What a B2B Case Study Actually Proves: An Audit of Ten Examples.

I read ten published customer stories from one software vendor against the same checklist. Nine quote a named person. Four give a real before number. None gives a measurement window. Here is what a case study can and can't tell you.

What Happens Next

B2B & SAAS PLAYBOOKS Issue 51

The roadmap for this issue promised an analysis of thirty case studies. I read ten. They are all from one publisher, Zapier, whose customer-stories page is a public, named, consistently formatted set, which made it auditable. That makes this a narrow sample and a single-vendor one, so it can't tell you what converts. It can tell you what these documents contain, and, more usefully, what they leave out.

I applied the same eleven questions to each: is the customer named, is there a before number, how many distinct results, is a time period stated, is someone quoted with a title, who built it, and does anything go wrong. The tallies below are mine, from reading the pages.

The Scorecard

Ten Zapier customer stories, same checklist
Test	Result across ten
Customer named	10 of 10
Named person, with a title, quoted	9 of 10
A real before number	4 yes, 2 partial, 4 no
A full measurement window for results	0 of 10
Some period attached to at least one result	8 partial, 2 none
Distinct quantified results	1 to 4 per story
A real limitation or setback	1 of 10

Source: my reading of the ten pages listed below, November 2026. A "before number" counts only if the page states the starting value for the same metric. Results that restate one another (a percentage and the minutes behind it) count once.

The pattern is consistent. The stories are strong on identity: a real company, a real person, a real job title. They are weak on method. Not one says over what period the result was measured, and six of ten don't give you a before number that would let you check the percentage.

A case study proves a customer was willing to put their name on a number. It rarely proves how the number was made.

The Cracks Inside the Stories

Reading closely turned up small inconsistencies, which matter less for the individual stories than for what they say about how these are made.

The Miro story reports participation rising from 50%, while a quote in the same story says "from 43% to 93%." Otter.ai's Allen Lai appears with two different job titles. Vendasta's headline claims M in revenue "recovered," while the body also calls it potential revenue and uses the 282-day figure as both lost and saved time. ClickUp's 917 hours a month looks derived from 5,000 tickets times 11 minutes saved, a multiplication the page never shows. Vendavo's byline repeats a name.

None of this means the numbers are false. It means they were assembled for marketing, probably edited several times, and not reconciled. Treat any single figure as the customer's claim, and trust the story more where the arithmetic can be checked on the page.

The Strongest and the Weakest

The best of the ten on checkability is ClickUp's: a stated baseline of about 15 minutes of research per ticket across roughly 5,000 tickets a month, one named builder, a clear description of what he built, and a result you can reproduce roughly in a spreadsheet. ActiveCampaign gives a before number for churn, 25% in the first 90 days, and then an after. Brightspot states a baseline too: a single integration taking two to three months and tens of thousands of dollars.

The weakest tend to be the shortest. Smith.ai's story, about 200 words, has one clearly defined result, 250-plus hours saved weekly, and two headline stats that don't say what they measure. Vendavo's is a similar length with a 90% faster lead response and no starting value. Palo Alto Networks is the exception in a different way: its $150,000-a-year saving is a self-calculated estimate, but it is also the only one of the ten that admits the workflow outgrew the tool and moved to Azure as usage passed several thousand users.

Length doesn't predict quality, but it correlates with having room to show the baseline. The one-thousand-word ClickUp story shows its working. The 200-word ones can't.

What This Means If You're Writing One

My sample can't tell you which format closes deals. It can tell you what a skeptical buyer, who has read many of these, can and can't verify, and that's a design brief:

State the before number. Six of ten don't, and it's the single easiest way to look different. Give the window. "Per month since March" beats "up to 90%." Show the multiplication. If a headline figure is derived, let the reader see how. Name the builder and their role. The ClickUp and Miro stories do, and it makes the result feel like a person's work. Include one limit. The one story that admitted a tool's ceiling is the one that reads as most trustworthy. And keep the headline claim consistent with the body, because a 43% versus 50% slip is exactly what a careful reader remembers.

I can't tell you whether this converts better. That needs your own numbers: reply rate on stories shared in outbound, or deals where a story was cited. Which is this week's move.

Where These Numbers Come From

All ten pages are published by Zapier, which sells the automation software described, so they are marketing, not independent reporting. I read each as published; I did not contact any company. Counts and groupings are my own. Figures checked November 2026.

This Week's Move

Audit Your Own Best Case Study

Take your strongest customer story and answer the checklist above honestly: before number, after number, window, named person, who built it, one real limitation. Every "no" is a gap to fix before the next prospect reads it. Then track whether stories get opened in your next ten sales conversations, so you have a number of your own.

Coming Up — Issue 52

What WeWork's S-1 Told Investors, and What the Numbers Showed.