Written for one buyer at a time.

Arrived on a recommendation and is now vetting it independently.

• pillarWe ran our own methodology on our own website. It scored 41 out of 100.Five buyer archetypes read audienceintel.app independently and none of them was convinced. Published in full, deductions included, because it is the one test a supplier cannot select the result of.planned
• postQuestions to ask before you hire a marketing consultantEight questions a case study cannot answer, what a good answer sounds like, and the two evasions that should end the meeting.planned
• postA supplier who only shows you wins has told you nothing about their methodYou are reading the portfolio the supplier chose. What to ask for instead, what a real answer contains, and what a refusal tells you about the next twelve months.planned
• postHow to check a provider's claims: a working methodOpen the claim, find the source, check the fieldwork year, check who paid for it. Four steps, under an hour, and no need to be rude about any of it.planned
• postA referral gets you the meeting. It does not get you the instruction.Hinge finds referral is still the first way buyers discover a professional-services firm, and that buyer-perceived visibility of those firms fell by more than a third between 2020 and 2022. The website is where a recommendation is quietly confirmed or dropped.planned
• postFour things you should refuse to accept from any supplierA countdown on a professional engagement, a figure with nothing behind it, a page silently dropped from an audit, and a score that grades you against a crawler rather than a buyer. All four are checkable before you sign anything.planned
• postAsk what your audit did with the pages it could not loadA page that failed to capture is either recorded as absent or quietly removed from the denominator. One of those flatters the firm paying for the audit, and the report will not tell you which happened.planned
• postWhere did that number come from? What a good answer sounds likeTake one statistic off a supplier's slide and follow it back to the fieldwork year, the sample and the sponsor. Most do not survive the second step.planned
• postWhat a supplier does when they are wrong is the only reference worth takingCorrection behaviour is the one part of the record a supplier cannot curate. Ask for the error, the date it was found, who found it, and what was published afterwards.planned
• postAsk for the engagement that went wrong, not the one that went wellGartner's regret research puts high-regret purchases seven to ten months behind no-regret ones, with 89% of high-regret buyers citing conflicting objectives inside their own firm. A supplier who has never met that has never been in the room.planned
• postReference calls are a weak test. Here is a stronger one.Referees are chosen by the supplier, briefed by the supplier and grateful to the supplier. A published method, a documented failure and a claim chain you can follow are none of those things.planned
• postWhat it means when a firm will not put a price on the pageMcKinsey finds B2B buyers willing to commit six figures through self-service channels. Quote on request is a decision to be absent from that moment, and it is usually a decision about the seller rather than about the work.planned
• postA Referred Sceptic scorecard, itemised: 50 out of 100 and where every point wentThe archetype closest to a buyer arriving on a recommendation, scored line by line — the places where a transfer of trust from the person who named you survives the homepage, and the places it stops.planned
• postA case study, a testimonial and a result are three different thingsEdelman finds decision-makers trust a firm's thought leadership more than its marketing, and rate almost none of what they read as excellent. Here is the test that separates the three, and which one a buyer can actually check.planned
• postWho actually does the work when the person you were referred to is busy?Depth of bench is the question owner-managed firms are asked least and answer worst. Three ways to ask it so the answer comes back checkable rather than reassuring.planned
• postUrgency works on consumers. On a firm choosing an adviser it does the opposite.Scarcity is one of the six biases Google's Decoding Decisions found reliably moves buyers, across an experiment simulating 310,000 purchase scenarios. The same behavioural work explains why a countdown on a professional engagement reads as a reason to leave.planned
• postThe finding that survives every redesign: nobody's name is on the siteWillis and Todorov measured trustworthiness judgements forming from a face in a tenth of a second. A firm that sells judgement while showing no one is asking a buyer to trust an abstraction.planned
• postWhen the person who referred you cannot explain what the supplier actually doesAdamson puts the buyer's struggle to make sense of the choice at the centre of selling. A referral that cannot survive one clarifying question is a lead, not an endorsement.planned
• postTry to break a claim: three checks a sceptical buyer should run on any supplierOpen the claim, follow it to the source, check the sample and the year it was collected. Run the drill here first, then on the next three suppliers who send you a deck.planned
• postWhen a website reading is not worth buying, and how to make a supplier say soThree situations in which a perception reading will tell a firm nothing it can act on. Any supplier worth instructing can name theirs before you pay, which is the only moment the answer is worth anything.planned
• teardownAvis admitted the one thing every buyer already knewWhy an admission you cannot avoid is worth far more volunteered than extracted, and how to find yours before a prospect finds it for you.planned
• teardownOatly published the case against itThe sceptic's decision rule is: let me see the document. What changes when a firm hands it over unasked, and why the case against you is less damaging in your own words than in a competitor's.planned
• teardownThe apology that was better evidence than the advertising it interruptedHow a firm behaves once it has visibly failed is the most informative thing a sceptical buyer ever gets to see, and most firms spend it on lawyers instead.planned

Buying this kind of thing for the first time. Reads for reassurance.

