The lists ask you to admire, not to check
Search for B2B content marketing examples and you get roundups. The first page we pulled on 8 September held 8 roundups and 1 forum thread, and 7 of the 8 put a count in the title. 10 inspiring, 20 inspirational, 7 exciting, 6 successful.
They share a habit. The artifact gets a screenshot, the result gets an adjective, and the source of the result is the company that ran it.
So we picked 4 examples and sorted the evidence ourselves. Each one is public, each one you can open in a minute, and for each we say what it proves and what it doesn’t.
What an example can actually prove
An example can carry three kinds of evidence, and roundups mix them freely.
The first is the artifact. The thing exists, you can read it, and its quality is yours to judge. It’s the weakest claim and the easiest to verify.
The second is the process. The producer published how it was made: what was counted, when, over what population, paid for by whom. You can audit the method without trusting the conclusion.
The third is the result. Somebody says it worked. It’s the claim buyers care about, and it’s almost never measured by anyone but the seller.
Sort any list this way and it gets short fast.
When the operating manual is the marketing
GitLab publishes its internal operating manual on the open web. Its own page calls it “the single source of truth for how we operate at GitLab,” and a second page adds that the company “makes its handbook publicly accessible to all.”
The interesting part for a marketer isn’t the text, it’s the changelog. The handbook site carries a page listing “The last 100 Merge Requests to the Handbook”, which means the edits are inspectable, not just the current text.
What this proves: the artifact, plus an unusual amount of the process. You can watch a document change.
What it doesn’t prove: any pipeline number. GitLab publishes no revenue attribution for it, and we can’t supply one either.
Giving away the whole book
Basecamp published Shape Up, its head of product strategy’s book on how the company builds software, as a complete website. 15 chapters, a foreword, appendices and a conclusion, all readable without a form.
There is a paid edition, and the site says so plainly next to the free one, in three words: “Buy the print edition”. The gate is the paper, not the argument. That single decision is the whole case study, and it is the opposite of the default in B2B, where the long document goes behind a form. We worked through when a form earns its keep in which content to gate.
What this proves: the artifact, and a costly commitment. 15 chapters is real work given away, and the free version is the complete one.
What it doesn’t prove: that it sold software. Nobody has published that number.
A report that shows its own working
GitHub publishes an annual report on activity across its own platform, and the useful part for our purposes isn’t any number in it. It’s that the report says what it counted, where the data came from, and over what window, so you know the figures describe accounts on 1 platform in 1 year and not the profession.
That’s the part to copy. A stated source and a stated window tell you what the answers can and can’t cover, and they let you argue with the method instead of guessing at it.
What this proves: the process, in full. You can criticise the sample without guessing at it.
What it doesn’t prove: that research pays for itself as marketing. It’s an annual commitment with a stated method, which is a different claim.
Watch what a forum does to a result
The three above are institutional. Here is one from a practitioner, in public, with numbers.
On a B2B marketing forum, in a thread asking for the scrappiest thing that ever worked, a marketer at a logistics software company described posting handwritten sticky notes to roughly 15 target accounts, on a printed one-pager, pointing at an article about the reader’s industry rather than the product. The reported outcome: 9 replies, 5 meetings, 2 customers. Stated cost, about $12 in stamps.
Now watch what the thread does with it. 3 separate commenters ask for the mechanism rather than the result, and all 3 ask the same thing: what did the note actually ask the reader to do?
And on a different story in the same thread, a reader says the quiet part: they doubt it ever happened.
What this proves: nothing, on its own. What it demonstrates is the audit an interested reader runs in about 90 seconds, unprompted, on a story that cost $12. Your buyer does the same to your case study.
Why none of these is a controlled test
None of the four is a controlled test, and that isn’t a flaw we fixed by choosing badly.
That matters because the counterfactual is missing every time. GitLab hired well before and after the handbook. Basecamp had an audience before the book. The mailer went to 15 accounts already chosen as targets, and 9 replies out of 15 hand-picked names is a different number from 9 out of 15 strangers.
We read an example as a demonstration that something is possible, and nothing more. It isn’t evidence that it’ll work for you, and a supplier who presents it as evidence is telling you what they want you to conclude.
Using an example while you’re choosing a supplier
Examples are still worth asking for. Ask for them differently.
Ask for 1 you can open. Not a screenshot, not a deck slide, a live link. Most B2B portfolios are shut, and judging closed work covers what a redacted sample can still show you.
Ask who measured the result. If the answer is the supplier, you have a claim, and claims price differently from measurements.
Ask what the piece cost in your time, not theirs. Interviews and approvals are the expensive part, and the arithmetic sits in white paper in-house or outsourced.
Ask what didn’t work. A supplier with 4 wins and 0 failures has either a short memory or a long story.
Turning the four questions on us
We publish portfolio pieces written for ourselves, because client work doesn’t go on public display without written agreement. That’s the artifact test, and it’s the only one of the 4 we pass outright.
We have no third-party measurement of results, and we don’t quote one. Our rates are published, so the cost side is checkable before you write to us.
And the failure question has a standing answer. Work goes wrong for us when nobody will give a writer access to the facts, which is why we turn that work down rather than write around it.
4 examples, 3 kinds of evidence, 1 uncomfortable conclusion. The example everybody wants is the one nobody has.
Sources
- The GitLab Handbook, About the Handbook and the importance of a handbook-first approach. Single source of truth, public accessibility, and the changelog page listing the last 100 merge requests.
- Basecamp, Shape Up. Fifteen chapters plus front matter and appendices, published in full; a print edition is sold alongside.
- GitHub, Octoverse. Used here only as an example of a report that states its source, its population and its window; no figure from it is quoted.
Send us the format and the subject, and say which of the 3 kinds of evidence the finished piece has to carry. A scope and a price follow inside one working day, plus a straight answer about the third kind.