The interview is (nearly) everything

Last issue I argued that writing was never the bottleneck; that the weeks between a finished draft and a published page are where stories quietly die.

So let's go back to the beginning. To a part of the job that has not changed at all.

If a model will faithfully shape whatever material you hand it, then the quality of the source decides everything downstream.

And at the very top of your customer pyramid, this is not just about 'a source', it's a critical relationship.

01  Ten customers

Years ago, at a conference, I fell into conversation with someone from one of the biggest and most trusted names in enterprise IT. The 80/20 rule came up (80% of your business from 20% of your customers), as it does. She said that in their case it was closer to 90/10.

I said "wow! You must really look after that ten percent."

They looked at me over the top of their specs. "You've misunderstood me. Ninety percent of our revenue comes from ten customers."

Not ten percent. Ten. Individual. Customers.

Contracts of that size will do that to a business; and when a handful of them are governments, the upside and risk get larger still. An extreme example, yet the shape of it is not unusual at all. Most of us sit somewhere on that curve: a few relationships carrying an uncomfortable share of the revenue, and a long tail behind them.

So here's the thing:

Simple answer. Zero. None. Nada.

For customers that matter most, you don't automate it. Are you really going to send a questionnaire to the CIO of the account that pays for a fifth of your company? You send the best interviewer you have, prepared to the teeth, and you treat the forty-five minutes as what it is: a trust transaction with your most important relationship, which happens to also produce a story.

02  The message we send

A story about your fortieth-largest customer is a sales & marketing asset. A story about your largest is an event in a relationship that predates the campaign and will outlive it. The words may end up looking similar; what you spend getting them, and who you send, is a message the customer receives whether or not you meant to send one.

Nothing tells a senior executive what you think they are worth quite like the format of your request.

03  What a journalist does that a question list cannot

Interviews run from a program question list are fine. They produce accurate, complete, entirely predictable stories, because the method can only retrieve the answers to questions somebody thought of in advance.

But the story you came for might not be the best your customer has to tell.

Here are five things a good interviewer does which cannot be encoded:

They arrive knowing enough to be worth talking to. Read the quarterly results, the language, the last three announcements, the competitors. Executives grant access to folks who have done the homework and politely withhold it from everyone else. Forty-five minutes with a CXO is not just a diary slot; it is a decision they made about whether you were worth it.

They ask fewer questions and listen harder. Question quality beats question quantity; the interviewers who come back with the good stuff are usually the ones who got a third of the way down their sheet.

They follow the flicker. Around minute nineteen someone says something slightly off-script: the Friday afternoon it all went wrong, the decision they nearly did not make, the colleague who told them it would never work. That is the story. A list cannot notice it, and a model summarising the transcript afterwards cannot either.

They leave silences. Deliberately, past the point of comfort. Customers fill them with the thing they were not going to say.

They ask what nearly went wrong. Nobody volunteers it. Everybody has at least one.

And while the advocate is still glowing, this is where the yes is won; ask who else has to sign off there, in the room, rather than in week six.

None of this is nostalgia for a craft AI has made redundant. These instincts are worth even more than they were five years ago, because they are the only part of the process that has not been commoditised.

04  What a recording can and cannot give you

None of which means the only legitimate source is a bespoke interview. Quite the opposite.

There is a recording of the QBR. A webinar the customer spoke on, a renewal call, a research program full of verbatims, a support thread where somebody described the problem better than we could have. Feed that in and something comes out; and handled properly it can be genuinely good.

With brand context loaded, customer research sitting alongside it, and real editorial judgement on top, a QBR recording will yield a story worth publishing.

Product marketing want as many of those as they can eat, and they should have them.

The long tail of stories that never got told (speed, money, bandwidth) is a real loss, and closing that gap is one of the better things happening to this profession.

What a recording cannot do is give you what was never said in it. Thin material does not announce itself; hand a model very little and it will still return eleven hundred fluent, plausible words, and it will not warn you that the input was thin.

Judging source material is our job, and it's critical.

05  Why this matters more than it did

Trust used to be delegated. Buyers outsourced a large part of it to analysts, and vendors bought their way into that judgement.

That is collapsing. Analyst reports are now used by around one buyer in eight, down by nearly two thirds since 2022, while three quarters reach instead for peer evidence and reviews. And when they use AI to move faster, they do not believe that either; almost all of them check its answers.

Trust has not disappeared. It has moved to named, verifiable people describing what actually happened to them.

Which puts the weight on specifics: a real name, a real number, a before state a reader can picture. Those are not editorial decisions. They were either said out loud in a conversation, or they were not.

06  The other 90%

No one has a pot of gold large enough to place the best interviewer in front of four hundred customers. Or enough energy for the resulting internal politics. Which is precisely why most customer stories never get told at all.

Yet the long tail can have a different process: one that is genuinely good enough, and never pretends to be a hand-crafted interview. It can be remarkably good. It needs two things.

Decide early which process this story is getting; nothing costs more than discovering in week six that you ran the wrong one.

And make the volume process a good experience for the customer. The advocate who enjoyed the time you asked of them is the one who says yes next time, takes the reference call, and introduces you to a peer.

That is the "nearly" in the title.

This is why I built Proofpoints. Not to replace the interview, but to make the whole pyramid workable from whatever source you actually have; a forty-five minute conversation with a CIO, or a QBR recording nobody has listened to twice. The same standards applied to both, and the budget freed up to put your best interviewer where they genuinely matter.

07  What comes next

This is the fourth issue of Proof Positive. We have established that everyone is now an editor; that not every story deserves the same investment; that the time never went where we thought it did; and that the material we start from decides everything that follows it.

Next issue: big logo, small story. Everybody wants the famous name on the slide, and it is almost never the story that changes a buyer's mind.

Subscribe if you have not. Forward to one person who is about to email a questionnaire to somebody far too important to receive one.

§

Robin Hamilton is the founder of Proofpoints, the customer storytelling platform launching in 2026. Previously, cofounder and CEO of inEvidence. Proof Positive is published fortnightly.

Two weeks from now, in your inbox. Fortnightly, no spam, easy unsubscribe.