Stay consistent on LinkedIn
without writing anything.

The e-commerce brands spending the most on paid ads are often getting the worst results. They run a campaign for ten days. The numbers aren't where they want them. They change the creative, tweak the audience, adjust the bid strategy — all in the same week. Then they wonder why nothing stabilises. The algorithm needs data to learn. Every time you make a significant change, you reset the learning phase. You're not optimising — you're interrupting. I've seen brands spend $40K a month and get worse results than brands spending $8K, purely because of how often they touched the campaigns. The discipline isn't in the setup. It's in leaving things alone long enough to see what's actually working. Most agencies don't tell their clients this clearly because it sounds like “do nothing.” But doing nothing, strategically, is sometimes the highest-value thing in the room. If your paid performance is all over the place — before you change anything, ask how often you've been changing things. The answer is usually the problem.

A client came to us in January with a €15K/month Meta budget and a cost per purchase sitting at €94. Their previous agency had been running 22 active ad sets simultaneously. Twenty-two. The logic, apparently, was to “test everything at once.” What it actually did was split the budget so thin that nothing had enough data to learn from, and the algorithm was optimising for the wrong signals because it had nothing else to go on. We cut it to four ad sets in week one. Same budget. Kept the two creatives that had the most spend history, wrote two new ones based on their top-performing organic content. By week six, cost per purchase was €51. Same product. Same market. Same budget. Just less noise, and enough patience to let the data actually tell us something. The problem with “testing everything” is that you learn nothing. Real testing means changing one variable, giving it time, and reading what happened. Everything else is just expensive guessing.

In 18 months, every agency that hasn't rebuilt its delivery around AI will be repricing or closing. I'm not saying this to be dramatic. I'm saying it because I've spent the last two years running an agency that does exactly this — and I can see the gap widening in real time between the agencies that have changed how they work and the ones that are still debating whether they need to. Here's what I actually observe: Agencies that haven't integrated AI into delivery are doing the same volume of work with the same number of people at roughly the same cost. Their margins are flat. Their capacity is fixed. When a client asks for a faster turnaround or a lower price, they have no room to move. Agencies that have rebuilt around AI are doing more work with fewer people, delivering faster, and still charging the same or more — because the output quality has gone up, not down. They're not using AI to replace thinking. They're using it to remove the parts of delivery that were never worth charging for in the first place. The mistake I see most often is treating AI as a tool you bolt onto an existing process. A copywriter who uses ChatGPT to write a first draft and then edits it is not an AI-enabled agency. That's just a slightly faster version of the old model. The real shift is structural — it's about redesigning the workflow from the output backwards, asking what actually needs a human and what doesn't. When you do that honestly, the answer is uncomfortable. A lot of what agencies charge for — research, first drafts, formatting, reporting — can be done faster, cheaper, and often better by a well-prompted model. The agencies that accept this and rebuild around it will have a structural cost advantage that compounds every six months as the models improve. The agencies that don't will find themselves in a market where their pricing no longer makes sense to clients who've started to understand what AI can do. I'm not predicting the death of agencies. I'm predicting a split — between the ones that used this moment to rebuild something genuinely more efficient, and the ones that waited to see how it played out. Waiting is a decision. It's just usually not recognised as one until it's too late.

Rebranding to attract better clients rarely works. I know because I've been involved in enough of them to see the pattern on the other side. The brand changes. The prices go up. The website looks significantly more expensive. And then — six months in — the founder is still having the same conversations with the same type of client, just with a nicer deck to show them. Here's what actually changes the client quality: changing who you talk to, before you change what you look like. If your pipeline is full of the wrong client, your positioning is probably talking to them. And a new logo doesn't fix that. A new message does. The most effective repositioning I've ever been part of didn't involve a rebrand at all. It involved a founder getting precise about who she wanted to work with, changing how she described her work, and updating one page on her existing website. She raised her day rate by 40% and closed two of the best clients she'd ever worked with — both of whom told her, unprompted, that the way she described her work was what made them reach out. The work was the same. The signal was different. Spend the money on clarity first. Then spend it on craft.

Fourteen months in and I still can't predict which features users will love. Last quarter we shipped something we'd been planning for three months. Two weeks of engineering, design reviews, testing. Launched it quietly to our active users. Seven of them used it in the first month. The week before that launch, I added a small filter function in an afternoon because a user asked for it in a support ticket. Didn't even announce it. Thirty-one users were using it within two weeks. Three of them mentioned it unprompted on calls as a reason they wouldn't switch to a competitor. I've stopped trying to predict what will land and started paying closer attention to what users do rather than what they say they want. What users ask for in feedback sessions is shaped by what they think is possible. What they actually do in the product is shaped by what they need. Those two things rarely match. The filter function they use constantly was never on any user's feature request list. It was just a problem someone had on a Tuesday that I happened to fix before the weekend.

