How to measure marketing ROI when you've got twelve leads a month, not twelve thousand

The formula, and the benchmark that comes from nowhere

Marketing ROI is (revenue you can attribute to marketing minus what the marketing cost) divided by what it cost. Spend £1,000, attribute £5,000 in revenue, and that's 400% (or "5:1" if you prefer the ratio, which most articles do). Simple. Two things make it lie.

First, the cost. If "cost" is only the ad spend and not the agency fee, the software, the design, and the evenings you put in, the number is flattered before you start. Second, the revenue. A 5:1 ratio is poor if your gross margin is 20% (there's £1,000 of profit left on £5,000 of work before you've paid anyone) and excellent if it's 75%. A retainer that stays 24 months makes it look worse than it is; a one-off project that eats delivery time makes it look better. The honest version is: revenue times your gross margin, minus the fully loaded cost of getting it. That's the number to compare against how long you can wait to be paid back.

Now the benchmark. You'll have read that "a 5:1 return is good and 10:1 is exceptional". I traced it. Investopedia states it, cites one source, a marketing agency's blog post, and that post asserts it as "the middle of the bell curve" with no study, no sample, no industry, no methodology. The 2026 articles repeating it cite nothing at all. It's folklore quoting folklore. Two smaller ones while I'm at it: "the SBA says spend 7-8% of revenue on marketing" is a misattributed 2018 survey (the same source also mentions a 1.08% figure nobody quotes), and "email returns £36 for every £1" is a UK survey of 197 marketers from 2019, self-reported, not measured. None of this means marketing doesn't pay. It means you're being compared to nothing, and the only benchmark worth having is your own.

What a lead is worth (the maths owners skip)

Three numbers, from memory and your bank statements, five minutes: customers won in the last twelve months, enquiries in the last twelve months (rough), and the average first-year value of a new customer. Won divided by enquiries is your win rate. Win rate times first-year value is what a lead is worth. Say 14 wins from 60 enquiries at £3,500: a 23% win rate, so every enquiry is worth about £800 to you before it's even qualified, and you can sensibly afford a couple of hundred pounds to generate one. Now every other number on this page means something in pounds.

Rough beats nothing. Knowing your numbers is what lets you test: change one thing, watch one number, keep or bin it. That's the science bit. Marketing is art, science and maths combined into one, and the maths is the part most owners skip because nobody showed them it takes five minutes. (The free Weekly Marketing Scorecard has this calculator built in, and the Pipeline Calculator runs it forward into what a good month needs to look like.)

The six numbers, every Friday

Most small firms track the wrong things. In a 2026 survey of over 500 UK businesses, the metric most often named as "most valuable" was website traffic (30%). Sales got 16%, leads 13%, and return on investment just 6%. More than a third said they weren't satisfied with how they measure marketing, and 36% said they needed more training in analytics. So here's the whole training, in six lines:

1. Visits to the pages that matter. Your service and landing pages, not the whole site. The post about the office dog doesn't count.
2. Where they came from. The top three sources, by name: search, referral, LinkedIn, direct, email.
3. Action rate on those pages. Enquiries (or form starts, or calls) divided by visits. This is the number a page rewrite moves.
4. Leads, by source. Every real enquiry, with the answer to "how did you hear about us?" next to it.
5. Proposals or quotes sent. Proof that leads became conversations. Form-fills up but proposals flat is the small-business version of the trap the enterprise people call "MQLs up, SQLs flat".
6. Sales won, count and pounds.

Weekly AND monthly. At twelve leads a month, one quiet week is a third of your pipeline, and a monthly-only review finds out four weeks late. But the month is where you read the season: a quiet August in a finance-heavy client base is a calendar, not a crisis. The test of a quiet month is whether the pipeline is still being built (leads becoming proposals) and the traffic and the value you're putting out are still there. If they are, it still wins.

Vanity metrics: the numbers that make you feel busy

Eric Ries popularised the term in The Lean Startup back in 2011: the headline totals that look like progress and support no decision. Fifteen years on, the marketing versions are the same: followers, impressions, reach, total page views, "engagement", open rates. Not lying to you, just not talking to you.

Two tests catch every one of them. Does it only ever go up? A number that can't fall can't tell you anything, which is why follower counts are decoration. If it moved this week, would you change what you do? If the honest answer is "no, I'd just feel differently", it's a mood, not a metric. Leads, proposals and sales pass both tests. Impressions fail both. Email open rates fail the second and are faked by Apple's Mail Privacy Protection anyway: watch replies and clicks instead. Rankings for their own sake fail too; ranking first on Google doesn't guarantee the click any more, so measure enquiries from search, not positions.

