About to drop £10k on marketing? Answer these first.

£10,000 feels like progress. It can also be an expensive way to confirm what you already suspected: the offer is not clear enough to convert. Before you spend, answer these five questions honestly. Honesty is cheaper than CPC.

I am not anti-marketing. I am anti-spending money to discover, publicly and at scale, that your homepage still cannot finish a sentence. Paid acquisition is a amplifier. Amplifiers do not invent clarity, product value, or follow-up systems. They reveal whatever you already have-sometimes with unfortunate speed.

Early-stage teams often treat a marketing budget like a rite of passage: we raised, therefore we advertise. That sequence skips the part where someone proves a cold visitor can understand the offer and take a meaningful next step. If you want a deeper take on why growth stalls look like marketing issues, read your product doesn’t have a marketing problem-it has a clarity problem.

Why these five questions before any media plan

Media plans are comforting. They have charts, audiences, and weekly pacing. They also assume the destination page and the offer are ready. Most early startups fail that assumption quietly. Traffic arrives. Bounce rates look “industry average.” Nobody books a call. The post-mortem blames creative fatigue or attribution. The quieter truth: the offer was never decision-ready.

Good spend answers a question. Bad spend produces dashboards. The five questions below force the former. They are deliberately uncomfortable. If you cannot answer them in a room without opening a slide deck full of aspirational personas, pause the budget.

1. Who is the one customer this budget must reach?

If the answer is “anyone who might need us,” pause. Broad audiences inflate cost-per-click and dilute creative. Narrow your ideal customer profile first. Specificity is not exclusion for its own sake-it is how you stop paying to confuse strangers.

Write the customer as a decision-maker with a context, not a demographic. “Marketing managers at Series A B2B SaaS companies who are drowning in tools and need pipeline they can explain to a CFO” is usable. “Growth-minded professionals” is a billboard looking for a product.

Stripe did not start by advertising to “everyone who accepts payments.” Early focus was developers who hated payment integration pain. Linear spoke to teams who felt issue trackers were slow and noisy. That sharpness made creative and landing pages easier-because the audience was not a committee of maybes.

• Name the role, company stage/size, and the trigger that makes them search now.
• Write an explicit exclusion list (who you will not target with this £10k).
• Confirm you can reach them in a channel you actually understand-not a channel that merely looks modern.

2. What is the one outcome we promise in one sentence?

If the landing page needs a paragraph to explain the product, ads will not save it. Clarity is a prerequisite for paid efficiency. Write the sentence as if a tired buyer has ten seconds and no patience for your category vocabulary.

Weak promise: “The all-in-one platform for modern revenue teams.” Stronger: “Turn messy CRM notes into a weekly forecast your leadership trusts.” The second can live in an ad. The first requires a webinar.

Match the ad promise to the first viewport. If the ad says “cut onboarding time in half” and the page opens with a brand manifesto, you paid for a bait-and-switch. Nielsen Norman Group has spent decades showing that people scan for relevance and next steps-not for your mission statement. Honour that behaviour.

3. What proof makes that promise believable?

Metrics, case snippets, demos, founder credibility-pick something real. Unproven claims raise customer acquisition cost because people correctly assume risk when you ask them to take it on faith.

Proof does not have to be a Fortune 500 logo. Early-stage proof can be: a quantified pilot, a short clip of the product delivering the outcome, a named customer quote with a specific result, or a transparent before/after. Fake social proof trains buyers to distrust you-and trains your team to believe theatre equals traction.

Place proof next to the claim on the landing page-not in a carousel three scrolls down. For conversion-focused page structure, Baymard Institute research on ecommerce UX is a useful reminder that trust and clarity beat decorative complexity-even when you are selling software, not shoes.

4. What is the conversion event we will actually measure?

Not “brand awareness.” A booked call, activated trial, or paid conversion-tied to a follow-up system that someone owns. If the event cannot be named, the budget cannot be judged.

Define the event tightly. “Signup” is often vanity if activation is broken. Prefer: “trial user completes first meaningful outcome within 24 hours” or “qualified demo booked with ICP title.” Then instrument it. Then assign a human to respond when it fires. Orphaned leads are how £10,000 turns into a CRM graveyard.

1. Name the primary conversion event in one line.
2. Name the secondary learning metrics (CTR, landing bounce, activation).
3. Set a kill threshold before you launch (e.g. cost per qualified demo above X after Y spend).
4. Confirm who replies within one business day-and what they say.

If your website cannot carry a cold visitor to that event in a clear path, fix the path first. See why most startup websites fail in the first 10 seconds and landing page review: is yours helping or hurting conversions?.

