Why Founders Get Positioning Wrong, and How to Fix It

Craig Brown spent three years at HSBC before concluding he was in the wrong place entirely, and went looking for something that was the opposite of a 263,000-person organisation. He landed at an early-stage startup as roughly its twelfth employee, built the customer success function, then moved into product marketing at a point when only a few hundred people in the UK were doing the job at all. In 2021 he founded Troubadour, which he describes as the UK’s first dedicated product marketing consultancy for B2B tech startups. In this episode he talks to Rob Woodhead about the single most common diagnostic error founders make, why a broad ICP quietly wrecks your customer research as well as your budget, why ambitious positioning often removes urgency rather than creating it, and where he does and does not trust an LLM with market research.

Key takeaways

• Most startups diagnose a messaging problem when the underlying issue is positioning, and fixing the words will not fix the target.
• An ICP is not just a company profile. It has three parts: the organisation, the person inside it, and the specific use case, and vagueness in the use case is what usually makes the whole thing broad.
• A broad ICP damages research as much as budget, because twenty interviews with twenty different customer types produce no trends to act on.
• Most startups hold more usable signal than they realise in existing product data: who adopts fastest, who upgrades, who does not churn.
• Highly ambitious positioning tends to reduce urgency, because a buyer cannot see how the product helps them this week.
• Companies now known for expansive positioning, including Stripe, began by solving one narrow problem for one specific role.
• In customer interviews, emotional and social reactions signal the problems buyers will actually pay to solve.
• LLMs are unreliable when asked to gather competitive intelligence and considerably more reliable when asked to analyse data you supplied.
• Surrounding yourself with people slightly further ahead changes what you believe is possible, which is a practical business input rather than a soft one.

Episode chapters

• 01:28  Three years at HSBC and the decision to leave
• 03:39  Falling into product marketing when almost nobody in the UK did it
• 05:18  Why product marketing suited him, and the role of empathy
• 08:26  Going independent and launching Troubadour
• 12:59  Messaging problem or positioning problem?
• 15:22  What actually makes a good ICP
• 17:42  Finding signals in the data you already have
• 21:13  How the work changes from pre-seed to growth stage
• 26:03  What founders should put in place early
• 28:43  Why the customer interview is still king
• 30:45  The ambition trap, and why big claims kill urgency
• 34:42  How Stripe and Facebook actually started
• 37:31  The ChatGPT positioning question neither of us can answer
• 40:00  Where AI helps in product marketing, and where it hallucinates
• 44:33  Community, and surrounding yourself with people further ahead
• 51:48  Comfort zones, and deciding to make a bold claim
• 56:31  What is next for Troubadour

Is it a messaging problem or a positioning problem?

Almost always positioning, in Craig’s experience, and the confusion is understandable because the symptoms all look like messaging.

The website is not clear. Buyers take a long time to work out what the product does. Campaign click-through rates are low. And the handful of prospects who do request a demo arrive still not really understanding the proposition, so a chunk of the call gets spent on explanation that should have happened earlier.

Every one of those points at the words, which is why founders reach for more precise messaging. But the questions underneath are positioning questions. Are you clear on who you are targeting? Do you understand the specific pains worth solving? Do you understand the buyer’s job at a technical level, well enough to reach a problem commercial and urgent enough that someone will pay a subscription to make it go away?

So the conversation Craig frequently has runs: yes, you have a messaging problem, and before we touch the messaging we are going to talk about positioning. That is where he sees the largest gaps.

What makes a good ICP?

Specificity in three places rather than one.

The answer he usually gets is a company profile, something like SaaS companies between 50 and 300 people. That is one component of three. The others are the person or people inside that company, and, critically, the use case. Use-case vagueness is usually what makes an ICP broad even when the company definition looks tight.

What he wants to understand is the end user, the problem they face, how they would actually use the product, and what an ideal outcome looks like for them. That combination is what constitutes an ideal customer.

He also checks something before any of that: whether the team even agrees. The ICP is often undocumented, and different people hold different versions of it. Founders frequently say the ICP is not the problem, and it frequently is.

The cost of getting this wrong is doubled. Scattered go-to-market resources across too many segments is the obvious half. The less obvious half is that research becomes impossible. Twenty customer interviews across twenty different customer types produce no comparable trends, so you cannot tell who feels the pain most acutely or who is most willing to pay.

This is the same problem we work through with clients before a build starts. You can see how we approach it and the products that came out of it.

How does positioning work change as a startup grows?

The goal reverses. Early on you are widening deliberately; later you are narrowing from evidence.

At pre-seed, narrowing is not the objective. Casting the net wide, generating signals and validating ideas is. Craig lists concrete tests: A/B testing two messages in cold outbound, or the same message to two buyer types. Pitching at different networking events to see which audience reacts with more genuine enthusiasm, and which needs more explanation before they understand. Publishing different kinds of content to see which buyer engages.

At growth stage the work shifts to interrogating data you already hold, and this is where startups diverge sharply. Some have built this analysis into their processes and arrive at a workshop ready to distinguish bad-fit from good-fit from best-fit customers. Others send a CRM export where, out of five thousand accounts, only fifteen to twenty per cent have complete fields, and the first task is enrichment rather than analysis.

Craig notes this is getting more common rather than less, because teams are leaner. A Series B company may still have only one or two full-time marketers, which means less capacity for the maintenance work that makes this analysis possible later.

Why do customer interviews still matter?

Because the hard data is the easy part to obtain, and the qualitative data is the part that is genuinely scarce.

Product analytics will tell you who is active, who upgrades, and who spends most. What it will not tell you is why. That requires talking to people, and Craig thinks the one-to-one interview will resist replacement for a while yet, including by AI, because of how much nuance sits in a single conversation and how much depends on hearing something and deciding to follow it.

