Klaviyo AI Customer Research: How a $9B Marketing Platform Learns from 150K Brands

TL;DR#

Klaviyo (NYSE: KVYO), the publicly-traded marketing automation platform that IPO'd in September 2023 at roughly a $9B valuation, serves more than 167,000 ecommerce brands and powered $63B+ in attributed customer revenue across the 2024 holiday quarter. Klaviyo AI shipped at scale in 2024–2026 — Smart Send Time, Predictive Analytics, generative subject-line and copy features, Brand Voice tuning, and the Klaviyo AI Agents framework — and Klaviyo crossed $1B in trailing-twelve-month revenue in 2025 at +34% YoY. The interesting part for product and CX leaders is not the feature list. It is the B2B2C feedback architecture: Klaviyo researches two distinct customer layers at once — the 167K brands paying for the platform, and the hundreds of millions of end consumers whose purchase behavior trains every predictive model. That dual-audience problem is the structural challenge every B2B2C SaaS hits on the way to billion-dollar AI features. This case study covers Klaviyo's 2026 scale, the AI surface area, and what PM and research teams can copy from how Klaviyo runs ai customer interviews across both layers.

---

Klaviyo's 2026 Scale: 167K Brands, $63B in Attributed Revenue#

Klaviyo is the largest pure-play marketing automation platform built for ecommerce, and the 2026 baseline is worth grounding because the scale defines the research surface.

Three implications shape how Klaviyo has to do research:

• The brand base is heterogeneous in a way few B2B SaaS bases are. A solopreneur on the free tier and a public DTC brand on enterprise point at the same product surface. Long-tail signal can mislead enterprise roadmap calls, and vice versa.
• There are two customers, not one. The brand pays Klaviyo. The brand's end consumer — the shopper who opens the email, clicks the SMS, abandons the cart, or buys — is the one whose behavior trains every predictive model.
• AI features have to work for tail brands. Most B2B AI products quietly design for the top 10% of accounts. Klaviyo's models have to deliver value for a brand with 2,000 subscribers and one with 20 million. Data sparsity at the bottom is the harder engineering problem.

---

Klaviyo AI: What Shipped in 2024–2026#

Klaviyo AI is the umbrella for the AI features Klaviyo has rolled into the platform since 2023. As of 2026, four surfaces touch every brand customer.

Smart Send Time decides — at the individual subscriber level — the optimal moment to send a given email or SMS based on each subscriber's past engagement. The brand operator selects "Smart Send Time" instead of a fixed clock time, and the model handles per-recipient scheduling. Smart Send Time predates the generative-AI wave by years and is still Klaviyo's most-used AI feature.

Predictive Analytics ships customer-level predictions out of the box: predicted Customer Lifetime Value, churn risk score, predicted next order date, average order value, and average days between orders. Predictions surface in segment builders, so a brand can build an "at-risk high-CLV customers" audience in the same UI as a "purchased in last 30 days" audience. The integration matters more than the math — predictions are useful only when they're trivially actionable.

Klaviyo AI generative features (subject line and email/SMS copy generation, segment name suggestions, and Brand Voice — a tuning surface where the brand teaches Klaviyo's models its tone) shipped through 2024 and reached general availability in 2025. Brand Voice is the load-bearing one: it's the difference between generic AI copy and copy that sounds like the brand. Klaviyo charges nothing extra for generative features on most tiers, betting AI gets adopted only when it's invisible to the operator.

Klaviyo AI Agents is the 2025–2026 surface, the closest analog to what Sidekick is for Shopify merchants. The agents take natural-language instructions ("build a winback flow for customers who haven't purchased in 90 days") and produce drafts the operator can edit and ship. The deliberate scoping is the interesting design choice: Klaviyo's agents are narrow — they work inside Klaviyo flows, segments, and campaigns — and Klaviyo has explicitly avoided shipping a general-purpose marketing chatbot. That scoping came from brand research showing operators didn't trust broad agents with their sender reputation. Compare to the deflection-first chatbots common in insurance and support — Klaviyo deliberately did the opposite.

---

How Klaviyo Researches at the Brand-Customer Layer#

Klaviyo runs a research operation split along the brand customer base, and the split is the most copyable part of the playbook for any B2B2C SaaS.

