Intercom's Net Revenue Retention Jumped 34 Points After a Pricing Change. HubSpot's Fell for Three Straight Quarters.
TL;DR: Intercom (now Fin) switched from per-seat to per-resolution pricing and watched net revenue retention climb from 112% to 146%. HubSpot, still seat-based, watched its own retention slide from 103.5% to 102% and lowered guidance to flat. That seems like a clean verdict for usage-based pricing — until Twilio's and MongoDB's usage-based numbers, reported the same quarter, show the same pricing model landing very differently with the market.
What happened to Intercom's (Fin's) net revenue retention after it switched pricing models?
Intercom spent years selling Fin, its AI support agent, the way it sold everything else: per seat. Growth stalled — five consecutive quarters of declining net new ARR, by outside accounts, before the company acted. In 2025 it tore out that model and moved to outcome-based pricing: $0.99 per resolution, charged only when the AI closes a ticket without a human touching it. Sacra's research on the company puts the retention effect at a jump from 112% net revenue retention to 146%, a figure that surfaced publicly through Intercom CFO Dan Griggs' interview on the Run the Numbers podcast and has circulated consistently since across several independent write-ups of the same podcast conversation.
The business results around that number are Fin's own, on the record. Fin passed $100 million in annual recurring revenue by March 2026, growing at roughly 350% year over year, with about 8,000 businesses using it. On May 12, 2026, the company renamed itself from Intercom to Fin — its ~1,400 employees now work for a company with the AI product's name, not the 15-year-old chat-bubble brand's. CEO Eoghan McCabe called Fin "our future, and the future of this new customer agent category... It's driving our renaissance and is about to be the largest part of our business." A month later, Salesforce signed a definitive agreement to acquire Fin for approximately $3.6 billion, with CEO Marc Benioff calling it a move to "complement Agentforce with powerful service agent capabilities." A retention number like Fin's aggregates every account's expansion and churn into one line — the account-level version of that same signal often shows up first in support-ticket data, long before it reaches a topline metric.
Why did HubSpot's net revenue retention decline in 2026?
HubSpot never changed its pricing model. It still sells primarily on seats, with a layer of AI "credits" added on top. Its net revenue retention for full-year 2025 was 103.5%, up slightly from 101.8% in 2024. From there it moved the other way: 103% in Q1 2026, down sequentially as expected but up half a point year over year, then 102% in Q2 2026, down a full point year over year. Management's own framing: "continued seat and credit expansion was offset by budget-related headwinds." The company's full-year guidance moved with it — from an expected 1-to-2-point year-over-year improvement in net revenue retention at the start of the year to, by the Q2 call, "roughly flat year-over-year."
Is usage-based pricing always better for retention than seat-based pricing?
Set next to Fin's jump, HubSpot's slide looks like a clear argument for tearing up the per-seat model. Two companies reporting in the same window, moving in opposite directions, on opposite pricing models — that pattern would be worth taking seriously if it were the whole picture. This isn't the same question as whether a blended NRR number can hide a wide split — it's whether the pricing model itself explains the split. Two other companies reporting the same quarter, both already usage-based, complicate it.
Twilio's Dollar-Based Net Expansion Rate reported at 116% for Q2 2026, with 75% of the company's revenue coming from usage-based fees. Analysts covering the print pushed back on the number itself: roughly 5 of those 116 points came from rising carrier fees passed through to customers, not from customers actually consuming more of Twilio's product. Strip the fee inflation out and the quarter-over-quarter acceleration was closer to 1 point. The headline retention number and the retention the business actually earned were not the same figure.
MongoDB's Atlas consumption business posted 122% net revenue retention for Q2 FY2027, reported alongside revenue of $771.8 million against a WallStreet estimate of $735.16 million and a record $127 million in net new Atlas dollars, per CFO Mike Berry. The company beat on every headline number that quarter. Its stock fell almost 12% anyway, on investor concern that Atlas growth itself, even at a healthy retention rate, wasn't accelerating fast enough.
