On this page
Why this is harder than it looks
An enquiry from a homeowner is rarely a single event. They see a van outside a neighbour's house, look you up a fortnight later, read the reviews, ask in a village Facebook group, then search your name and click whichever result is first — which might well be your own advert. Last-click reporting will attribute that job to a paid search click, and it is not wrong exactly, it is just describing the last two seconds of a three-week process.
You cannot fully solve that, and pretending otherwise is how people end up with dashboards they do not trust. What you can do is make each channel's contribution visible enough to make sensible decisions, and stop the specific failures that make the data actively misleading. In practice that is four installations and one habit.
It is worth naming the minimum viable version, because the perfect setup is what stops people building any setup at all. A trade firm that can answer three questions is already ahead of most of its competitors: how many enquiries arrived last month, which channel each of them came from, and how many became quoted work. Everything further — keyword-level call attribution, offline conversion import, consent-modelled reporting — refines those three answers rather than replacing them. Build in that order, and resist the tool that promises to do all of it before you have a reliable count of enquiries.
One structural warning before the mechanics. Tracking that lives entirely inside one advertising platform will always flatter that platform, because it can only see the conversions it was involved in. Keep the definitive record somewhere neutral — your job management system, or a spreadsheet if that is what you will actually maintain — and treat the platform's own reporting as a view of the account rather than as the truth about the business.
Call tracking that survives contact with reality
There are two mechanisms and most firms need both.
Static numbers means one dedicated number per channel: one on the vans, one on the Business Profile, one in the parish magazine, all forwarding to your real line. It is cheap, it never breaks, and it answers the big question of which channel produced the call. It cannot tell you which keyword did.
Dynamic number insertion swaps the number displayed on your website depending on how the visitor arrived, holding it for that session. That is what gives you keyword-level and campaign-level attribution on calls, which is the difference between knowing that search works and knowing which searches work.
The things that break it are consistent and worth checking on day one:
- The number rendered inside an image or a logo, which the script cannot swap and a mobile user cannot tap.
- Hard-coded tel: links in a header or footer template that the swap does not reach, so half your calls report as direct.
- Aggressive page caching serving a stale number to everyone.
- No duration threshold. Set one, at thirty or sixty seconds, or wrong numbers and hang-ups will be counted as leads and your bidding will chase them.
- Missed calls counted as nothing. They are not nothing. They are enquiries you paid for and lost, and they belong in the report as their own line.
On recording: if you record calls you need a lawful basis under UK data protection law and you must tell people it is happening, which in practice means an announcement before the call connects and a line in your privacy policy explaining why you record, how long you keep it and who can listen. Do not record card details. This is not onerous, but it is not optional either.
Keep your primary published number consistent wherever your business is listed. Channel-specific tracking numbers are for channels you control, not for replacing the number that appears in your directory listings and structured data, which is a consistency signal worth protecting.
Forms, WhatsApp and everything else
Every route in needs its own event, or the channels people actually prefer become invisible.
- Form submissions fire on a confirmed submission or a thank-you URL, never on the button click. Button-click tracking counts validation errors and double taps as leads. The quickest way to check yours is to submit the form with a required field deliberately left blank and see whether a conversion registers anyway. That test belongs in the handover of any new website.
- Click-to-call taps on mobile as a separate event. Not every tap becomes a connected call, so treat it as a strong signal rather than a lead, and reconcile against the phone system.
- WhatsApp and messaging link clicks, which for many trades now account for a real share of enquiries and are completely invisible otherwise.
- Email address clicks, for the same reason.
- Business Profile website clicks, separated by adding UTM parameters to the website link on the profile itself.
Then mark the ones that are genuine enquiries as key events in analytics and import them into the ad platform, rather than letting each tool count its own version. And deduplicate: a customer who fills in the form and then rings ten minutes later is one enquiry, not two. Without a rule for that, every well-behaved keen customer inflates your numbers.
Consent, and the bit you cannot skip
In the UK, storing or reading non-essential cookies and similar identifiers requires consent under the Privacy and Electronic Communications Regulations, and any personal data you then process falls under UK GDPR. Analytics and advertising tags are not essential cookies.
The state most trade websites are in is the worst of both worlds: a banner that appears, is dismissed, and does not actually gate anything. That gives you no compliance benefit and no data benefit. Either implement consent properly — tags held until a choice is made, a genuine reject option as easy as accept, and Google's consent mode signals wired in so modelling can fill some of the gap — or do not pretend. Doing it properly costs you some observed conversions and gives you numbers you can defend.
Two practical points that follow. Your conversion counts will drop when you implement consent correctly, and that drop is not a performance problem, so record the date and stop comparing across it. And your privacy policy needs to actually describe what you collect, including call recording and any CRM you push data into.
