Voice of customer vs voice of self — the conflation killing your copy

1. The bug: VOC-only copy that sounds like nobody

You've seen this page. The headline names your problem with uncomfortable accuracy. The subheadline lists three pains you'd recognize anywhere. The bullet list is so calibrated it almost feels like the copywriter sat in on your last therapy session. And then — somewhere around the second scroll — you realize you have no idea who is selling this to you.

Not the company name. The voice. The person on the other side of the page.

This is the modern data-driven landing page. High-empathy diagnosis, zero authorship. It reads like it was assembled by a very smart system that listened to two hundred sales calls and produced the median sentence. Which, increasingly, is exactly what's happening.

The bug is not voice of customer research. VOC research is real, valuable, and probably under-practiced by most operators. The bug is that the copywriting community has spent fifteen years building extraordinary tooling for one half of the loop and almost nothing for the other. VOC tells you what your buyer is dealing with. It does not tell you who's supposed to be speaking.

When you only run one half, you get copy that sounds like nobody. And copy that sounds like nobody converts worse than copy that sounds like someone slightly off-thesis but recognizably human.

2. What VOC is, what it isn't, what it can't tell you

Voice of customer is a research discipline. The clean definition: the language a buyer uses, in their own mouth, to describe the problem your offer addresses. You capture it from customer interviews, won/lost sales call transcripts, review mining, support ticket archives, churn surveys, and increasingly from forum threads where your buyer talks unfiltered to peers.

The output is not copy. It's a corpus. Phrases like "I just need it to stop being a fire drill every Monday" or "I keep buying these and they sit in the drawer" — verbatim, in the buyer's own register. From that corpus, you derive the headlines, the objections to handle, the desires to lean into, the metaphors that already live in their head.

Practitioners like Joanna Wiebe at Copyhackers built much of the modern playbook around this. Conversion Copy Co and the broader conversion-copywriting community have refined it. Krystle Church and a generation of B2B SaaS copywriters made it standard practice. None of this is wrong. Most of it is excellent.

What VOC research will tell you:

• The exact words your buyer uses for their problem.
• Which fears block the purchase decision.
• Which outcomes they actually buy (often different from what they say they want).
• The language register they trust — clinical, casual, technical, plainspoken.
• Which objections are real versus performed.

What VOC research will not tell you:

• Who is speaking on the page.
• Whether your sentences should run long or stop short.
• What you believe that the rest of your category doesn't.
• The line nobody else would write because it isn't theirs to write.
• The signature you can sign at the bottom of the page.

That second list is voice of self. It's a different research problem with a different toolkit, and conflating it with the first one is the bug we keep shipping.

3. Voice of self — the operator behind the words

Voice of self is the language of the person doing the addressing. Tempo. Register. The words you reach for and the words you'd never use. The belief underneath the brand. The specific way you notice things that nobody else in your category notices the same way.

It is not "brand voice" in the agency-deck sense — three adjectives, a tone slider, and a list of words to avoid. That artifact has the shape of voice without the substance. You can tell because the deck could be applied to any of nine competitors and nobody would notice.

Voice of self is harder to capture because it lives one layer underneath the surface signature. It's not the costume. It's the person wearing the costume. And most voice exercises stop at the costume.

This is where the the Mirror framing pushes back on a lot of the voice-discovery industry. A typical brand-voice workshop produces adjectives. Bold. Warm. Authentic. Adjectives don't help anyone write the next sentence. What helps you write the next sentence is knowing your tempo, your default register, the line that only sounds right in your mouth, the contradiction you carry, the thing you believe that the rest of the category doesn't.

That's voice of self. It's structurally different from VOC because the buyer can never give it to you. They can tell you what they're wrestling with. They cannot tell you who you are.

4. Why the conflation happens

The data-driven copywriting movement is real and largely correct. Before it, most landing pages were written by guessing what the buyer cared about. Wiebe and others made it standard to check — to mine the actual language of the actual buyer, then write copy that sounds like the buyer thinks. This was a step-change improvement. We're not arguing with it.

The conflation creeps in because the movement was so successful that it became the methodology, and the second half of the loop quietly fell off the radar. If you can sit through ten interviews and produce a landing page that sounds like the customer, why bother with the slower, weirder, harder-to-defend work of figuring out who you are on the page?

Two reasons.

One: the customer voice is shared. If your competitors run the same VOC research on the same customer base, you'll arrive at uncannily similar headlines. We've seen entire categories where the top five pages all open with versions of the same sentence — because they all listened to the same buyer. The customer voice is a commodity input. It cannot be your differentiator.

Two: people don't only buy diagnoses. Especially in creator-economy, coaching, consulting, and founder-led B2B, people buy operators. They buy the person they think is on the other side of the offer. A page that diagnoses them perfectly but reads as if no one is selling it leaves the most expensive trust signal — recognizable authorship — on the table.

The fix isn't picking one. It's running both.

5. The clean separation

Here's the cleanest way to hold the distinction in your head:

Voice of customer tells you the JOB. Voice of self tells you WHO'S DOING IT.

