/* global window */
// Content object for the article renderer (blog/blog-article.jsx).
// Produced by the Writer→Editor→Humanize→Proofreader chain; draft in blog/pipeline/drafts/verra-onboarding.md
window.__ARTICLE__ = {
slug: "verra-onboarding",
laneLabel: "CASE STUDY",
kicker: "ONBOARDING",
readMins: 9,
dateLabel: "Apr 2026",
title: "An onboarding rewrite, traced to the clips that justified it",
deck: "How a B2B SaaS insights team rebuilt onboarding off research evidence — and tied every change back to a specific participant clip, so stakeholders could audit the recommendation instead of trusting it.",
tags: ["onboarding", "saas", "methodology"],
toc: [
{ id: "problem", num: "01 · THE PROBLEM", title: "The recommendation that dies in the room" },
{ id: "study", num: "02 · THE STUDY", title: "What the interviews actually said" },
{ id: "trail", num: "03 · THE TRAIL", title: "Building the evidence trail" },
{ id: "rewrite", num: "04 · THE REWRITE", title: "Every change had a parent" },
{ id: "room", num: "05 · THE ROOM", title: "Playing the tape" },
{ id: "travels", num: "06 · WHAT TRAVELS", title: "The reusable part" },
],
body: [
{ t: "p", html: `The onboarding rewrite shipped on a Friday. The deck that justified it was three slides long. What made it survive the room wasn't the slides. It was that every line in the recommendation had a clip number next to it, and anyone who doubted a line could click it and watch the participant say the thing themselves.` },
{ t: "p", html: `That's the whole story, really. A B2B SaaS team rebuilt the first-run experience of their product, and the thing that got the rewrite approved was not the quality of the redesign. It was that nobody had to take it on faith.` },
{ t: "p", html: `We see a particular failure often enough that it has a shape. A research team runs a good study, writes a sharp recommendation, walks into the stakeholder room, and watches the recommendation die anyway. Not because it was wrong — because the people who had to fund it could not see where it came from. "Customers want a risk-free way to try the product" is a sentence. It's also unfalsifiable in the moment. The VP who has to move the roadmap can't tell a real signal from a researcher's hunch dressed up in a chart, so they default to the safe thing, which is doing nothing.` },
{ t: "p", html: `This is a story about closing that gap. Not by arguing harder. By making the evidence trail walkable.` },
{ t: "h2", id: "problem", num: "01 · THE PROBLEM", text: "The recommendation that dies in the room" },
{ t: "p", html: `The team was a research-and-insights function inside a B2B SaaS company, the kind of group that buys and runs a lot of studies. In our recent work with research buyers, that profile is common: across a recent set of consumer-insights interviews, 90% of respondents either led research decisions or actively influenced them, and 80% had commissioned external research three or more times in the prior year. These are not casual buyers. They're professional skeptics, which is exactly why they get ignored when they can't show their work.` },
{ t: "p", html: `Their product had an onboarding problem. New accounts signed up, poked around, and a large share never reached the moment where the product actually did the thing it promised. That pattern is not unusual. Across more than 2,600 companies, Amplitude's 2025 benchmark data found that over 98% of new users churn within two weeks when they never hit a value milestone, and the average activation rate across SaaS and AI tools sits around 37.5% (AgileGrowthLabs, 2025; shno, 2026). The cost of a bad first run isn't subtle. It's most of your funnel.` },
{ t: "p", html: `The team knew the number was bad. What they did not have was an account of why, in a form a roadmap owner would act on. They had a hypothesis: the onboarding asked people to commit before it let them see value. But a hypothesis is not a mandate. They needed the why to be visible, and they needed it auditable, because the fix was going to cost engineering time the company did not want to spend.` },
{ t: "h2", id: "study", num: "02 · THE STUDY", text: "What the interviews actually said" },
{ t: "p", html: `So they ran AI-moderated interviews with people in their buyer's seat, then read the transcripts against the behavioral data they already had. The interviews were short, around two minutes each on average, which is roughly what these sessions run. The point was not session length. It was to get people talking, in their own words, about the moment they decide whether a new tool is worth the trouble.` },