Closing the loop on what became a job
This is the step that separates accounts that optimise for enquiries from accounts that optimise for revenue, and hardly anyone does it.
When someone arrives from a Google ad, the click carries an identifier in the landing page URL. Capture it in a hidden field on your form, store it against the record in whatever you use to track jobs, and when that job is won, send it back to Google Ads as an offline conversion with the value of the work.
What changes is the objective. Instead of an algorithm being told that forty enquiries happened, it is told that six of them became jobs worth an average of several thousand pounds, and which searches produced those six. It will reallocate accordingly. The requirements are modest: a hidden field, a CRM field, and someone marking deals won consistently — usually the hardest part, which is why it belongs in the automation build rather than being left to willpower. Volume matters too; below roughly thirty won jobs a month the signal is thin, though even a monthly manual upload beats nothing.
The same identifier lets you do the simpler and equally valuable version: once a month, take last month's enquiries, mark each one quoted, won or rubbish, and see which campaigns produced the rubbish. That reconciliation finds problems no dashboard will show you.
A naming convention that lasts three years
Dull, and it will save you a fortnight of confusion later. Decide the format once and write it down.
| Thing | Convention | Example |
|---|---|---|
| Campaigns | Channel – intent – service – area | gads-emergency-drainage-coventry |
| UTM source | The platform, lowercase, always the same word | google, facebook, newsletter |
| UTM medium | A short fixed list you never add to casually | cpc, organic, email, print |
| Tracking numbers | One per channel, documented in a sheet | Van livery, profile, leaflet drop |
| Conversion actions | Type then detail | call-60s, form-quote, whatsapp-click |
The two rules that matter: lowercase everything, because analytics treats Google and google as separate sources, and never reuse a name for a different purpose. Historical data is only useful if the labels meant the same thing last year.
What tracking will never tell you
Being honest about the ceiling keeps you from over-interpreting.
It will not capture word of mouth, van sightings, the board on a job or the recommendation in a WhatsApp group, and for many established trade firms those are a large share of the work. A “how did you hear about us” question at quoting stage is unreliable — people genuinely do not remember — but it is still worth asking as a tiebreaker against the numbers.
It will not resolve the multi-touch problem. Last click will keep taking credit that belongs partly to a review someone read three weeks earlier, and your organic work will look smaller than it is. Where you can, judge channels by what happens to total enquiries when you change one of them, rather than by the attribution table alone.
And it will not fix response time, which is often where the actual money goes. The frequently cited research on responding within five minutes comes from a 2007 study of six companies, written up in Harvard Business Review in March 2011, and its headline figures are odds ratios rather than probabilities — they are routinely misquoted as though a lead is a hundred times more likely to convert. The underlying point survives the misquotation: a caller who does not get through rings the next firm. You do not need a study to justify answering quickly, and tracking is what tells you how often you are not.
Get these installed before you increase spend anywhere, not after. Every figure on the benchmarks page depends on this groundwork existing, it is the first thing we set up on any advertising engagement, and it is why our methodology starts with measurement rather than campaigns.
Rather have someone else do this?We run exactly this for trade and construction firms across the UK. Call and we will tell you what is realistic in your area, including when the answer is that it is not worth it.
Call 07443 392243WhatsAppCommon questions
Will call tracking numbers hurt my local SEO?
Not if your primary published number stays consistent across your website, your Business Profile and your directory listings. Use dedicated numbers for channels you control and keep dynamic swapping confined to website sessions. The risk comes from replacing your main number everywhere, not from tracking itself.
Do I legally have to announce that calls are recorded?
You need a lawful basis for recording under UK data protection law and you have to inform the people on the call, which in practice means an announcement before connection and a clear explanation in your privacy policy. Keep recordings only as long as you need them and do not record payment card details.
Is Google Analytics enough on its own?
For a trade business, no. Analytics sees website behaviour, and most trade enquiries end in a phone call it cannot observe. You need call tracking beside it, and ideally the won-job data fed back so the advertising optimises towards revenue rather than form fills.
My conversions dropped after I installed a cookie banner. Is that a problem?
It is expected. A properly implemented banner stops tags firing until someone consents, so you observe fewer conversions than before while the same number of real enquiries arrive. Note the date it changed, stop comparing across it, and judge performance on enquiries and jobs rather than on the tracked count alone.
How long does this take to set up?
A working setup for a single-location trade firm is typically a day or two: tag container, conversion events, call tracking, consent, and the naming convention documented. Offline conversion import takes longer because it depends on job outcomes being recorded consistently, which is a habit rather than an installation.