VOC defines the page's brief: what the buyer cares about, what objections to handle, what register to write in. Voice of self defines the page's author: tempo, signature lines, which sentences could only have come from this operator and not the next one over.

A landing page is a piece of speech. VOC tells you what to say. Voice of self tells you how it should sound when you say it. They're not redundant. They're the two inputs.

This is also why "AI brand voice" tools that ask you to upload writing samples don't work for early-stage operators. They can do voice-of-self extraction if you've already produced enough material to extract from. Most operators shaping a new offer haven't, which leaves them with no voice-of-self research at all and they default back to pure VOC. Hence the no-one-on-the-page problem.

6. A worked example

Let's run the same thesis through a single page, two ways.

The brief: sell a $497 productized service to founders who keep launching newsletters that nobody reads.

VOC-only headline (well-researched, signs nobody):

Stop launching newsletters into a void. Get the audience research, naming framework, and first three issues your launch needs.

It's fine. It diagnoses the problem in language the buyer would recognize. It's also indistinguishable from any other page selling newsletter-launch help. You could swap the operator's name and nothing about the language would feel wrong.

VOC + voice of self (same brief, signed by a recognizable operator):

Most newsletter launches don't fail because the writing is bad. They fail because the writer hasn't decided who's speaking yet. We do the deciding part first — then the launch is just announcing what's already true.

The job (newsletter launches that fail) is still named. But now there's a thesis underneath — the writer hasn't decided who's speaking yet — that only this operator would write. The verb deciding is doing real work. The cadence is theirs. A reader who's seen this operator twice before will recognize the page before they read the byline.

The second page does not abandon VOC. It still opens on the buyer's pain. It just refuses to leave the operator off the page.

7. The two-input flow — where to actually get each

These are different research problems with different tooling.

For voice of customer: the established stack. Customer interviews (5-12 is usually enough for a landing page), won/lost sales call review, review mining on G2 / Capterra / Amazon depending on category, churn surveys, support ticket archives, and increasingly forum scraping for unfiltered peer-to-peer language. Build a verbatim corpus. Pull headlines, fears, desires, and objections directly from it. The Copyhackers / conversion-copywriting community has documented this thoroughly — the methodology is mature.

For voice of self: much less mature. Most operators have either nothing (a vague sense of their tone) or an over-engineered brand-voice deck full of adjectives. Neither helps you write the next sentence. The cheap, repeatable version is a structured excavation — a guided question flow that surfaces tempo, register, signature lines, beliefs, contradictions, and the words you'd never use. the Mirror productizes one version of this; bespoke brand-voice consulting is the expensive version. Either way, the deliverable should be something concrete enough that you could hand it to another writer and they could sound like you on first try.

You run both inputs into the same brief. VOC fills the what to address slots. Voice of self fills the who's addressing it slots. The page that comes out of both is harder to write than the page that comes out of one — and far harder to confuse with anything your competitors are shipping.

8. When to lead with which

Both inputs run on every page. But the lead — the headline, the first paragraph, the thing the buyer reads before deciding whether to keep scrolling — depends on the audience.

Lead with VOC when:

• The buyer needs to feel diagnosed before they trust you. Most B2B SaaS, regulated industries, and category-defining offers fall here.
• The category is mature and the buyer has language for the problem already.
• The operator is not the product. (Enterprise software. Most agencies. Consumer goods.)

Lead with voice of self when:

• The buyer is choosing the operator as much as the offer. Coaches, consultants, course creators, founder-led consulting, premium creator brands.
• The category is so saturated with VOC-mirrored copy that diagnosis no longer differentiates. (See: every productivity newsletter on Earth.)
• The offer's value is partially or wholly the operator's judgment, taste, or perspective.

In practice, even VOC-led pages need voice-of-self running underneath the surface — the body copy, the testimonial framing, the close. And voice-of-self-led pages still need VOC informing what problems to name. The lead question is which input gets the headline. The full-page question is both, always.

9. The cheap version of voice-of-self research

VOC research has a mature, well-documented playbook. Voice-of-self research mostly doesn't, which is part of why the conflation persists — there's no equivalent toolkit for half the loop, so people skip it.

We built the Mirror for this exact gap. It's a $27 productized session that runs nine adaptive questions and outputs a voice brief — tempo, register, signature lines, beliefs, the words yours / never yours, and a foundation file you can paste into any AI tool when you're drafting copy. It's not a replacement for VOC research. It's the missing input that sits next to it. We've made the case elsewhere on why most brand voice tools miss the layer underneath — same argument, applied to the broader voice-discovery industry.

If you have $5k+ and a long timeline, hiring a brand-voice consultant works too. If you don't, this is the cheap version. Either way, the goal is to leave the session with a deliverable concrete enough that a reader of your page would recognize the operator behind it. That's the bar.

VOC tells you the job. Voice of self tells you who's doing it. Run both. The pages that come out of both are the ones that don't sound like nobody.