The best voice-of-customer research I have run, I ran without speaking to a single customer. I had to sharpen the messaging for a product I could not put in front of its own switchers, so I pulled the switcher's voice from the only people who heard it every day: the colleagues sitting one seat closer to the customer than I was.
This is a common trap for anyone who writes to convert. You know that good copy comes from the customer's own language, and the standard advice is to go interview them.
Then reality arrives. You do not have access to the accounts.
You do not have the weeks a proper interview study takes. MIT's John Hauser, who wrote the standard reference on this with Abbie Griffin, found that 20 customers identify over 90% of the needs 30 would surface, which is still 20 people to recruit and schedule.
Or the organization treats its customer list as something marketing is kept at arm's length from, for reasons that are entirely sensible from where legal and support are sitting. The insight you need is real, and the front door to it is bolted.
I hit exactly this working as a consultant to the web team inside one of the largest players in marketing software. My job was to make the pages that greet a prospect actually speak to why they would leave a competitor and come over.
To do that well I needed the voice of a switcher: the specific frustration that makes someone start shopping in the first place. The switchers themselves were behind the access wall. The colleagues who fielded that frustration all week were not.
The customers you cannot reach
Start by being honest about the shape of the problem. A prospect deciding whether to switch is at the center of everything, and almost nobody on the marketing team ever hears them think out loud.
What you get instead is aggregate data: which page they bounced from, which plan they picked, how long they lingered on pricing. Useful, and completely silent on the one thing copy needs, which is the sentence the prospect would say if you asked them what was wrong with what they had.
Now look one ring out from that customer. Support answers the ticket the moment something breaks. Sales sits across from the objection every time a deal wobbles. Product fields the feature request that is really a complaint in disguise.
These people are soaking in customer language all day, and none of them is ever asked to write it down in a form the marketing team can use.
Your colleagues are a customer-research panel nobody thinks to survey. Sales and support field more real objections in a week than an interview study collects in a quarter.
Ask the people who hear the objections
The way you ask matters as much as who you ask. My instinct might have been to book a meeting, get everyone in a room, and whiteboard it.
That would have been a mistake. Hauser's comparison of the two formats found the expected group synergy absent, with two one-on-one interviews surfacing about as many needs as one focus group, 51% against 50%.
People recall objections poorly on the spot with an audience watching; the good ones surface when someone has a minute to think back over the last dozen conversations they had. So I ran it async.
I sent a short, specific prompt to each function, phrased for their vantage point.
To the sales side: what is the frustration a prospect names most often when they explain why they are looking to move. To support: what is the recurring failure that makes someone start questioning the tool they already pay for. To product: which requests are really a workaround for something the competitor does badly.
One question, tuned per audience, answerable in five minutes from a phone. Within a couple of days the thread had filled with the kind of detail no survey would have surfaced.
Pin every insight to the person who said it. Attribution is what turns a synthesis into evidence instead of one more confident opinion.
Synthesize into a taxonomy of pain
A thread full of raw answers is not yet an output. The work now is to read across every response and find the pattern: the handful of distinct pains that keep recurring under different wording.
One person's story about a botched migration and another's about a stalled setup are the same underlying pain wearing two costumes, and the synthesis job is to name that pain once and file both stories under it.
As I clustered, I carried each contributor's name along with their input, so a pain that surfaced from three functions arrived with three sources stapled to it rather than my summary alone.
What came out the other end was a short, named list. Each entry held the pain in the prospect's own language, the role that heard it most, and the promise that would answer it.
Six or seven of these, no more, sized so the whole team could hold the set in their heads and point to it in any messaging debate. The formal version of this lands in the same range, with Griffin and Hauser's structure keeping two to ten top-level needs at the strategic tier and the detail nested underneath.
That list became the reference everyone reached for. When a designer questioned a headline or a paid-media colleague pitched a new angle, the argument stopped being about taste and started being about which entry in the taxonomy the work was serving.
The real output of research is a shared taxonomy of pain. The campaigns you build on it expire in a quarter; the taxonomy keeps paying out for years.
From taxonomy to messaging and tests
With the taxonomy in hand, the hard part of the messaging is already done. Each pain implies a promise, and each promise, phrased in the reported language rather than in house marketing-speak, becomes a candidate headline.
Those candidates turned into a set of landing-page test concepts, each one explicitly mapped to the pain it was answering. That mapping is what makes a losing test valuable: when a concept fails, you learn something precise about a real objection rather than just retiring a headline you happened to dislike.
Those concepts are creative, and creative earns its place by getting scored against results rather than argued over in a meeting; the way I run that scoring is in the creative scoring system.
Upstream, this same research fed the quarterly plan I describe in the learning agenda, where the pains became ranked questions worth testing.
And it was only possible because the colleagues I leaned on already trusted me enough to answer, which is the relationship groundwork I cover in breaking silos as a growth lever. Ask thirty strangers for their hardest-won knowledge over Slack and you get silence.
Running it without customer access
None of this requires a research budget or a single scheduled interview. When the customer is out of reach, borrow the ears of the people who reach them for you. The method:
- Map who in your company actually talks to customers, and notice that it is rarely marketing. Usually it is support, sales, and product.
- Send each function one sharp, async question aimed at their vantage point, answerable from a phone in a few minutes.
- Collect the answers with the source attached, so every claim keeps a name and a role behind it.
- Cluster the responses into six or seven named pains, each carrying the prospect's language and the promise that resolves it.
- Turn each pain into a message and a test, mapped back to the pain, so every result teaches you something about a real objection.
The reason this works is almost unfair. The insight most teams pay an agency to go chase is already sitting inside the building, distributed across the people who field it and never thought anyone in marketing wanted to hear it.
The customer stays out of reach and you hear them anyway, by asking the person one seat closer and then doing the patient work of turning their scattered answers into something the whole team can build on.
If reaching your customers is the problem, the people who already reach them are the answer, and turning that into copy is the kind of work I do. Let's talk.