From keyword to revenue: measuring what trade marketing is really worth

Ask a trade business owner whether Google Ads is working and they will usually answer from the phone, not the dashboard: it has been busy, or it has not. Ask the dashboard and it will usually tell a better story. The difference is rarely the campaign. It is measurement that was set up once, often by someone else, and never checked against the job book.

The single most common gap is what Google Ads is told to count. A tap on a phone number, an email address or a WhatsApp link can be recorded as a "conversion". It tells you someone meant to get in touch, not that they did: many taps never connect, some are existing customers, and the same person can tap three times. In 32% of SoudCoh's research sample, those taps were counted as conversions.

Where taps were counted, they made up 45% of everything Google Ads reported as a conversion for the median advertiser, and the Conversions column overstated real enquiries — connected calls and submitted forms — by close to double (+92%). A campaign that reports nearly twice as many conversions as it produced real enquiries shows a cost per enquiry at roughly half the truth. Every decision made on that figure is made on the wrong number: the budget, the bid targets, the verdict on the agency. We look at the mechanics in how phone-tap conversions inflate trade leads.

There is a second, quieter cost. Automated bidding optimises to whatever sits in the primary Conversions column. When taps are there, Smart Bidding learns to find the people most likely to tap, which is not the same as the people most likely to book. The account gets better at producing the wrong thing, and the report says it is improving.

The arithmetic is worth doing on your own account. If taps make up 45% of the conversions in a report, the true cost per enquiry is about 1.8× the reported figure — a campaign that looks comfortably inside its target may be well outside it. Nothing about the campaign has changed; only the denominator has. That is why the first question in any review of a trade account is not "how are the campaigns going?" but "what is being counted?"

For plumbers and electricians the phone is the business. Calls arrive by two routes, measured by different machinery. A caller can tap the number in the ad itself, which Google records as a call from the ad, or read the website first and ring the number on the page, which is only recorded if the site shows a tracked number to paid visitors. Both are enquiries, and both come from the same budget. Most trade advertisers record calls in some form, but only 32% capture both routes.

The website route is the one most often missing, and it is the harder one to set up well. The usual method is a tracked number that is swapped onto the site only for visitors who arrived from an ad, so the call can be tied back to the click, the landing page and, once matched, the keyword. The business's main number stays unchanged everywhere else — on the Google Business Profile, directories, the van and the invoices — so nothing that customers already rely on breaks. Test it from a phone after every website change, because a swapped number that silently stops swapping looks exactly like a quiet week.

An account that records only calls from the ad misses the people who read three pages first — often the most considered buyers, sent by the pages working hardest. An account that records only website calls misses the quick tap-to-call from the ad. Either way, budget drifts away from whatever is not being counted. Track both, and count each caller once.

The balance between calls and forms depends more on the job than on the device. For the median advertiser, calls were 58% of mobile enquiries and 41% of desktop enquiries — but the trade matters more: plumbers and electricians get mostly calls on both devices, cleaners mostly forms on both. Urgent, hard-to-describe jobs such as a leak or a fault arrive by phone whatever the screen; planned, easy-to-describe jobs such as a bond clean or a carpet clean arrive as forms even from phones. Trades that live on calls need the number, call tracking and someone answering; trades that live on forms need a short form, an instant acknowledgement and a fast reply. Every site still needs both doors.

Once calls are recorded, three rules decide which ones count, and the defaults rarely fit a trade business.

Call length. Google counts a call from an ad as a conversion only if it lasts 60 seconds or more, unless someone changes the setting. About 49% of answered calls from ads last less than a minute. Short calls are a mix: wrong numbers and hang-ups, but also the quick "can you come Thursday?" that books a job. A fixed 60-second rule hides the second group from bidding and keeps the first out of reports — neither is ideal. Listen to a sample of short calls before deciding where the threshold should sit, and revisit it by service: a blocked-drain call is often short and decisive (short ad calls and the 60-second rule).

Repeat callers. About 19% of calls to a tracked number come from someone who has already called — chasing a quote, confirming a time, calling back after a missed call. They matter to the business, but they are not new enquiries. Count unique callers, otherwise cost per enquiry looks better than it is. Google's "one per click" counting stops a single ad click being counted twice, but a caller who rings back later can still count again; apply a unique-caller rule deliberately wherever your call tracking allows it.

After-hours calls. About 26% of calls from ads arrive before 7am, after 6pm or on a weekend. If the roster is set to office hours, those calls are paid for and then missed. Report missed and after-hours calls separately, because they are an operations problem no bid change can fix: either answer them — a call-answering service usually costs less than the ads that produced the call — or schedule the ads to the hours someone can pick up.

Return on ad spend is the number every owner wants, and for most trade advertisers it is not measuring what it claims. When a conversion action is set up, Google suggests a default value of one dollar per conversion unless someone changes it. 77% of trade advertisers leave that default on every enquiry. The "conversion value" column then simply counts enquiries in dollars, and the ROAS column divides that count by spend.