5. What will we learn if the campaign fails?

If failure only means “ads didn’t work,” you learned nothing useful. Design the spend to teach: messaging, audience, or offer. A failed test with a clear lesson is still cheaper than a vague success.

Structure the £10k as experiments, not as a single heroic push. Example split: message tests on a small budget before scaling winners; one audience hypothesis at a time; landing variants that change one belief gap, not seventeen colours. When something fails, you should be able to say which hypothesis died.

• Hypothesis A: ICP X responds to outcome Y more than feature Z.
• Hypothesis B: social proof near the CTA lifts demo rate more than a longer feature list.
• Hypothesis C: a tighter exclusion list lowers CPC waste even if volume drops.

This is how teams like those profiled in First Round Review talk about early go-to-market: learning velocity beats vanity scale. Your £10k should buy learning velocity.

How to allocate the £10k once the questions are answered

There is no universal split. There is a sensible order.

1. Allocate part of the budget to message testing, not only media.
2. Fix the landing path before scaling winners.
3. Kill channels that cannot explain the product better than the page.
4. Reserve budget for follow-up capacity (someone has to talk to humans).
5. Scale only the combination of audience + message + page that produces the named event at an acceptable cost.

A fictional example: Northline, a compliance tool for fintech startups, nearly spent £10k on LinkedIn ads with a “AI-native compliance OS” headline. After the five questions, they rewrote for “fintech ops leads pre-Series B who dread audit season,” put a 90-day pilot metric on the page, and measured “audit checklist kick-off booked.” Half the budget tested two messages. One won. They scaled that. The other half would have been spent teaching the internet a slogan nobody needed.

Checklist before you load the ad account

• One ICP written with exclusions.
• One outcome sentence used in ad + hero.
• Proof placed beside the main claim.
• Primary conversion event defined and instrumented.
• Follow-up owner named.
• Failure hypotheses written in advance.
• Kill threshold agreed by founders-not “we’ll see how it feels.”

If you are still deciding whether spend is even the right next move, the most expensive mistake early-stage founders keep making is scaling an unclear offer. And if positioning itself is mushy, reposition in 60 minutes before you fund the amplifier.

Channel choice after the questions-not before

Founders often start with “Should we do LinkedIn, Google, or influencers?” That is a later question. Channels are pipes. The five questions decide whether the water is worth pumping. Once answered, choose the pipe that reaches your ICP with the least translation loss.

• Search works when intent exists and your page matches the query’s job-not when you hope to invent demand with a vague brand term.
• LinkedIn/social works when the message is sharp enough to survive a scroll and the CTA is obvious.
• Partnerships work when the partner’s audience is your ICP, not merely adjacent and flattered.
• Content works when it answers a real job and routes to a conversion event-not when it exists to “feed the algorithm.”

If you cannot explain why a channel reaches your one customer with your one outcome, you are choosing fashion. Fashion is a poor CFO. For a broader take on spend sequencing, where the first £1,000 often goes wrong is the smaller-budget cousin of this article.

A sample £10k learning plan (not a template religion)

One workable shape for a B2B seed-stage team with a clear offer:

1. £1,500: landing and analytics hygiene-event tracking, speed, mobile CTA visibility, proof placement.
2. £3,000: message and audience tests on one channel (three ads × two audiences, capped).
3. £3,500: scale the winning combination only after the conversion event hits an agreed cost band.
4. £1,500: founder-led follow-up capacity-calls, Loom replies, onboarding help for early converters.
5. £500: contingency for the surprise that always appears (tracking breaks, creative fatigue, a form bug).

Notice how much of that is not “media.” Learning systems include the page, the measurement, and the humans who respond. Buying clicks into a silent CRM is how teams conclude their category cannot be advertised. Often the category is fine. The system is hollow.

Red flags that you should stop mid-campaign

• CTR is fine but the conversion event never fires-inspect the page and the offer, not only the bid strategy.
• Sales says every lead needs a 20-minute “what we do” primer-your creative is over-promising or your page is under-explaining.
• You keep widening the audience to “get volume”-you are buying cheaper confusion.
• Nobody owns reply latency-leads older than a day are archaeological artefacts, not opportunities.
• You cannot state which hypothesis is being tested this week-you are pacing spend, not learning.

Stopping feels like failure. Continuing a broken test feels like diligence. Diligence without a kill switch is just denial with a credit card. Operator writing on First Round Review often returns to this: speed of learning beats stubbornness dressed as grit.

Answer the five questions on one page. If the page feels thin, the budget will feel thinner after the invoice. And if the real issue is that founders need something other than more marketing theatre, read what founders actually need isn’t more marketing.