The honest limit: nobody has a rigorous study proving social engagement drives B2B revenue for firms like yours, and nobody has one disproving it either. Treat engagement as a signal that something resonated, never as the scoreboard. The scoreboard is the chain: attention, enquiries, proposals, wins. Measure the hand-offs between them and every drop tells you what to fix: the offer, the page, the speed of your reply, the qualifying, the follow-up.

"If a number only ever goes up, or you'd change nothing when it moves, it's not a metric. It's a mood."

Attribution for humans

Set the tools up. Google Analytics and Search Console are free, and the more data you collect the more hypotheses you can test, which is the whole game. This site runs GA4 with consent mode, and here's the honest part the vendors skip: Google's own documentation says that when visitors decline the cookie prompt, Analytics is simply missing their data, and its "behavioural modelling" that estimates the gap only switches on above roughly 1,000 consented events a day. A small site never gets there. Your analytics is a sample, not a census.

Then ask. Every enquiry, every time: "how did you hear about us?" At this scale that one question beats the software, and there's evidence for why. In a controlled experiment by SparkToro in 2023, links shared through WhatsApp and similar private channels showed up in analytics as "direct" traffic with no source at all; Facebook Messenger lost the referral on three visits in four. A B2B firm that compared what customers said against what its attribution software said found a podcast credited with over half its revenue by customers and zero by the software (that one's a vendor case study, so take the size with salt, but the direction matches everything I've seen). Use analytics for what people did on your site. Use the question for where they came from. Treat both as evidence and neither as the whole truth. (One caveat on the folklore: "84% of conversions are dark social" is an old sharing statistic, not a measurement of anything you can count.)

What marketing reports hide

I'll say this carefully, because I produce reports for a living and the good ones are honest. Investopedia has the line: agencies "get around weak ROI figures by adding in more of the soft metrics". The mechanics are simple. A report leads with what went up. Impressions went up, reach went up, engagement went up, so page one is impressions, reach and engagement, and leads and sales are on page four, if they're there at all. It's the typical vanity stuff, and the test is the same as for your own numbers: if it can be shown to lead to enquiries and sales, fair enough, that's the job. If it's there to justify the next invoice, that's not the job.

• Ask for the report to lead with enquiries by source, proposals, wins, and cost per lead and per proposal, split by channel, not blended into one flattering total.
• Ask "what decision does this number support?" of anything on page one. Practitioner research on agency reporting says clients want three or four numbers and a recommendation, not a wall.
• Ask what isn't working. A report with no bad news is a brochure.

If you want the full anatomy of how this goes wrong, that's the marketing theatre page.

Measuring the stuff that compounds

Brand, content, the daily letters on this site: none of it measures cleanly at this scale, and anyone who says otherwise is selling a dashboard. The best evidence we have for the long game is the IPA Databank analysis by Binet and Field, built on close to a thousand campaign cases over thirty years, and its lesson is that brand building pays over years while activation pays over weeks, and you need both. What that means for a five-person firm is a set of leading indicators you watch without trying to prove: traffic in general, branded searches for your name, direct visits, replies to your emails, and the moment on a call when someone says "I read your thing about…". If people are aware of you, know exactly what you do, and think of you in a good light, there's no harm in that. Being remembered for what you do, well, is most of the job.

And borrow one honest idea from the finance textbooks: slowing a loss is a win too. If enquiries were falling and the marketing turned a drop into a plateau, the marketing worked, even though the graph didn't go up.

The fifteen-minute Friday

Card out, six numbers in, two honesty checks, one line on what you'll change next week. Then the pipeline review: what's coming, and whether the marketing is feeding it. Fifteen minutes, defended like a client meeting. Weekly catches the bad week; monthly reads the season; the lead-value maths turns it all into pounds. That's how you measure marketing when you're the whole marketing department, and it's more than most of the department does.

Sources: Investopedia and its cited source (the origin chain of the 5:1 rule) · Ridge Solutions' tracing of the SBA "7-8%" claim to a 2018 Web Strategies survey · DMA UK Marketer Email Tracker 2019 (197 marketers) · CMO Survey (Duke/Deloitte/AMA) sector spend figures · LOCALiQ UK State of Digital Marketing 2026 (500+ UK businesses; marketing-engaged sample) · Google Analytics Help (attribution models; consent mode and behavioural modelling thresholds) · SparkToro and Really Good Data dark-social experiment (2023) · Refine Labs self-reported attribution case (2023, vendor) · Eric Ries, The Lean Startup (2011) · Databox agency reporting research (2019) · Binet and Field, IPA Databank. Named as unverifiable and not used: the 5:1/10:1 benchmark, "84% of conversions are dark social", the "weekly reviews improve completion by 43%" claim, and any industry conversion benchmark presented as a target. No representative UK survey exists on how many small firms measure marketing at all, so none is claimed.