The brief a client gives you is rarely the problem they actually have. Not because the client described it wrong. Because the problem they described is the symptom, and nobody asked what was underneath it. “We need to redesign the onboarding flow.” Fine. But why are users dropping off? Is it the flow, or is it that the product is asking for commitment before it's demonstrated value? Those are different problems. One is a UX problem. The other is a product positioning problem wearing a UX costume. We get briefed on the UX problem. We could deliver a beautiful new onboarding flow, invoice for it, and leave. The client would be happy for about six weeks. Or we could spend an hour before the brief asking different questions. What do users say when they churn? Where in the flow does drop-off actually happen — not where the client thinks it happens, but where the data says it happens? What's the one thing a retained user does in week one that a churned user doesn't? That hour usually changes the brief entirely. The agencies that get hired for the big strategic work aren't the ones with the best portfolios. They're the ones who made the client feel understood before the project even started. It's not a technique. It's just being curious before being useful. If you're about to invest in a product design project and want a partner who asks the uncomfortable questions before the work starts — book a 20-minute call. No pitch, no deck. Just an honest conversation about whether it makes sense. Link pinned in the first comment.

The decision that changed everything wasn't a product decision. For the first year and a half we were completely focused on the product. Every conversation internally was about features — what to build next, what to cut, what the roadmap looked like. We had a good product. We had users who liked it. We had an MRR number that was moving in the right direction. What we didn't have was a clear answer to the question: who is this actually for? We thought we knew. We'd written an ICP document. We had personas. We had a target market that we'd defined in a workshop and put in a slide deck. But when I look back at that period honestly, we were building for a category of person, not for a specific type of situation. The turning point came from a conversation with a churned user. I called her myself — something I hadn't been doing nearly enough — and asked her one question: what were you trying to do when you stopped using the product? She told me. It was specific. It was a situation we'd half-designed for but never quite solved. And when I went back and looked at the data, the users who had stayed the longest, expanded their seats, and referred other people to us — they were almost all in exactly the situation she described, dealing with exactly the problem she'd named. We had been building a product for everyone who might plausibly need it, when we should have been building a product for one specific person in one specific situation. We made one decision: we restructured onboarding entirely around that situation. Not a product change. Not a new feature. Just a different path through what already existed, designed for the person who was already getting the most value from it. Activation went from 31% to 68% in six weeks. I think about that churned user a lot. She didn't tell us what to build. She told us who we were already building for — we just hadn't been paying close enough attention to notice.

We quoted a client £85K. They hired someone internally for £55K. Eighteen months later they called us back. I wasn't smug about it. I understood the decision. £30K is real money, especially when you're looking at a spreadsheet and both options say "brand and creative." But they're not the same option. The internal hire was talented. She worked hard. She also spent four months learning the industry context we already had, another three months navigating internal politics we would have been insulated from, and the last five months trying to ship work that kept getting revised by stakeholders who'd never had a proper creative brief in their life. The rebrand launched. It looked fine. Fine is expensive when you're trying to grow. When they called back, they didn't negotiate the price. They asked when we could start. I tell this story not to win the argument I lost eighteen months ago — I lost that one. I tell it because the conversation about cost almost never includes the cost of time. And for founder-led businesses, time is the one thing you can't buy back. The question isn't what does the agency cost. It's what does it cost to not move fast.

How we structure every PR campaign so journalists actually respond: Step one: we don't pitch the company. We pitch the moment. Journalists don't write about companies. They write about things that are happening right now that their readers need to understand. The company is context. The moment is the story. Before we write a single pitch, we map out what's changing in the client's industry in the next 90 days. Funding rounds. Regulation. Competitor moves. Macro shifts. Then we position the client as the person who can explain what it means. Step two: we build the journalist list before we build the pitch. Most agencies write the pitch first and then find journalists to send it to. We do it the other way. We identify the twelve journalists who cover this intersection of topics, read their last thirty pieces, and write the pitch to the specific interests of each person. Twelve tailored pitches outperform two hundred generic ones every time. Step three: we give them something they can use without us. Data. A quote that stands alone. A counterintuitive angle they haven't covered. The easier we make their job, the more likely they are to respond — not because journalists are lazy, but because they're running on no time and a hundred pitches a day. Step four: we follow up once. One email. One week later. If they don't respond after that, the pitch wasn't right for them. We move on and rework the angle, not the journalist list. Coverage isn't about relationships. It's about relevance.

I turned down a $120K contract last month. Here's what I told them. The founder was sharp. The business was growing fast. The numbers were a mess but fixable. Everything looked right on paper. Then I asked who I'd be reporting to. He said: "Mostly me, but you'd need to run the big decisions past the board." I asked what "big decisions" meant. He said: "Anything that involves changing how we do things." I said I didn't think I could help them. A CFO who needs permission to change how things are done is a CFO in name only. You're paying for the title and the liability, not the function. And I've spent enough of my career in rooms where the right answer was obvious and the politics made it unsayable to sign up for that again voluntarily. He came back a week later and asked what it would take. I told him: direct access to the board, a seat in the room when decisions get made, and the authority to say no to things that are going to hurt the business — even when they're his ideas. He said that felt like a lot. I said that's what a CFO is. We didn't move forward. That's fine. The founders I work best with already understand that the finance function only works if it's allowed to be honest. If you're looking for someone to make the numbers look better, I'm not the right person. If you're looking for someone to tell you what the numbers actually mean — even when it's uncomfortable — that's a different conversation. PS — If this is the conversation you’ve been trying to have inside your business, I made a short video on what a CFO function actually looks like inside a founder-led company. Watch it here: [link]