• pillarWhat five different buyers see when they open the same websiteOne page, read five times by five different buyers. They stop in different places, and the disagreement between them is the part worth reading.planned
• postShould you put your photo on your business website?Trustworthiness is judged from a face in about a tenth of a second, and a longer look does not revise the verdict — it hardens it. A site with no named human gives a buyer nothing to judge but the copy.planned
• postUrgency copy is costing you the clients it was meant to winScarcity reliably moves a consumer buyer — it is one of the six biases Google's behavioural work found dominating purchase loops. On a firm choosing a professional adviser, against a population where 68% already believe business leaders mislead them, the same device works in reverse.planned
• postWhat happens after you click runWhat the free reading does, in order, and what it will not tell you at the end of it.planned
• postWhat should a website cost a firm your size?No honest answer is a single number. What actually differs between two quotes, which of those differences a buyer ever sees, and the line items you end up paying for twice.planned
• postThe marketing words you are allowed not to knowPlain English for the terms an agency will use in the first meeting, and an honest note on which of them are doing real work.planned
• postWhat a website audit report should contain — and what padding looks likeMost reports are a list of technical faults with a logo on the front. What to check before you pay for one, and the three sections that carry all the value.planned
• postHow many agencies should you actually talk to?Three per cent of customers bought when shown twenty-four jams; 30% bought when shown six. Collecting nine quotes is not diligence — it is a way of never deciding.planned
• postThe obvious questions you are allowed to ask a marketing supplierNobody in the room will think less of you for asking what a metric means. Eight questions any supplier should be able to answer on the spot, with what a good answer sounds like.planned
• postWhat a bad website proposal looks likeSix things that appear in proposals from suppliers you should not hire, and the reason each one is a tell rather than a preference.planned
• postWhy every marketing website feels like it is selling you somethingSix in ten people report grievance toward business, and 68% believe business leaders deliberately mislead. Your suspicion is well calibrated — here is how to test a claim anyway.planned
• postYou probably do not need a new website. You need a different first page.Most visitors leave a page inside ten to twenty seconds, and the impression that decides it forms before any of the copy is read. What that means for the quote in front of you.planned
• postWhat normal looks like: the pages a firm your size is expected to haveNot best practice and not a wish list — the set a buyer treats as the minimum, and what they conclude when one of them is missing.planned
• post“Nobody can find us” is usually a different problem than you thinkBuyer-perceived visibility of professional-services firms fell 37% in two years. Being findable and being remembered are separate jobs with separate fixes, and most firms buy the wrong one.planned
• post“We get all our work by referral” — the part that stopped being trueReferrals are still the number one way buyers discover a firm. They are no longer where the buyer stops looking, and the next stop is your website.planned
• postWhat happens when a buyer asks an AI about your firmGartner puts AI use somewhere in a recent purchase at 45%, from a sample of 646 buyers. TrustRadius's January 2026 wave finds 94% of its AI users fact-check what they are told. Both facts point at the same thing: what your site puts in writing.planned
• postThe regret is not usually about the work. It is about the buying.Sixty per cent of technology buyers regret nearly every renewal, and high-regret feeling peaks before anything is delivered. What that tells a first-time buyer to check first.planned
• postThe sentence missing from almost every contact pageWhat happens after the form is sent: who reads it, when, and what they do next. Almost no contact page says, and the first-time buyer is exactly the reader who needs to know before pressing anything.planned