A one-dollar default is harmless until someone reads the ROAS column or switches to a value-based bid strategy. At that point Google optimises for the number of enquiries while everyone believes it is optimising for revenue. Only 9% of trade advertisers told Google Ads what a job was actually worth — and none did it through website or call tags. The only real values came from won jobs uploaded after the fact (why a default conversion value makes ROAS meaningless).

That is not a failure of effort; it is how trade work is priced. A website cannot know what a job is worth when the form is sent, because the price is set on site, days later. So real value has to come back from the job book: record each won job against the click or call that produced it, and upload it to Google Ads as an offline conversion. Until then there are two sound choices — leave value empty and bid on enquiries, or feed it real job values — and one mistake: reading a placeholder ROAS as if it were revenue.

Even with real values, trade revenue is lumpy. In the median business, the biggest tenth of won jobs carried 41% of recorded revenue. A switchboard upgrade, a full re-pipe or a leak that became a bathroom can swing a month's return more than any campaign change. Judge campaigns on rolling quarters, look at the median job as well as the total, and keep a separate eye on the services that produce the large tickets, because they are rare enough to be cut by mistake.

The practical response is to connect jobs to the ad group that produced them — even a monthly spreadsheet is enough to start — and to give the services that produce the large tickets their own campaigns and budgets, so they are not starved by cheaper, faster-converting work. Judge calls and forms alike by the jobs they win. Budget should move on revenue per ad dollar once the data is complete enough, never on a single month and never on cost per lead alone.

The Google Ads report shows one total per campaign. Inside a Search campaign, that total can include three very different audiences: people searching on Google, people searching on partner sites that use Google's results, and people reading something else entirely on the Display Network. In 21% of the sample, campaigns labelled "Search" were also serving ads on the Display Network.

The Display option is switched on by default when a Search campaign is created in the interface, and Google can shift delivery towards Display without any entry appearing in the change history. The campaign still says "Search", the report still shows one total, and the leads get worse. Where Display ran inside Search campaigns, it took 23% of their impressions but only 1.5% of the spend — inflating reach while barely touching the budget line anyone checks.

Cheap impressions flatter the averages, and cheap clicks flatter cost per click. What matters is enquiries. A Display click inside a Search campaign produced a real enquiry at about 4% of the rate of a Google Search click in the same campaigns. Display clicks come from people reading a news article, playing a game or checking the weather, and some are accidental taps. On a fixed daily budget, every dollar spent there is a dollar not spent on someone searching for a plumber (the Display Network inside trade Search campaigns).

Search partners are subtler. They are real searches, on other sites that use Google's results, and their clicks are cheaper. But for the median trade advertiser, search-partner clicks converted at only 16% of the Google Search rate, so each enquiry cost more, not less. A trade business pays for enquiries, not clicks. Check enquiries per dollar by network in your own account; if partners lose, switching them off is one tick-box. If you want Display at all, give it its own campaign with its own goal.

The common advice is to send every ad to a specific service page rather than the home page. Service pages earn their keep: they make ads more relevant, lift Quality Score and give Google and AI assistants something specific to cite. But a new address is not a conversion fix on its own; what turns a click into an enquiry is what the page shows. Put the phone number and a short form in the first screen, name the job and the suburb, and show proof.

The same goes for suburb pages. A per-suburb campaign can send each click to a page that names the suburb, but the page has to be real — local jobs, local photos and travel times, not a template with the suburb name swapped.

Across the trade websites SoudCoh has built, the components that recur are the ones that answer a homeowner's questions before they call: do you do this exact job, do you come to my suburb, and what happens after I press send? 80% of those sites explain what happens after someone enquires, and most show suburb coverage and give each service its own page. None of those needs a big budget; they need specific content.

Measurement matters here too. Most trade enquiries come from mobile clicks — 76% across the sample — so pages should be judged on a phone: a tap-to-call button that works, a form that fits on one screen, and a page that loads before the searcher gives up. Then measure each page on enquiries per click, split into calls and forms, and fix the first screen before building more pages. When comparing pages, compare like with like — the same service, the same kind of search — over enough clicks to mean something. We cover the page itself in what a trade website needs to turn clicks into calls and forms, and trade-specific builds on our pages for plumbers, electricians and cleaning businesses.

Some research now starts with a question to an AI assistant rather than a search. Measuring it is hard. On SoudCoh's own website, visits that analytics could attribute to AI assistants were 0.2% of sessions — well under one in a hundred — because most AI-influenced visits arrive with no referrer and are filed as "Direct". When someone copies a name from an AI answer into a browser, or taps a link inside an app that drops the referrer, nothing tells analytics where they came from. So analytics understates AI discovery badly.

What the assistants cite can be observed, and it follows a clear pattern. When ChatGPT-style assistants answer "who does marketing or websites for [trade]", they mostly cite agency pages built for that exact trade — not agency home pages or directories. The lesson carries over to any trade business: assistants answer the question they were asked, so a page that plainly states who it is for, what the work involves, where, and what it costs is more likely to be quoted than a general home page. If you want to be recommended for a service, publish a page for that service.