That's a Contradiction, not a coincidence: Fin's 146% and Twilio's discounted 116% are both usage-based numbers from the same year, and the market read them completely differently — one as proof of a business transformation, the other as a figure that needed a footnote before anyone would believe it. MongoDB's 122% didn't even need a footnote and still wasn't enough to hold up the stock. The pricing model didn't decide how any of these four numbers got received. What customers were actually paying more for, and whether the market believed that growth would keep compounding, decided it.
What does this mean for a company choosing between pricing models?
Fin's story is real and dramatic, and it's also a company that used a pricing change to fix a growth problem that had gotten severe enough to threaten the business. HubSpot's decline is real too, and it's the ordinary friction of a mature, seat-heavy customer base facing budget scrutiny. Between those two poles, Twilio and MongoDB show that a usage-based number, by itself, doesn't come pre-certified as a good one. It's a version of the same lesson from Clari and Salesloft's $450 million combined ARR standing next to Gong's own $500 million: a bigger or cleaner-looking number doesn't automatically mean a stronger position. Knowing which of these four stories a single retention line is telling takes the intelligence layer underneath the number, not just the disclosure itself.
Caveats
Fin's own net revenue retention has never been disclosed directly by the company in an SEC filing or an official press release — Intercom, and now Fin, was private throughout this entire period, and the 112%-to-146% figure traces to Sacra's research and a CFO's podcast interview rather than a primary company disclosure. That's a materially different sourcing than the fully audited, SEC-disclosed figures used for HubSpot, Twilio, and MongoDB in this post, and it's presented here as the best available account of a real, well-documented business event (a pricing change, ARR jump to $100M+, and $3.6B acquisition), not as an audited number. The four companies here also don't compete with each other and aren't a controlled comparison — a support-automation product, a CRM, a communications API platform, and a database each carry a different natural retention ceiling before pricing model ever enters the picture. The claim in this post is narrower than "usage-based pricing wins." It's the same gap between an accurate number and the full picture that shows up anywhere metrics get read in isolation: pricing model alone, without knowing what's driving the number, tells you less than it looks like it does.
—Steven Rencher, Founder of Monadux
FAQ
What happened to Intercom's (Fin's) net revenue retention after it switched pricing models?
Intercom, now renamed Fin, moved its AI support agent from per-seat pricing to a $0.99-per-resolution outcome-based model in 2025. Research from Sacra, corroborated by CFO Dan Griggs' account on the Run the Numbers podcast, puts the resulting net revenue retention jump at 112% to 146%. The company's ARR passed $100 million by March 2026, and Salesforce agreed to acquire it for roughly $3.6 billion in June 2026.
Why did HubSpot's net revenue retention decline in 2026?
HubSpot's own earnings calls report net revenue retention falling from 103.5% for full-year 2025 to 103% in Q1 2026 to 102% in Q2 2026, with management citing budget-related headwinds offsetting continued seat and credit expansion. The company revised its full-year guidance from an expected 1-to-2-point improvement down to roughly flat.
Is usage-based pricing always better for retention than seat-based pricing?
Not reliably. Twilio's headline 116% net expansion rate for Q2 2026 included roughly 5 points from carrier-fee pass-through rather than real usage growth, and MongoDB's clean 122% net revenue retention for Q2 FY2027 didn't stop its stock from falling nearly 12% the same day. Both are usage-based companies whose numbers the market treated very differently from Fin's.
What is Twilio's net revenue retention and why is it debated?
Twilio reported a 116% Dollar-Based Net Expansion Rate for Q2 2026. Analysts noted that roughly 5 of those points came from rising carrier fees passed through to customers rather than genuine usage growth, putting the real quarter-over-quarter acceleration closer to 1 point.
What is MongoDB's net revenue retention?
MongoDB reported 122% net revenue retention for Q2 FY2027 (the quarter reported in September 2026), alongside a revenue beat and a record $127 million in net new Atlas consumption dollars. The stock still fell almost 12% that day on concern that Atlas growth wasn't accelerating fast enough.