Table of Contents
Braze’s Q4, like Klaviyo’s results, begs the question: If SaaS is dead, then why are both of them pushing 30% YoY growth?
Braze is saying two things clearly: First, make AI usable inside the workflow, not as a bolt-on. Second, MarTech still gets a budget when it is tied to first-party data and measurable outcomes.
One proof point regarding the AI claim: CEO Bill Magnuson said Operator and Agent Console shipped to GA months earlier and that more than two-thirds of customers were actively using Operator within a few weeks.
Just the numbers: Q4 revenue $205.2M, +28% YoY, +8% QoQ. FY2026 revenue +24% YoY. DBNR 109. RPO $1.0B (+30% YoY). cRPO $642M (+27% YoY). For a refresher on Bookings vs. RPO vs. ARR, see my older Braze analysis here.
The Good
Demand is stable and commitment is up.
Dollar-based net retention (DBNR) was 109. It’s still down from the high of 111%, but improving, which is nice.
Here’s the thing: DBNR is a proxy for likeability. If you didn’t lift a finger to sell to existing customers, would they buy enough to offset those who left you or are spending less?
100% is table stakes. At 100%, revenue is flat, but there’s no intrinsic incentive for the customer to buy more. You want it in the 110%+ category because it signals a growing product. They like you, and they want more of you.
RPO crossed $1B (+30% YoY), and cRPO was $642M (+27% YoY). The takeaway is that the backlog is growing.
Customers spending $500K+ annually rose to 333, and they now represent 64% of ARR (up from 62% last year). That’s the cohort that tends to drive expansions and attach new products first.
Braze also initiated a $100M share repurchase authorization. It’s not a capital-return story like Twilio, but it is a cohort signal. Like LINK and Sinch, Braze is using buybacks to communicate confidence.
The Interesting
The best “why you pay” moment was the build-versus-buy story.
A large customer spent eighteen months and more than $10M trying to replace Braze and only got to about one-third of the functionality. Bill’s point was that customer engagement at scale is hard to recreate because it is an integrated systems problem: data ingest, consent, orchestration, channel execution, and reliability.
On AI, Bill’s framing was straightforward: Operator and Agent Console are meant to get customers to use more of what they already bought. He described marketers using Operator and Agent Console to create campaigns, build canvases, and stand up agents, then learn by watching the system act. The claim is that this increases confidence and usage in higher-consequence workflows.
Decisioning Studio is Braze’s most ambitious proprietary intelligence claim. Bill framed it as reinforcement learning trained on customer data, paired with foundation models for reasoning.
The Unknown
A tech-wide story: Except for the companies truly selling AI products, for the rest, AI is everywhere, except in the numbers (see my RCS/AI comparison here). CFO Isabelle Winkles said premium messaging has been the main source of channel-mix pressure in recent years, and while newer AI products may be slightly better than company-average margin, they’re starting from a small base, so it takes time to show up.
They may be building something valuable, but it won’t show up in the income statement right away. If the product is sold on long contracts and the revenue is recognized over time, higher usage today won’t immediately lift reported revenue. So, I’d watch renewals, pricing, and cash, not excitement about adoption.
OfferFit is in the promising-product, messy-economics phase. Isabelle said margins are pressured because they still need to staff implementations and onboarding, and the path forward is more self-serve tiers over time. That implementation burden needs to drop.
For thoroughness: Braze disclosed a material weakness in IT general controls. I’m treating it as a likely growing pain for now, but it’s worth tracking until remediation is complete.
Finally
Braze’s quarter is easy to like if you believe in the core thesis: First-party data plus orchestration still wins, and AI increases the value of the system that can execute reliably at scale.
The work now is conversion. If Operator and Agent Console truly increase adoption, the next two places it should show up are retention and expansion. And if Decisioning Studio is going to be more than an interesting product line, it needs to become a margin story, not a services story.