This is a composite. The shape of the engagement is drawn from a pattern we have seen play out across several UK marketing and creative agencies — not from one specific client, and not with any client’s permission to name them. We prefer composites to named studies for two reasons: the named studies in our industry tend to be varnished, and the clients in question are small enough businesses that their identifiable problems are their private business. What follows is the accurate shape of the work, with the identifying details changed.

The agency

Twelve people. Based in a Midlands market town. Formed by two founders about seven years before we met them, out of a freelance practice that got too busy. Work is a mix of brand strategy, campaign design, and a growing line of digital production. Clients are SMEs and a handful of mid-market retailers. Revenue somewhere in the low seven figures. Profitable, though not as profitable as the two founders would like.

On the day of our first discovery call, one founder said: “I don’t think we’re bad at what we do, but I think we’re bad at the space around it. The meetings before, the handovers during, the invoicing after. It’s chaotic in a way that makes me think we’re losing money.”

This turns out to be a common shape.

The presenting problem

What had finally pushed the founder to write in was a specific incident. A prospect had reached out through a LinkedIn DM to one of the account directors, who had caught it late on a Friday and dropped it in a Slack channel with a “can someone pick this up?” at 6:47pm. Nobody did. Three weeks later, the same prospect appeared on the homepage of a competing agency as a signed client.

The founder was not sure how many of these there had been. He suspected, rightly, that this one was not the first.

The presenting problem was lost enquiries. The underlying problem, as often, was larger.

What we found during diagnosis

In the week between our first call and the written diagnosis, we sat on a call with the account directors (two of them), the operations lead (one), and each of the founders separately. We also spent an afternoon looking at six months of their enquiry history across the four channels they admitted to and the two they had forgotten they used.

The findings:

Enquiries arrived across six channels, not four. The website contact form, a dedicated sales email, direct emails to the account directors’ addresses (the addresses were on their LinkedIn profiles), LinkedIn DMs to the founders and directors, Instagram DMs to the agency’s public account, and a contact number that rang the operations lead’s desk phone. The operations lead admitted that the desk phone was often unanswered because “the calls are almost always sales”.

There was no single record of enquiries. Each channel had its own thread of history, visible only to the person whose inbox it was. The operations lead kept an informal spreadsheet but hadn’t updated it in six weeks.

Response times varied from minutes to never. We measured this by tagging a sample of fifty enquiries over three months against their eventual outcome. Enquiries acknowledged within an hour converted at 34%. Enquiries acknowledged between one and twenty-four hours converted at 19%. Enquiries acknowledged after twenty-four hours converted at 7%. Enquiries never acknowledged — and there were eleven of these in the sample — converted at 0%.

Onboarding was scrambled, separately. Winning the client was where the trouble began. Each new client was onboarded differently by the account director who had won them. Documents were requested by email threads that later had to be dug up. The first month of a new engagement was described, independently, by four different people on the team, as “the hard bit”. Two clients had churned in the previous year citing a chaotic start.

The founders were triaging in the evenings. Both of them. One had a Slack channel on mute on his phone during dinner that he checked anyway. The other had a habit of reading LinkedIn DMs in bed. This was the quiet cost the agency was paying: two people’s evenings, daily.

What we wrote in the diagnosis

The diagnosis document, sent three working days after the interviews, was three pages. It described what we had heard; named the six channels; quantified the conversion penalty of slow response (we estimated, conservatively, £180,000 a year in lost revenue, not including the onboarding churn); and described three possible scopes for the engagement.

Narrow. Fix the enquiry intake only. A single inbox, channel consolidation, response-time SLA and automation. About four weeks of work; £3,500 fixed; £700 a month retainer for maintenance. Would address roughly 70% of the diagnosed problem.

Standard. Fix the enquiry intake and the onboarding, as a connected engagement. About eight weeks of work; £5,800 fixed; £1,100 a month retainer. Would address roughly 90% of the diagnosed problem, and was the option we recommended.

Broad. Fix intake, onboarding, and a third piece — the handover from the account director to the creative team, which we had noticed in the interviews and which surfaced during onboarding — as a three-part engagement. About twelve weeks; £8,400; £1,500 a month retainer.

The founder chose Standard. We did not try to upsell him into Broad; the third piece wasn’t urgent yet and would have benefited from the first two being in place for a couple of months before we touched it.

What we built

Over seven weeks (one more than we estimated, for reasons we’ll come to), we built the following.

A single enquiry pipeline in Airtable. Six channels consolidated:

  • Website contact form → webhook → Airtable.
  • Dedicated sales email → IMAP listener via n8n → Airtable.
  • Account directors’ direct emails → a lightweight forwarding rule + an n8n filter that recognised “new client” or “work with you” phrasing → Airtable, with a flag for the director to confirm.
  • LinkedIn DMs → a polling worker that ran every fifteen minutes against the founders’ and directors’ accounts, using their own credentials, with clear opt-out → Airtable.
  • Instagram DMs → Meta’s Graph API via a verified business integration → Airtable.
  • Phone line → a number-forwarding rule sending missed-call alerts and voicemails (transcribed) → Airtable.