The brand layer breaks into three cohorts: the "tail" (long-tail Shopify, BigCommerce, and Wix merchants under 5,000 contacts), the "growth" cohort (mid-market DTC brands at 50K–5M contacts with a marketing team of 2–10), and "enterprise" (public DTC brands, retail chains migrating from legacy ESPs). Klaviyo crossed 3,400 customers >$50K ARR by Q3 2025 — small in count, disproportionately large in revenue.

The research model maps to those cohorts:

• Tail brands get product-led research. In-product surveys, NPS prompts at meaningful moments (after first send, after first revenue attribution), churn-reason capture when a free-tier account goes dormant. Signal volume is enormous, per-account depth is shallow, and synthesis is the bottleneck. This is the cohort where AI-moderated interviews replace static surveys — you cannot run 1:1 calls with 150,000 brands, but you can ask follow-ups at scale.
• Growth brands get hybrid research. In-product behavior data, structured CS calls, and dedicated user research against strategic feature bets. Cadence is continuous and synthesis happens weekly — see the continuous discovery habits framework.
• Enterprise brands get high-touch advisory research. Customer advisory boards, named-account research programs, beta access in exchange for product co-design participation.

The mistake other B2B SaaS companies make at this scale is running one research process for all three cohorts. Klaviyo runs three.

---

The Freemium-to-Enterprise Feedback Architecture#

Klaviyo's pricing tiers go from a free plan (up to 250 contacts) to seven-figure enterprise contracts, and the feedback architecture spans the full range. It's built around four ideas.

Capture intent at signup, not after onboarding. Klaviyo's signup flow asks new accounts what they're trying to accomplish — first goal, current marketing stack, store platform, send volume target. Intent capture happens before the operator has invested time, and it lets Klaviyo segment the activation experience by use case. Same logic behind replacing contact forms with conversational intake: capture the why upfront, route based on context.

Instrument the activation moments, not the feature usage. The metric that matters is not "did the operator click the Predictive Analytics tab." It's "did the operator ship a flow that generated attributed revenue within 14 days." Klaviyo's analytics are wired around outcome moments — first send, first attributed dollar, first SMS, first flow live, first AI-generated subject line approved.

Route disengagement into qualitative research, not just retention dashboards. When a paying brand's monthly send volume drops by more than 50% versus their 90-day baseline, that's a churn precursor. The conventional response is to flag the account for CS outreach. Klaviyo's stronger move is to route a structured conversational follow-up asking what changed — and what they did instead. The "what they did instead" answer is the only one that maps to roadmap. Same pattern as the conversational signals that beat usage data alone.

Feed end-consumer signal back into product, not just into the brand's reports. The 600M+ shoppers who interact with Klaviyo-powered emails and SMS every month leave behavioral signal — opens, clicks, purchases, unsubscribes, replies. Most of it flows back to the brand as analytics. But Klaviyo also uses the aggregated signal to train the predictive models, and the model release cadence is published so brands can plan around it. Treating end-consumer behavior as a structured data layer is what separates a marketing platform from a marketing tool.

---

What B2B2C SaaS PMs Can Learn from Klaviyo#

The single biggest takeaway from Klaviyo's operation, for any B2B2C SaaS team, is that you have to instrument both customer layers and avoid letting one substitute for the other. The failure mode is assuming strong brand-customer satisfaction means the end-consumer experience is working — or, worse, that strong end-consumer behavioral metrics mean the brand operator is happy. Different populations, different success criteria, different methods.

Five concrete lessons:

1. Build segment-aware research processes. The tail, growth, and enterprise cohorts need different methods. One process for all three is the most common B2B research anti-pattern.
2. Capture intent at signup, not after activation. A two-question conversational intake at account creation is worth more than a 12-question survey six weeks in. Operator memory of "what I was trying to do" decays fast.
3. Treat end-consumer behavior as a structured product input. If end-consumer signal flows only back to your direct customer as analytics, you're leaving the most important feedback layer on the table.
4. Route disengagement into conversational follow-up, not just CS escalation. When usage drops, ask the operator what they did instead. The "instead" answer is the roadmap input.
5. Scope AI features narrowly and ship the unsexy ones first. Smart Send Time predates ChatGPT by years and is still Klaviyo's most-used AI feature. Narrow, embedded, invisible-to-the-operator AI compounds.

Teams running feature prioritization without the guesswork and jobs-to-be-done interviews at scale can adopt the same dual-layer architecture without a Klaviyo-sized budget — the constraint is methodology, not headcount.