• postThe stock photograph is not neutral. It costs you something.A brand cue changes what the brain prefers even when the product is chemically identical. A generic image is a cue too, and it is not saying what you hope it says.planned
• postIs a free website report just a way of getting your email address?Often, yes. Four tests that separate a lead magnet from a reading, including what a report should be willing to tell you before it asks for anything. Apply them to this one too.planned
• teardownDe Beers did not sell a stone. It sold a rule for how much to spend.Our reading of the campaign that supplied a price expectation to buyers who had none, and what it implies for any firm whose client cannot guess what the work should cost.planned
• teardownIntel Inside: a badge for a component nobody could evaluateWhen a buyer cannot assess the work itself they assess the marks on the outside of it — which is precisely what the accreditation logos in your footer are doing, well or badly.planned
• teardownSpecsavers sells the consequence of choosing badly, not the benefit of choosing themA teardown of regret used as an instrument, and the version of it a regulated firm can use without becoming the thing its own clients distrust.planned

Scanning for something actionable. Leaves the moment it is not.

• pillarWebsite audit for professional-services firms: the seven checks a buyer makes firstMost visitors leave a page inside ten to twenty seconds, and the majority arrive with a supplier already in mind. Seven structural checks, each either present on the page or not, runnable in an hour without a tool.planned
• postWebsite audit checklist for law firmsOnly 8% of UK SMEs with a legal issue consult a solicitor at all, and fewer than 15% think solicitors are cost-effective. The checklist answers both objections on the page, before anyone has to raise them in a meeting.planned
• postWebsite audit checklist for accountants and IFAsFees, regulation, and the two things a prospective client looks for that most practice websites leave out entirely.planned
• postSEO audit vs buyer-perception reading: what each one catches, and what neither doesA technical audit checks whether a page can be found. It does not check what the page says to the person who found it — and a site can be flawless on the first and lose every reader at the same paragraph.planned
• postPut a number on your fees page this weekFewer than 15% of UK SMEs think solicitors offer cost-effective solutions, and only 8% ever consult one. A published price answers that objection while the buyer is still reading, rather than in a meeting they never book.planned
• postDelete half your service lines from the homepageSix jams outsold twenty-four by roughly ten to one. A firm presenting fourteen practice areas with equal weight is running the twenty-four-jam experiment on its own buyers.planned
• postYou are competing for one of about three slots. Write the page accordingly.Eighty-three per cent of technology buyers shortlisted three or fewer suppliers; the average list held 2.7. What that changes about the page a prospect lands on.planned
• postYour team page is a photo gallery. Make it a decision aid.Trustworthiness is judged from a face in about a tenth of a second, and referrals are still the number one way buyers find a firm. A grid of unnamed faces answers neither. Names, faces, and who to call for what.planned
• postReplace three testimonials with one case study that names a numberUnattributed praise is free to produce, so it carries no information. Eighty-six per cent of decision-makers say strong thought leadership earns an invitation to tender; only 15% rate what they actually read as excellent.planned
• postFix these four things first, in this orderRanked by the size of the gap each one closes, not by how quickly it can be done: load time, then the headline, then the named human, then the single action. With the test that tells you whether it landed.planned
• postStop putting “book a call” in front of everything a buyer needsDecision-makers now use ten interaction channels per purchase, and 39% will spend over $500,000 through self-service or remote channels. The gate is not qualifying anyone; it is filtering you out.planned