{ t: "p", html: `The dominant finding was not price. That surprised people, because the team had assumed the friction was cost. It wasn't. The single largest barrier, present in about 65% of the conversations, was a demand to try before committing — a risk-free proof of concept before integrating anything new. Cost mattered too, but it showed up less than half as often. The thing standing between signup and value was trust-to-try, not willingness or budget.` },
{ t: "figure",
fig: { key: "transcript", props: { time: "P-03", speaker: "RESEARCH BUYER", children: `"Initially, we would like a demonstration to verify and see what type of data is collected and how it's collected."` } },
ref: "FIG 01",
caption: "Verbatim from the activation-barrier interviews — one of eleven clips where a buyer asked to verify before committing." },
{ t: "p", html: `Another said they would only move forward if "the company or an interview" offered a free trial, "in case my company didn't like it." A third, talking about how feedback shaped what they built, said input from potential customers "shaped the direction of my project to make sure I'm giving them great products at fair pricing." None of these people were asking for a discount. They were asking to be allowed to verify before they leaned in.` },
{ t: "p", html: `That maps onto what the activation literature has been saying for a while. Users who don't engage within the first three days carry roughly a 90% likelihood of churning, and products that get someone to the "aha" moment inside five minutes see about 40% higher 30-day retention than products that take fifteen minutes or more (AgileGrowthLabs, 2025; shno, 2026). The team's onboarding was doing the opposite. It asked for commitment up front and parked the proof somewhere after it.` },
{ t: "h2", id: "trail", num: "03 · THE TRAIL", text: "Building the evidence trail" },
{ t: "p", html: `Here's the part that mattered, and the part most teams skip.` },
{ t: "p", html: `The team did not write "customers want a risk-free trial" and call it a finding. They built a trail. Every recommendation in the rewrite pointed back to a specific artifact: the clip where a participant said the thing, the transcript line, the behavioral signal in the funnel that corroborated it, and the segment where the pattern held. Five layers, and a claim only counted if it had all five.` },
{ t: "pullquote", text: "A finding you can play back at the timestamp, read in the transcript, and watch repeat across a segment is evidence. A sentence in a deck is an opinion wearing a chart." },
{ t: "p", html: `This is the discipline. When the recommendation said "let people reach a working result before asking them to configure their account," it didn't float there. It carried a clip number, the count of participants who voiced it, the activation-funnel drop that confirmed it, and a confidence indicator that accounted for the people who said nothing either way. The VP did not have to trust the researcher. They could click the line and watch P-03 ask for a demonstration before committing. The argument stopped being "trust me" and became "watch this."` },
{ t: "figure",
fig: { key: "insight", props: { accent: true, title: "Trust-to-try beats price as the activation barrier", clips: 11, participants: 13, segments: 3, confidence: 0.65, body: `Adoption friction — the demand for a risk-free proof of concept before integrating — outranked cost roughly two to one, and held across segments. Traceable to specific demonstration-request clips, not a researcher's read.` } },
ref: "FIG 02",
caption: "The load-bearing insight, with its evidence layers attached: clips, participants, segments, and an honest confidence read." },
{ t: "p", html: `That's the difference between a recommendation that gets funded and one that gets tabled. Not rhetoric. Receipts.` },
{ t: "h2", id: "rewrite", num: "04 · THE REWRITE", text: "Every change had a parent" },
{ t: "p", html: `The redesign followed the evidence, line by line.` },
{ t: "p", html: `The old flow asked for account configuration, then team invites, then a guided setup, and somewhere past all of that, a result. The rewrite inverted it. Get the person to a real, working output first — on a sample or their own pasted-in data — before asking for anything that felt like commitment. The proof-of-concept moment moved to the front, because that's where the clips said it had to be.` },
{ t: "p", html: `Every change had a parent. Moving the working result before configuration traced to the demonstration-request clips. Cutting three fields from the first screen traced to a behavioral signal — the drop-off concentrated on that screen — plus the broader pattern that each extra onboarding minute costs conversion. Adding a "try on sample data" path traced directly to the participant who would not move without a way to verify "in case my company didn't like it." Nothing in the rewrite existed because a designer liked it. Each piece had a participant behind it.` },