---

How Other Public Platforms Compare on B2B2C Research#

Klaviyo's structure is worth seeing against the other public commerce and martech platforms running similar challenges in 2026.

Every public platform is moving toward conversational AI surfaces, and every one is reorganizing research to feed those surfaces. For Klaviyo, the dual-audience research problem is the strategic edge — see also the Shopify case study on multi-sided research, the DocuSign workflow research playbook, and how Anthropic researches enterprise AI buyers. For named-carrier insurance research on similar B2B2C dynamics — brokers as the brand layer, policyholders as the end consumer — the Lemonade conversational AI case study and Root Insurance's conversational risk interview are the cleanest parallels.

---

Frequently Asked Questions#

What is Klaviyo AI?#

Klaviyo AI is the umbrella brand for the machine-learning and generative-AI features built into Klaviyo's marketing automation platform, including Smart Send Time, Predictive Analytics (CLV, churn risk, predicted next order date), generative copy features for subject lines and email body, Brand Voice tuning, and the Klaviyo AI Agents framework that takes natural-language instructions for building flows and segments. Most AI features ship on all paid plans without an extra AI add-on SKU.

How does Klaviyo's Smart Send Time work?#

Klaviyo Smart Send Time is a model that decides — at the individual subscriber level — the optimal moment to send a given email or SMS based on each subscriber's past open and engagement behavior. Instead of the brand operator picking a fixed clock time, the operator selects "Smart Send Time" and the model handles per-recipient scheduling. Smart Send Time predates Klaviyo's generative-AI features by several years and is the default recommendation for most automated flows.

How does Klaviyo do customer research across 167,000 brands?#

Klaviyo runs a segment-aware research operation that splits the brand base into three cohorts — long-tail brands (under 5,000 contacts), mid-market growth brands (50K–5M contacts), and enterprise brands (3,400+ customers generating >$50K ARR). Tail brands get product-led research via in-product surveys and behavioral instrumentation, growth brands get hybrid research combining behavior data with structured customer success interviews, and enterprise brands get high-touch advisory research including customer advisory boards and beta co-design. The most important decision is running three processes, not one.

What is B2B2C feedback architecture and why does it matter?#

B2B2C feedback architecture is the practice of researching both the direct customer who pays for a platform (the brand) and the end consumer who uses the experiences the platform powers (the shopper), and treating both signals as structured product inputs. It matters because B2B2C platforms — Klaviyo, Shopify, marketing platforms generally — fail when they optimize only for the brand layer and ignore end-consumer behavior, or vice versa. A single research process for both produces decisions that work for neither.

How much revenue did Klaviyo customers attribute through the platform in 2024?#

Klaviyo brand customers attributed more than $63B in customer revenue to Klaviyo-powered campaigns and flows across the 2024 holiday quarter (Black Friday through Cyber Monday and the December holiday window), per Klaviyo's published BFCM recap. Klaviyo's own trailing-twelve-month revenue crossed $1.07B by Q3 2025, up 34% year over year.

How is Klaviyo different from generic email service providers?#

Klaviyo is purpose-built for ecommerce, deeply integrated with commerce platforms (especially Shopify, BigCommerce, and Wix), and ships customer-level predictive analytics and conversational AI agents as native features. Generic email service providers — the Mailchimps, Constant Contacts, and broader marketing-automation suites — are designed for any business sending email, not for the specific data model (orders, products, customers, sessions, abandoned carts) that ecommerce brands depend on. The integration depth and the ecommerce-native data model are what put Klaviyo on a different trajectory from horizontal ESPs.

---

What This Means for AI-First Research at B2B2C Platforms#

Klaviyo's 2026 position — 167,000+ brands, $63B+ in attributed customer revenue in a single holiday quarter, $1B+ in TTM revenue, and a four-surface AI product — is the result of taking ai customer interviews seriously across two distinct customer layers for over a decade. The brand layer pays the bill. The end-consumer layer trains the models. Both feed back into product decisions, and the company that treats either as secondary watches the gap to companies that don't widen every quarter.

For PM, research, and CX teams on smaller budgets, the playbook is portable. Perspective AI is the platform built for that architecture — AI-moderated customer interviews for the brand layer, and a single research surface for product, CX, and research teams to share. Built for product teams and built for CX teams, Perspective AI gives PM and research leaders the same dual-layer research capability Klaviyo built in-house — without the years of engineering investment.

Start a research study | See how Perspective AI works for B2B2C platforms | Compare AI customer research platforms