Short answer: Tag your outbound links with UTM parameters, record the landing page and referrer with every captured lead, and stop relying on asking people how they heard about you. Attribution will never be perfect, but going from no idea to a rough idea is the step that changes decisions.
Most small businesses spend money on several channels and can name exactly none of them as the one that works. The result is a defensive budget: keep everything running because cutting something might cut the thing that matters. That is expensive, and it is fixable with a couple of afternoons of setup.
Why the How Did You Hear About Us field fails
It seems like the obvious solution, and it produces unreliable data for three reasons.
People genuinely do not remember. A visitor saw you in a local group in March, searched your name in June, clicked an ad, and reported Google. All three touches were real, and the answer credits the last one.
People pick the easy option. Google is the default answer whether or not it is true, because it is the first item in the list and requires no thought.
Most people skip it. Optional fields get skipped, and required ones make the form longer, which is the exact tax described in reduce form fields to get more leads.
Use it as a soft signal if you like. Do not build budget decisions on it.
Start with UTM parameters
A UTM parameter is a tag on the end of a link that tells your analytics where the click came from. You control it, which is what makes it reliable.
The three that matter:
- utm_source is where it came from. facebook, newsletter, partner-site.
- utm_medium is the type. social, email, cpc, referral.
- utm_campaign is which push it belongs to. spring-offer, van-livery, autumn-guide.
Tag every link you control: email footers, social bios, paid ads, sponsorships, QR codes on printed material. Do not tag internal links between your own pages, which corrupts your session data.
The discipline that actually matters is a naming convention. Pick lowercase, use hyphens, write it down, and stick to it. Six months of Facebook, facebook, and FB in the same report is worse than no report, because it looks authoritative while being wrong.
Record source data with the lead itself
Analytics tells you about sessions. It does not tell you which enquiry came from where unless you connect the two.
The fix is to capture context at the moment of lead capture: which page they were on, what referred them, and any campaign tag present. Attach it to the lead record so that when you look at a booked job in three months, the trail is still there.
Clerkzo leads carry their conversation with them into the leads inbox, so alongside contact details you have the page context and the visitor's own words. Combined with the pipeline states, New through to Won or Lost, and search across the whole inbox, you can answer the only question that matters: which source produced work, not just enquiries.
Notifications help here too. Sending lead data to a webhook lets you push source information straight into a spreadsheet or CRM, where it accumulates into something you can actually analyse.
Track to revenue, not to enquiries
The commonest attribution mistake is stopping at lead count. One channel produces forty enquiries and looks like the winner. Another produces eight. If those eight convert at a much higher rate and carry bigger jobs, the ranking inverts and you have been optimising the wrong number all year.
So track the whole chain: source, enquiry, qualified, won, and value. Even a manual spreadsheet reviewed monthly beats a sophisticated dashboard nobody opens. The pipeline in your leads inbox does most of this work if you keep it honest and actually close leads out rather than letting them linger.
Accept the limits
Attribution is directional, not forensic. Several things will always be murky:
Dark social. Someone shares your link in a private message and the click arrives with no referrer. It gets filed as direct. Nothing you can do.
Multi-touch reality. A customer sees you four times before enquiring. Last-click attribution credits one. First-click credits another. Neither is the truth, but consistently using one rule still lets you compare periods.
Offline influence. Word of mouth, a van seen at a job, a mention at a school gate. Some of your best sources are structurally untrackable. Ask about those in conversation rather than in a form.
The point of attribution is not certainty. It is being able to say, with reasonable confidence, that one channel is producing three times what another is, and reallocating accordingly.
A setup you can finish this week
- Write your UTM naming convention on one page and save it where the team can find it.
- Tag every external link you control. Social bios, email signatures, ad campaigns, printed QR codes.
- Make sure every captured lead records its landing page and referrer.
- Add a source column to your pipeline and keep it filled in.
- Once a month, list won jobs by source and value. Look for the obvious.
- Change one thing based on what you see. Then leave it long enough to matter.
That is it. No attribution model, no data warehouse, no consultant. Most businesses find something surprising within the first two months, usually that a channel they had been ignoring is quietly producing their best work.
Where chat fits in
An always-available chat widget improves attribution as a side effect, because it converts anonymous readers into identified leads on pages that would otherwise produce nothing measurable. A blog post that gets traffic but has no form is invisible in your lead data. Add a chat that captures enquiries and that same post starts showing up as a source, which sometimes reveals that your content is doing far more work than anyone believed. That is one of the quieter arguments in chatbot vs contact form.
Frequently asked questions
Do I need paid analytics tools?
No. UTM tags plus a lead record that stores page and referrer covers most of what a small business needs.
First touch or last touch?
Pick one and stay consistent. Consistency over time is worth more than picking the theoretically superior model.
How long before the data is useful?
Long enough to cover your normal sales cycle, and then some. Judging a channel on two weeks of data is guessing with extra steps.
Should I still ask how people heard about us?
Ask it in conversation, on the call, where you can follow up. Not as a required form field. See how to capture more leads from your website for the wider picture.