• pillarCheck what your website tells a buyer, before you change anythingWhat the free reading looks at, what it deliberately withholds, and the four things it will not do — stated before you run it rather than discovered afterwards.planned
• postFive signs your website is losing you clients, checkable todayEach one is structural — either it is on the page or it is not. No opinions, no redesign, no meeting.planned
• postWhat the free reading includes — and what it doesn'tSaid plainly, because a visitor who mistakes the free check for the full report never buys the full report.planned
• postCut your enquiry form to three fields. Here is the reasoning.Behaviour needs motivation, ability and a prompt in the same moment. Every extra field is a tax on ability — what to keep, what to delete, and how to tell it worked.planned
• postIf your site takes over three seconds on mobile, fix that before the copyFifty-three per cent of mobile visits are abandoned past three seconds. Copy rewritten for readers who never arrive is wasted work.planned
• postOne action per page, and it is not “get in touch”Behaviour needs motivation, ability and a prompt in the same moment. A page offering four equal actions supplies no prompt at all; a page whose only action is a form supplies no motivation. Here is the repair, in order.planned
• postIf a stranger cannot say what you do from your H1 alone, rewrite itA visual impression forms in around fifty milliseconds; the headline is what the reader checks it against. Hand the page to someone outside your industry and ask what you sell. It is the cheapest test on this list.planned
• postRead your own four main pages in one sitting. They probably contradict each other.Pages written in different years by different people, each true when it was written. A buyer reads them consecutively and finds the disagreement the firm never sees. One pass, no tools, free.planned
• postHow to tell a website change workedThree measurements you can take straight after a change, the one that will lie to you, and what a real improvement looks like when it arrives.planned
• teardownThe Ronseal line is the hardest thing on this list to copyIt reads as a joke about plain English; our reading is that it is a decision rule aimed squarely at the buyer who will not reach your second paragraph.planned
• teardownFour identical price-comparison sites, one meerkatA teardown of retrieval rather than persuasion: when every offer in a market is the same, the firm that comes to mind first is the one that gets checked at all.planned
• teardownA rail-safety message that did not read like a rail-safety messageHow a mandatory-feeling message got itself processed at all, and why the same problem faces every firm publishing something it is obliged to publish.planned

Reading the evidence before forming a view. Rewards citation depth.

• pillarHow B2B buyers actually judge your website: what the research saysThirty primary sources on how a professional-services buyer reads a website — what each one measured, on whom, in which year, and where the evidence runs out.planned
• postThe fifty-millisecond first impression: what the research actually foundThe most-quoted finding in web design, read back against the papers it comes from — what Lindgaard's fifty milliseconds measured, and the purchase claim it will not carry.planned
• postWhy survey data misses what buyers actually doAsking people what they think is a different measurement from watching where they stop. Two instruments, two answers, and a rule for which one to believe when they disagree.planned
• postRetired sources: the statistics that outlived the studies behind themSeveral of the most-quoted figures in B2B marketing were withdrawn by their own authors, disavowed, or never traceable to a document at all — and they are still in circulation. How to spot one before it reaches a board pack.planned
• postThin slices: what Ambady and Rosenthal measured, and why a website is only an analogyAn effect size of r = .39 across thirty-eight studies of people judging people. Used here as an analogy for how a homepage is read, and said so rather than dressed up as a website finding.planned
• postProcessing fluency: why an easier page reads as a more competent firmReber and colleagues tie ease of processing to judgements of beauty, truth and trust. What that implies for the service page a fee-earner wrote in a hurry.planned