{ t: "p", html: `This is also where honesty about confidence earns its keep. The strongest signal, the trust-to-try barrier at roughly 65%, was multiply corroborated: it showed up in the verbatims, in the theme prevalence, and in the funnel. The team built on it hard. Thinner signals, the ones resting on two or three participants, got flagged as hypotheses to validate after launch, not load-bearing decisions. A good evidence trail does not just show what you know. It's honest about what you're guessing.` },
{ t: "h2", id: "room", num: "05 · THE ROOM", text: "Playing the tape" },
{ t: "p", html: `The review went differently than these reviews usually go.` },
{ t: "p", html: `Normally the researcher presents, the room nods, and then someone with budget asks the question that kills it: "How do we know this is real and not just a few loud people?" That question has no good answer when all you have is a chart. It has a very good answer when every bar on the chart is a link.` },
{ t: "p", html: `Someone did ask a version of it. The team pulled up the clip. The room watched a real buyer, in their own words, describe needing to verify before committing. Then a second clip. Then the funnel data showing where people actually dropped. The question didn't get debated. It got answered, on screen, with the participant doing the talking. The recommendation moved from "the research team thinks" to "here's a customer saying it, and here are eleven more." You can't argue with the tape the way you can argue with a researcher.` },
{ t: "pullquote", text: "The fastest way to win a stakeholder room is to stop defending the finding and start playing the clip. Let the customer make your argument." },
{ t: "p", html: `The rewrite got approved. Not because it was persuasive. Because it was auditable. Anyone who wanted to check the work could check it, and that permission to verify — the same thing the customers had asked for — was what made the room say yes.` },
{ t: "h2", id: "travels", num: "06 · WHAT TRAVELS", text: "The reusable part" },
{ t: "p", html: `Strip the specifics and a method is left that travels to any team trying to turn research into a decision someone will actually fund.` },
{ t: "p", html: `Treat every recommendation as a claim that has to be sourced. If a line in your deck doesn't have an artifact behind it — a clip, a transcript moment, a behavioral signal, a segment — don't present it as a finding. Present it as a hunch, clearly labeled. Build the trail before you build the slides, because the trail is what survives contact with a skeptical VP.` },
{ t: "p", html: `Be honest about confidence. The 65% signal and the 20% signal are not the same animal, and a stakeholder can smell it when you treat them alike. Lead with what's multiply corroborated. Flag the thin stuff as something to validate, not something to bet the roadmap on.` },
{ t: "p", html: `And give the room permission to verify. The instinct under challenge is to defend the finding harder. The better move is the opposite: hand them the receipts and let them check. The teams that win the room are not the ones with the best rhetoric. They're the ones who can answer "how do we know" by clicking a line and letting the customer answer it for them.` },
{ t: "p", html: `The onboarding rewrite worked, in the end. But the reusable part is not the redesign. It's that a research recommendation got funded because every word of it could be traced back to a person saying the thing, on the record, at a timestamp. That's the whole job. Make the evidence walkable, and the decision makes itself.` },
{ t: "references", items: [
{ n: 1, html: `AgileGrowthLabs (2025). "User Activation Rate Benchmarks 2025." Average SaaS activation rate ~37.5%; AI/ML 54.8% vs FinTech 5.0%; a 25% activation improvement → 34% revenue; ~90% churn likelihood without engagement in the first three days. agilegrowthlabs.com` },
{ n: 2, html: `shno (2026). "SaaS Onboarding Statistics for 2026." Over 98% of new users churn within two weeks without a value milestone (Amplitude 2025 benchmark, 2,600+ companies); 5-minute aha-moment window ≈ 40% higher 30-day retention vs 15+ minutes; ~3% conversion drop per extra onboarding minute. shno.co` },
] },
],
related: [
{ href: "/blog/what-counts-evidence.html", title: "What counts as evidence", meta: "Index · Methodology" },
{ href: "/blog/synthetic-vs-human.html", title: "Synthetic vs. human research: when each one wins", meta: "7 min · Index" },
],
};