Most agencies send their first invoice before they know who buys from you, why, or what those people are worth.
Search for a digital marketing agency for small business and the results are landing pages and directories. Discovery is the part that matters most when choosing a digital marketing agency in the UK. One London agency lists four retainer packages from £845 to £4,225 plus VAT on a six-month minimum term, chosen on a 30-minute call.
A 30-minute call picks a package. It does not find out who your customers are.
We run an agency in Northampton, so this is partly a note against our own trade. A 2021 CallRail survey of 600 small business owners found that, among those unhappy with a past agency, 46% said the agency was not a strategic business partner. A further 48% said the agency could not help beyond lead generation, and 43% said the programme was not customised to their brand.
Those three complaints are one complaint. The agency did not learn the business before it started spending the budget.
Why the first invoice usually arrives before the first real question
The first invoice arrives early because agency cash flow depends on it, and because a 30-minute sales call is cheaper than two weeks of interviews. A retainer signed in week one covers the account manager's salary from week one. A discovery phase that produces no campaign for a fortnight looks, from inside the agency, like unbilled time.
Owners accept it because they are busy. In the same CallRail research, 74% of owners said they would rather spend time on the core business than on marketing, and 44% were too pressed for time to plan it at all. An agency that says "leave it with us" is offering relief.
The question is what it does with the budget once the owner has stopped looking.
The answer, too often, is a template. CallRail's own report on agency relationships found that only 32% of agencies take the time to understand a client's brand position and voice. If two thirds skip the brand, it is a safe bet that more skip the customers.
What skipping discovery produces
Skipping discovery produces a campaign that could have been built for any business in your category. The keyword list is the one the tool suggested. The ad copy says "trusted local experts" because nobody asked what customers say when they ring.
The audience is "people aged 25 to 55 within 20 miles" because nobody looked at where the last 20 customers came from.
Generic campaigns still get clicks, which is why they survive the first monthly report. What they do not get is the right clicks. A plumber whose profit comes from boiler installations and whose losses come from emergency call-outs will be shown a campaign that generates emergency call-outs, because those are the easiest leads to count.
The CallRail survey asked owners with a poor agency experience what they wished the agency had done. The list reads like a discovery checklist that never happened.
xychart-beta
title "What owners wished their previous agency had done (CallRail, 600 US SMB owners, 2021)"
x-axis ["Customer service", "Vet new technology", "Marketing to sales workflow", "Refine value proposition", "Lead tracking", "Understand our customers"]
y-axis "Share of respondents (%)" 0 --> 60
bar [55, 51, 43, 40, 38, 29]
Source: CallRail SMB survey, reported by Search Engine Journal. Four of the six items ("refine the value proposition", "fix the marketing to sales workflow", "better lead tracking", "go beyond scope to understand our customers") are things a competent agency would do in its first two weeks, before any campaign is built.
What the first two weeks should cover
The first two weeks should produce a written picture of who buys, what they are worth, where they came from and what a bad one looks like. Nothing goes live until that exists. Here is what we think belongs in it, and roughly how long each part takes.
| Discovery task | What it answers | Who the agency needs | Time |
|---|---|---|---|
| Call with whoever answers the phone | What customers ask first, what they object to, which jobs get turned away | Receptionist, office manager or the owner | 45 minutes |
| Three to five customer interviews | Why they chose you, who else they considered, what nearly stopped them | Recent customers, ideally one who did not come back | 20 minutes each |
| Margin by service line | Which work is worth winning and which is worth declining | Accountant or the owner's management accounts | 1 hour |
| Source of the last 20 customers | Which channels already work, which the agency would be building from nothing | CRM, invoices, or the owner's memory | 1 hour |
| A description of a bad customer | Who to exclude from targeting and copy | The owner, in their own words | 30 minutes |
| Account and tracking audit | What can be measured today and what is being missed | Analytics, ads accounts, call logs | Half a day |
Six items, perhaps a week of agency time spread over two weeks of your calendar. None of it is expensive. All of it is cheaper than the first month of a campaign built on guesses.