• postEvery service line on one page: the jam experiment and its limitsSix jams outsold twenty-four by roughly tenfold, and later meta-analysis found the effect is context-dependent. Both halves matter when a practice decides what to stop advertising.planned
• postYour enquiry form is an ability problem, not a motivation problemFogg's B = MAP holds that behaviour needs motivation, ability and a prompt in the same moment. A ten-field enquiry form removes the one thing the page had already won.planned
• postWhat the Coke study says about your brand, and what it does notBrand knowledge changed stated preference and measured brain response for an identical liquid. It did not show that a better logo makes a worse adviser acceptable.planned
• postSystem 1 chooses the firm. System 2 writes the minutes.Kahneman's dual-process account applied to the partnership meeting where a decision already taken is given its reasons.planned
• postTen to twenty seconds: what the dwell-time research actually showsNielsen Norman's survival curve describes when people leave a page, not when they decide to instruct a firm. The distinction changes what you fix first.planned
• postThe shortlist is about three names long before anyone picks up the phoneTwo independent instruments, different samples, different years, converging on a very short list — and both drawn from technology buyers, which is stated every time they are used.planned
• postWhere the evidence is thin: three citations in wide circulation that will not take weightA client-only report held up through trade press, a vendor with a commercial stake in its own answer, and a survey whose question design moved between waves. Named, each with the reason.planned
• postCan a language model stand in for a buyer? What the Stanford agent study licensedAgents matched real people at 82–86% of those people's own retest consistency, against a 74% demographic baseline — on attitudes and economic games, not on choosing a supplier.planned
• postWhen a machine could first look at a website the way a person doesOctober 2024, at 14.9% on the leading agent benchmark against roughly 70% for humans. Why that date matters to any firm buying an AI-based analysis of its own site.planned
• postHow many buyers actually use AI during a purchase — and what the number countsGartner measured 45% in a recent purchase across 646 buyers. TrustRadius's own figure moved so far in twelve months that it reads as question design, not behaviour.planned
• postIndependent research is most of the buying time. Meeting suppliers is a sliver.Adamson's split is a statement about share of attention, divided across every supplier considered — not a claim about how far a decision has travelled. The difference is the whole argument.planned
• postTrust is the scarce input: the Trust Barometer read as a marketing constraintA UK Trust Index of 43 and a majority who believe business leaders deliberately mislead. What that does to the claims a small firm can credibly make about itself.planned
• postOnly 8% of SMEs with a legal problem call a solicitorThe Legal Services Board's demand research says the main competitor is doing nothing, or asking the accountant. That is the question a firm's website has to answer before it argues about fees.planned
• postFifteen years after the Zero Moment of Truth: what actually changed10.4 sources in 2011, 27 interactions per buyer journey by 2021, ten channels per purchase by 2024 — three different instruments, read side by side rather than stacked into one number.planned
• teardownMichelin gave away the guide before it sold the tyresWhy a genuinely useful reference work outperforms a brochure with a buyer who has not yet decided you are worth an email, and what the equivalent costs a firm that bills by the hour.planned
• teardownThe Rolls-Royce advertisement that made one measurable detail do all the workOne checkable claim carrying a full page of copy, read as a decision rule: give a researcher something they can verify and let them persuade themselves.planned
• teardownVolvo Trucks stopped describing the feature and demonstrated itWhy a filmed demonstration answers a question a specification sheet cannot, and what the equivalent looks like for a firm that sells judgement rather than machinery.planned

Has bought this before. Looking for the machinery, and for tells.