His specific advice on what to listen for is the most useful part. Watch for emotional and social dynamics. When a customer becomes animated about something, or when the problem touches how they are perceived by colleagues, that is usually a sign they care enough to invest in solving it. Those reactions predict willingness to pay better than a stated priority does.

He also argues for doing it continuously rather than as a project. Markets move, new solutions appear, regulation changes, macroeconomic conditions shift and personal values change, and all of that alters who gets value from a product and how they perceive it.

Why does ambitious positioning backfire?

Because scale and urgency pull against each other, and urgency is what closes.

Craig describes real pressure on founders, sometimes from investors and sometimes internal, to position around a large mission. His example is HR tech companies claiming to solve employee culture or reduce churn, when in reality they are one component among many contributing to employee wellbeing.

The trap is that these enormous goals feel less tangible, not more compelling. A buyer hears an ambition they broadly share and cannot see how the product helps them this week, so the thing becomes a nice-to-have and the conversation gets postponed until they have time to understand it properly. Which is to say, never.

Pitching at the level of the job does the opposite. This specific task, done this way, without the downsides of how you do it now. Craig’s argument is that this is not less ambitious, it is ambition that is legible.

The examples make the point efficiently. Stripe today positions as financial infrastructure for large organisations. It began by helping software engineers at e-commerce companies implement online payments, which had been a convoluted engineering problem. One use case, one role, one type of company. Facebook launched on a handful of US college campuses. Founders tend to want the positioning those companies have now, which is not the positioning either of them started with.

Can AI do competitor and market research?

Not reliably when it has to find the information. Reasonably well when you supply it.

Craig has tried using an LLM for competitive intelligence and describes the hallucinations arriving quickly: feature sets attributed to competitors that do not have them, capabilities omitted, a characterisation of a competitor’s ICP that was simply wrong. The correction overhead was large enough to make the exercise pointless.

The inversion works much better. He gathers reviews himself from sites like G2 and Capterra, five-star and one-star, feeds them in, and asks for qualitative and quantitative analysis: which job titles recur most, whether one use case dominates. He finds the accuracy high enough to trust, because he can sanity-check the output against material he has read himself.

So his working rule is that an LLM should save time analysing data you already trust, including interview transcripts and jobs-to-be-done research, rather than being asked to go out into the market and come back with facts.

Rob notes the same pattern in engineering: people who know what good looks like get faster, and people still learning cannot evaluate what they are given. Craig agrees that this is the constraint on getting value from these tools at all.

The equivalent problem in code is what our Vibe Code Audit addresses, and The Jekyll and Hyde of Vibe Coding sets out both failure modes. Evan Larbi makes a related argument about AI output in brand and narrative work.

Why does community matter for founders?

Craig cites Harvard Business Review research on the traits of successful entrepreneurs, and the finding that stayed with him is that successful founders surround themselves with people who are where they want to be, or on their way there.

His account of the effect is concrete rather than motivational. Since joining several communities he takes himself more seriously, has expanded what he considers possible, and is planning in-person events with partners, something he would not have contemplated a couple of years ago. He attributes that shift to watching people in his network do it first.

There is a loneliness argument too. Founders, particularly solo ones, are doing something nobody in their immediate circle does, which makes it hard to discuss. Sometimes the value is practical, someone saying they hit the same ceiling six months ago and describing how they got past it. Sometimes it is just being understood.

He is also direct that founders have to be somewhat delusional, because there will be many voices suggesting a thing is difficult or impossible, and listening to all of them means stopping. He thinks British culture has been historically poor at celebrating that kind of ambition compared with the Bay Area, though he sees it changing.

His own example is good. Researching the market before launching, he could not find another dedicated product marketing consultancy in the UK, and hesitated to say so because it felt arrogant. He decided to make the claim anyway. His framing is that the fears surfacing at the edge of your comfort zone are the signal you have reached it, rather than a reason to stop.

About the guest

Craig Brown is the founder of Troubadour, which he describes as the UK’s first dedicated product marketing consultancy for B2B tech startups. He works on positioning, messaging, ICP research, sales enablement and go-to-market strategy with companies backed by investors including Y Combinator, Octopus Ventures and LocalGlobe. He previously worked at HSBC Global Banking and Markets before moving into tech, where he built both customer success and product marketing functions at an early-stage startup. He is also a stand-up comic. You can find him on LinkedIn.

About the host

Rob Woodhead is a co-founder of Old.St Labs, a London-based software agency building web and mobile products for startups and innovators. He also runs Tech Startups in the Pub, a monthly London gathering of founders, which is where he and Craig know each other from.

Frequently asked questions

What is the difference between positioning and messaging?

Positioning decides who you serve, which problems you solve and why you are the right choice. Messaging is how you express that. Craig Brown’s experience is that most perceived messaging problems, including unclear websites and low engagement, originate in unresolved positioning.

What should an ICP include?

Three components: the type of organisation, the specific person inside it, and the use case. A definition covering only company size and sector is incomplete, and an unclear use case is usually what makes an ICP too broad to act on.

How do you narrow your ICP?

Start with existing product data. Look at who adopts fastest, which use cases drive the highest adoption, who does not churn, and who upgrades from free to paid. Those signals reveal which customers perceive the most value, which gives you evidence for narrowing.

Why is ambitious positioning a problem?

Because very large claims feel intangible to a buyer, who cannot see how the product helps them today. That removes urgency and turns the product into a nice-to-have. Positioning around a specific job creates urgency while remaining ambitious.

Can AI do competitor research?

Poorly, if it has to source the information, as hallucinated features and inaccurate competitor profiles are common. It performs much better analysing material you have gathered yourself, such as customer reviews or interview transcripts.

‍