contraction in the month it takes effect. A discount expiring runs the other way and posts as expansion, which catches teams out when a promotional cohort rolls off and growth looks better than the product earned.

Should a paused subscription count as contraction MRR?

Yes, for as long as the pause lasts. Chargebee's product docs group pauses and active-to-trial moves with downgrades on the contraction side, because recurring revenue stops without the account closing. If the customer never comes back, the movement converts to churn when you cancel the subscription, not when the pause started.