• pillarBuyer-perception data in your stack: the report, the API and the MCP serverHow the reading is produced — capture, rubric, weighting — and how to drive it from your own tools rather than from someone else's dashboard.planned
• postWiring buyer-perception data into Claude, Cursor and your own toolsThe MCP surface, the scopes it asks for, and what each tool actually costs to run.planned
• postHow the report is built: capture, rubric, weightingThe method, stated so you can disagree with it specifically rather than in general.planned
• postThe quarterly refresh: measuring perception driftWhat changes between readings, what it means when nothing does, and the change that is real but too small to act on.planned
• postHow a score of 38 is derived, line by lineA worked example on the site we know best — ours. Every stage weight, every deduction, and the rubric line each one came from, so the disagreement can be with a specific line rather than with the number.planned
• postFour things this method will not score — and why the exclusions belong in writingEvery instrument has a boundary. The ones worth buying state it before the invoice: what is out of scope, why, and what you should go elsewhere for instead.planned
• postIs it a measurement or an opinion? A test you can apply to any report you are soldA measurement names its instrument, its sample and its date; most agency audits name none of the three, and the ones that do are worth reading twice.planned
• postWhen a source is retired, what happens to the report you already haveSources get withdrawn. The question is whether your supplier has a procedure for it — versioning, notification, and a replacement citation that arrives with its sample attached. Gartner's 45%, n = 646, is what one of those looks like.planned
• postThe tells that a proposal is a template with your logo on itEach one is checkable in the document itself, before you spend a second meeting finding out.planned
• postWhat a single reading costs to run: capture, inference, storage, human passInference for a GPT-3.5-equivalent model fell roughly 280-fold in under two years, which changes what an analysis can honestly be sold for. The four cost lines, and what they imply for pricing to a firm turning over £1m to £5m.planned
• postTwo survey waves are not a trend lineTrustRadius's own AI-usage figure moved from 7% in January 2025 to 63% in January 2026. That is question design moving, not buyers — and joining the two points would be the most persuasive lie on the page.planned
• postA supplier's self-audit is not a case studyPublishing your own worst score proves a method can embarrass its author. It does not prove the method predicts revenue — and a supplier who presents it as though it does has just told you which claim to check next.planned
• postDepth of bench: what to ask when the firm you are hiring is one personThe load-bearing objection to a one-person supplier is not competence, it is continuity. The questions that actually resolve it, and the answers that should end the conversation.planned
• postPI cover, DPA, per-client caps: the terms a second-time buyer checks firstMost supplier terms are silent on the three things that decide what happens when the work goes wrong. What to demand in writing, in the order it matters, and what a twenty-six-word service description is actually telling you.planned
• postThe best evidence on most B2B sites is on the page built for machinesSpecification tables, version histories, machine-readable summaries: a technical buyer reads them first, and firms file them behind URLs designed for crawlers. The most fluent page on the site is the one nobody was meant to read.planned
• postWhat synthetic buyers can predict, and what they cannotThe Stanford work licenses simulating individuals over guessing from demographics — 82–86% of a person's own retest consistency against a 74% demographic baseline. It does not license predicting how anyone judges your website, and that limit belongs on the front page of any report built this way.planned
• postIf the shortlist holds 2.7 names, most of your marketing budget is noiseTwo independent instruments, different samples and different years, put the shortlist at roughly three names and formed early — here is what that does to the spend of an owner-managed firm turning over £1m to £5m.planned
• postWhen this reading disagrees with your analytics, which one is wrong?They are measuring different things, so the answer is usually neither — a rule for deciding which one to act on, and the case where the disagreement is itself the finding.planned
• postHow absence is recorded: the page that would not loadA firm scored only on the pages that loaded looks better than it is. So absence is marked explicitly, and the crawl limit that binds the reading is stated rather than left for you to discover.planned
• postA ban that catches one phrasing tests the phrasing, not the claimContent governance fails quietly: the automated check passes, and the retired claim is still on the page in prose. What a guard has to catch to be worth anything, and what a green tick is actually evidence of.planned
• teardownDomino's made the process visible, and the promise redundantA teardown of status as marketing: why an experienced buyer trusts a progress screen more than any reassurance you could write about the same work.planned
• teardownWhen the price is the messageOur reading of a campaign that made cost the whole argument, and what it says about firms that keep their fees back until the second meeting.planned
• teardownThe Economist posters left something out on purposeA teardown of deliberate difficulty — when making the reader do a little work is a qualification device rather than a failure of clarity, and when it is just obscurity.planned