Assistants are also unstable. Ask one the same buyer question three times and only a minority of the websites it cites appear in all three answers (25% in SoudCoh's testing). There is no fixed "AI ranking" to win. Being on the short list — through a page that directly answers the question, a directory profile and independent mentions — matters more than being first in any one answer.

For measurement, that means three habits. Ask every new enquiry how they found you, and store the answer on the lead record. Give AI assistants their own channel in analytics, so the visits that can be identified are not lost in "Referral". And measure visibility by asking a fixed set of buyer questions several times, on more than one assistant, recording who is named and which pages are cited — never by asking once. We explain how assistants choose in how ChatGPT chooses which business to recommend.

The question on the lead form matters more than it looks. A free-text "How did you hear about us?" box collects answers that cannot be counted; a short list — Google search, Google Maps, an AI assistant such as ChatGPT, a friend or neighbour, a van or sign, social media, other — can be. Ask the same question on the phone and record the answer on the same field, so calls and forms are counted the same way. Over a few months, the gap between what customers say and what analytics recorded is the most practical estimate most businesses will get of how much of their demand starts outside a search box.

None of the fixes above is expensive. What they need is order. Measurement in a trade business works as a stack, and each layer depends on the one beneath it: bidding cannot optimise to revenue if the lead record does not carry the click, and the lead record is little use if the Conversions column is counting taps. The five layers below are the ones SoudCoh sets up for clients, and any business can build them for itself, whether it is one owner with a phone or a multi-site brand with a call centre and a CRM.

Layer one: clean conversions. Open the conversion actions and sort them into the three piles — enquiries, signals and revenue. Only enquiries belong in the primary column. Check campaign-level goals too, and redo the audit whenever the website, tag manager container or call-tracking provider changes, because each can quietly add or break an action.

Layer two: both call routes. Record calls from the ad through Google's call reporting, and calls from the website through a tracked number shown to paid visitors, with the click and landing page attached to the website call and passed back to Google Ads. Count each caller once, above a duration set for the business, and report missed and after-hours calls separately.

Layer three: one lead record. Every form, call and booking should be written to one record with its source captured at the moment it arrives: the click identifier Google adds to the page address, campaign tags, landing page, device and — once the click is matched — campaign, ad group, keyword and search term. A tag that captures the click on landing and keeps it for weeks means a visitor who returns days later to enquire is still attributed. The same source field should flow into the CRM or job system, so sales follow-up and reporting read the same thing. Existing Google conversion tags stay in place; the lead record adds to them.

Layer four: outcomes. Each lead is marked contacted, quoted, won or lost, with the value of the job when it is won. The system matters less than the habit: a job-management tool, a CRM or a spreadsheet updated monthly will do. Record how revenue arrives — entered as it happens or imported from a sheet, and where the won date comes from — because two businesses that record revenue differently cannot be compared on return on ad spend.

A note for multi-site brands. A business with several locations, brands or call centres needs one more rule: every lead carries the location or brand that handled it, and every location uses the same definitions of an enquiry, a quote and a win. Without that, one branch's "enquiry" is another branch's "booked job", and no comparison between sites — or between the campaigns that serve them — can be trusted.

Layer five: value back. Upload won jobs to Google Ads as an offline conversion against the original click or call, created as a secondary action at first so it informs reporting without changing bidding until enough months have accumulated. Send the same lifecycle events to analytics exactly once, so channel reports can be read in jobs rather than form fills, and keep analytics key events restricted to real enquiries.

The checklist below is the audit SoudCoh runs before judging any campaign, written so that a business owner or office manager can run it themselves. Work through it in order. The first half checks what Google Ads is counting; the second half checks whether enquiries can be traced to jobs. Anything that fails is worth fixing before any campaign, agency or channel is judged.

Once the audit passes, read the numbers differently. Judge campaigns over rolling quarters rather than single months, because a few large jobs swing any month. Look at the median job as well as total revenue. Compare services on revenue per ad dollar, not cost per lead. And keep a line in every report for what the dashboard cannot see: missed calls, repeat callers and enquiries whose source is unknown. If you use an agency, those are fair questions to ask of its monthly report (what a good agency should report every month).

A useful monthly measurement report fits on one page. It shows enquiries split into calls and forms, cost per enquiry by service, missed and after-hours calls, the share of enquiries whose source is unknown, jobs won and their value, and revenue per ad dollar over the latest rolling quarter alongside the median job. Each line answers a different question, and together they show where an enquiry leaks between the click and the job: not answered, not quoted, not won, or won but never recorded. The fix for each leak is different, and only one of them is a change to the ads.

This paper is the measurement half of a pair. Its companion, Launch, discover, control: the first 90 days on Google Ads, covers how a campaign should be built, how it discovers what customers search, and how to control it without cutting off customers. For the wider picture across home-service marketing, read Marketing for home services in Australia; for how we set up call and form tracking, see tracking and measurement; and for how enquiries become booked work, lead generation. If you would like us to run the audit on your account, get in touch.