The phone call is the one most agencies miss
The person who answers your phone knows more about your customers than any analytics dashboard. They hear the first question every caller asks, the price objection, the job the business keeps turning away and the postcode that keeps ringing. One 45-minute conversation with them will write better ad copy than a fortnight of keyword research.
It also exposes the leak that marketing cannot fix. CallRail's call data across 1.1 million small business conversations found that up to 85% of customers whose calls go unanswered will not call back.
If a third of calls ring out at lunchtime, the agency's job for month one is a call-handling fix, and an honest agency will say so even though it does not sell one.
Margin by service line changes the whole plan
Margin by service line tells the agency which leads to chase and which to filter out. Most small businesses have three or four lines of work with very different profit. A dental practice earns more from an implant case than from a hundred check-ups. A kitchen fitter makes money on full refits and loses it on single-unit replacements.
An agency that has never seen those numbers will optimise for volume, because volume is what the platforms report. Google Ads will happily tell you cost per lead. It cannot tell you cost per profitable lead unless somebody has defined one. Defining that starts with Google Ads conversion tracking that records which leads were real.
That is what the first two weeks are for, and it is why we ask for management accounts before we touch a paid media account.
Where the last 20 customers came from
The last 20 customers are a better data set than any market research report, because they are the people who actually paid you. Go through them one by one. Referral, repeat, Google, a sign on a van, a Facebook group, a trade directory. In our experience the split surprises the owner at least half the time.
The point is to find the channel that already works and the one the agency is about to build from zero. If 14 of 20 came from referrals and none from search, a search campaign is a bet, and the agency should price it and report it as one.
What a bad customer looks like
A bad customer is the one who takes the most time, pays the least and leaves the one-star review. Every owner can describe them in thirty seconds. Almost no agency asks. Yet the answer shapes targeting (exclude the postcodes, exclude the "cheap" keywords), copy (state the minimum job size) and landing pages (say what you do not do).
Negative targeting is where small budgets go furthest, and it only exists if someone asked the question.
The 6 discovery questions to expect
A good agency's first call is mostly questions, because it does not yet know what your problem is. Searchlab's guide to agency red flags describes the opposite: a first call that is "90% them talking" with three surface-level questions before the contract comes out.
A real agency's first call is a diagnostic. It's mostly questions, because they don't know yet what your problem actually is.
Searchlab, Marketing agency red flags
Here are the six we think any agency should ask before it quotes. They are the mirror image of the questions to ask a marketing agency yourself. If you get through a first call and fewer than four of these come up, expect a generic campaign.
- Who are your best customers, and what do they have in common? Not "target audience". The three or four named customers you would clone if you could, and what links them.
- What does a bad customer look like, and how do they find you now? This tells the agency what to exclude and often reveals a channel that is quietly attracting the wrong work.
- Which service makes you the most profit per job, and which makes the least? If they do not ask about margin, they will optimise for whatever is easiest to count.
- Where did your last 20 customers come from? Any agency that proposes a channel without knowing this is guessing with your money.
- What happens between a lead arriving and a sale? Who answers, how fast, what gets quoted, how many go cold. Most leaks are here, and most agencies never look.
- What have you tried before, and why do you think it failed? Owners usually know. The agency that repeats a failed campaign with a fresh logo has not asked.
Notice what is missing. None of these are about the agency's process, its awards or its "proven framework". Searchlab's test for a generic proposal is a good one: if you can find and replace your company name with another and the document still works, nobody looked.
What to do with an agency that has already invoiced
If you have already signed, ask for the discovery write-up. Whether that write-up should be a paid strategy phase for a small business depends on how much is unknown. If none exists, ask the agency to do it now, in the next two weeks, and to pause any spend it cannot justify against it. A good agency will be relieved. A bad one will talk about momentum.
Directory listings will not help you judge this in advance. The Semrush agency list for UK small business has 114 agencies with self-written descriptions and star ratings, and at least one promises Google rankings inside 90 days, which nobody can promise. The only reliable test is the first call. Count the questions.
One decision rule covers most of this. If an agency can tell you what your campaign will look like before it has spoken to the person who answers your phone, it is selling you the campaign it sold the last client.
If you want a discovery phase done properly before any budget is committed, start with a marketing and tracking audit or get in touch and we will tell you what we would need to see first.