A light qualification pass, using a small language model (llama 3.1 8B, running on the agency’s own infrastructure — they already had a small server for another purpose, and we preferred to keep enquiries off a third-party API for simple privacy reasons). The pass extracted the prospective client’s name, company, approximate size, service area, and urgency, and drafted a first reply for the relevant account director to review.

A Slack integration that replaced the founders’ evening triage with a single 9am summary to the operations lead: new enquiries in the last 24 hours, anything sitting for more than two working days, anything that had been classified as urgent. No individual alerts. No buzzing at 6:47pm.

A ten-minute Monday morning ritual for the operations lead, in which she reviewed the week’s inbound, confirmed routing, and occasionally corrected the language model’s qualification tags. We designed this to remove the sense that the system was running without supervision — and because we have learned that teams trust automation more when one human has a small, predictable audit role.

For the onboarding side: two templated sequences (retainer client, project client), triggered from the CRM when a deal moved to “signed”. Branded welcome email, document-collection form, kick-off call scheduling link, and a per-client Notion page that pre-populated with the engagement’s specifics. Both sequences ran themselves; the account directors’ only new task was to review the kick-off agenda for the client’s specific situation.

What went wrong during the build

We underestimated one thing. The LinkedIn DM ingestion was more fiddly than we expected because LinkedIn’s posture toward third-party access has been moving during this period. The polling worker we had planned to build against the LinkedIn Sales Navigator API hit rate limits sooner than we liked, and we ended up using a more conservative approach that polled less frequently and asked the founders to confirm any DMs that looked like enquiries. This added about a week to the build.

We also discovered, during testing, that the agency’s Stripe account and their QuickBooks were out of sync in a way that meant several existing clients had been partially-invoiced. We flagged this and left it for their finance lead to clear up. We did not fix it for them — it wasn’t in scope, and tidying someone else’s accounting mid-engagement tends to create a mess we don’t want to own.

What happened after launch

The tracking story, honestly reported:

First-response time went from a mean of 28 hours to a mean of 41 minutes, inside three weeks. This was expected; the system was designed for it.

Lost enquiries — the ones the sample audit had shown never got acknowledged — went to zero. This is a stronger claim than we like to make, but the data was clean: every enquiry into the pipeline had an owner within the working day, every week for six months.

The agency’s new-business conversion rate, measured as signed contracts per enquiry, moved from 11.4% to 17.1% over two quarters. This is a larger effect than the response-time improvement on its own should produce, and we suspect the qualification and routing improvements contributed as much as the raw speed. We cannot prove this causally.

Founder evening triage dropped to essentially zero. The 9am summary replaced the evening checks. One founder later admitted that the loss of evening triage was harder to adjust to than he’d expected — a quiet addiction to the feeling of being on top of things. He described it to us, not entirely joking, as “grieving the chaos”. Both founders were, six months in, sleeping better.

The onboarding improvements took longer to show in the numbers. First-invoice-to-first-deliverable time moved from an average of 19 days to 9 days. Client NPS on the first-month question moved from 6.8 (where it had been for two years) to 8.9 over the following three quarters. Churn in the first year of a new engagement fell from 15% to 6%, but this change took a full year to become visible in the data and we would not claim it with certainty from a single year.

What we would do differently

Two things.

We should have insisted on addressing the creative-team handover (the “Broad” scope) at the same time as the onboarding. We talked the founder out of it, reasonably, on the basis that it wasn’t urgent. Within three months of launch, the handover had become the visibly-worst part of the operation, and we ended up doing it as a separate engagement later. We would, in hindsight, have preferred to build it as part of a connected whole.

We should have been firmer about a clean LinkedIn solution. The conservative polling approach we ended up with works, but it has required more maintenance than we would like. If we were building the same system today, we would either integrate more deeply with LinkedIn’s Sales Navigator (expensive but clean) or we would simply route LinkedIn DMs to a dedicated review step rather than attempt to ingest them automatically. The middle ground was the uncomfortable answer.

What the agency is paying us now

£1,100 a month. This covers: monitoring and occasional fix-ups to the pipeline, quarterly reviews, and a few hours of flexible work for small extensions (a new source added, a template changed, a report tweaked). The retainer is cancellable month-to-month. The founders have, at the time of writing, kept us on for fourteen months.

The setup phase, the bigger spend, ended when we said it would end. We have not charged them for fixes to the setup since then. This is how we try to make retainers fair — the thing they are paying for monthly is the ongoing care, not the amortisation of the original build.

The general shape

The specifics above are bespoke to this client. The shape — consolidated intake, lightweight qualification, templated but personal response, scheduled onboarding — is not bespoke. It is roughly what we have built for a half-dozen agencies of similar size, because the same problems produce the same solutions once you strip away the local details.

If the shape of the above sounds like your own business, the fit is probably good. Come and talk to us.