every saas pitch deck has a slide that reads "nrr: 120%." the partner nods. the founder moves on. the slide is true. it is also incomplete in a way that decides whether the next round closes.
nrr — net revenue retention — counts what last year's customers paid this year, including their expansion. expansion is mathematically allowed to mask everything else. a company can lose 35% of its logos, contract another 10%, and still print 120% nrr if the survivors doubled. the deck looks great. the underlying business is a floor that's falling out from underneath itself.
the number that doesn't lie is gross dollar retention. same denominator. expansion stripped out. just the dollars last year's customers paid you this year, capped at what they paid before. it answers a different question: not "how much money did the cohort produce" but "how much of the cohort did you keep."
what gdr actually measures
gdr = ending arr from a starting cohort, capped at the starting amount, divided by the starting cohort arr. if a $1m cohort produced $900k this year (no logo lost, but $100k of contraction), gdr is 90%. if half the logos churned and the rest expanded to fully replace the dollars, nrr is 100% and gdr is 50%.
both companies enter year one at $4m arr from the same cohort. company a runs a 130% nrr engine on top of a 72% gdr floor — expansion is doing all the work, the base is collapsing underneath. company b runs 108% nrr but holds 92% of its dollars. three years later the cohort runs are $1.5m and $3.1m, more than 2x apart, from the same starting point.
the partner who has seen one of these blow up will spot the gdr gap inside a week.
the bands that matter
world-class saas runs 92-95% gdr. it's the band where renewals are quiet, contraction is light, and the product is sticky enough that customers don't go shopping every renewal. series b companies usually need to be here by their next priced round if they want a clean term sheet.
85-92% is the series a bar. acceptable, defensible, has a clear path to the higher band as you mature.
75-85% is the warning zone. the next investor pass is going to ask whether the cohort is improving year over year. it can be, but you'll spend two diligence sessions defending it.
below 75%, the product isn't sticky. no nrr number rescues this. when expansion slows — and expansion always slows — the base will surface as a hole in the boat.
why founders avoid running it
three honest reasons.
first, the calculation requires a clean monthly close. you need to identify each customer in the cohort by start date, track their actual paid revenue per period, and cap at their starting amount. if your books are 30 days late or your customer ids don't reconcile between stripe and your crm, gdr is unreliable.
second, gdr is unflattering. nrr can be carried by a single big expansion. gdr is a survey of every account in the cohort and treats them all as equally important. that makes it a much harder number to spin.
third, gdr forces the conversation about why customers are leaving — and most founders don't want to have that conversation until it's structurally too late. by the time gdr is the headline, the product team is 18 months behind on the work that would move it.
what to do with the number
run gdr by cohort quarterly. plot the trend. if the latest cohort's gdr at month 12 is better than the cohort 12 months older at the same age, the product is improving. if it's worse, every other metric on the dashboard is misleading you. the cohort-age comparison is what investors will run themselves in diligence; running it yourself first means you walk in already knowing the answer.
a healthy operating cadence: the cmo and head of cs report gdr at the monthly business review, not just nrr. the question the team asks every month: which named accounts contracted, why, and what's the path back. the answer is rarely "more sales." it's almost always "the renewal conversation should have started 90 days earlier."
how zift handles this
zift computes gdr by cohort, by month, every fifteen minutes against the stripe + bank data it already ingests. on monday morning the briefing names the cohort that moved, the dollar amount of contraction, and the named accounts inside it. the partner-facing version of the same number sits in the board pack without anyone having to rebuild it.
if you're a finance lead at a series b team running multi-entity consolidation or multi-currency renewals, zift handles that too.
every nrr number is a story. gdr is the one that decides whether the story has an ending.