Launch, discover, control: the first 90 days on Google Ads

Most trade businesses judge Google Ads the way they would judge a new apprentice: by how the first few weeks go. That is understandable, and it is usually the wrong test. In SoudCoh's research sample of Australian plumbers, electricians, cleaners, HVAC businesses, pest controllers, roofers, removalists and other home-service businesses, the first month is the least representative. It is usually cheaper than the month that follows, and it runs on a keyword list written before anyone knew how customers search.

The owner's question is simple: is this working, and what should we do next? The answer depends on which of three jobs the account is doing. In the first weeks it is launching: getting the structure, settings, tracking and phone answering right, so that real enquiries arrive and are counted. From about the second month it is discovering: learning from the search-terms report which words customers actually use. From then on it is controlling: promoting the searches that pay, removing the ones that never will, and putting budget where enquiries and jobs come from. Running through all three is a fourth job most accounts never finish — tracking each keyword back to revenue, so decisions are made on jobs won rather than on clicks or cheap leads.

This paper follows those jobs in order: what the reported data shows, what it means, and what to do about it — whether you run your own account, employ someone to run it, or want to check an agency's work. The method is the same for an owner-operator with one van and a multi-site brand with a call centre; what changes is how many ad groups there are to review.

A trade business has three facts that every campaign should be built around: it does a defined set of jobs, it serves a defined area, and it earns its money when the phone is answered. The build should mirror them. In practice that means a services campaign with one ad group per job — blocked drains, hot water, switchboard upgrades, end-of-lease cleans, termite inspections — each with its own keywords, ads and landing page, so the search, the ad and the page say the same thing. Someone searching for a hot water replacement should see an ad about hot water replacement and land on a page about it, not on a home page that lists eleven services.

SoudCoh builds launch with a median of 18 keywords per ad group and three responsive search ads in every ad group, each written for that one service. Every keyword is staged as an exact-and-phrase pair, so there is effectively no broad match at launch. Campaigns are created paused — 97% of the campaigns SoudCoh's campaign builder creates — and switched on only after a person has read the ad groups, keywords, copy and landing pages against what the business actually does. None of this needs special software: a list of services, a spreadsheet and an afternoon of careful reading will do.

The settings matter as much as the keywords. When a Search campaign is created in the Google Ads interface, the defaults include search partner sites, the Display Network, and location targeting that reaches people interested in your area as well as people in it. For a plumber who works a set of suburbs, each default spends money on people who cannot become customers: someone reading a news site, someone on a partner site that rarely produces enquiries, someone interstate looking up a holiday rental near your suburb. SoudCoh's campaign builder switches search partners and Display expansion off on every Search campaign, and targets people physically in the service area. The companion paper, From keyword to revenue, shows what the Display setting costs when it is left on.

What the build should not be is big for the sake of it. A comprehensive build switches on hundreds or thousands of keywords, but for the median advertiser only 29% of switched-on keywords have triggered even one impression, and the top tenth of keywords by spend carries 53% of all keyword spend. Google matches each search to the closest keyword, so near-duplicates sit silent — harmless, but it makes keyword count a vanity metric. Typically, 17% of live ad groups received no impressions over a recent month: niche services and suburbs waiting for a rare search.

What makes an account work is whether searches, ads and pages line up, and whether someone reads the search-terms report every week. Be wary of any pitch that sells "more keywords" as the fix.

Search advertising captures demand that already exists. Someone with a leaking hot water system is going to call somebody today, and a new account that appears for that search should hear about it quickly. In the reported data, it does.

That makes the first week a useful diagnostic window. If a new account has spent money for several days and recorded no enquiry, the cause is far more often mechanical than strategic: a call-tracking number that does not ring through, a form whose confirmation never fires the conversion tag, a phone that goes to voicemail, or a conversion action that was never switched on. Check those before touching a single keyword or bid.

The second pattern surprises almost everyone: the first month of a new account is usually its cheapest. For the median new advertiser, cost per enquiry in days 31 to 60 was 31% higher than in days 1 to 30, and 68% of new advertisers paid more per enquiry in their second month than in their first. We explain the pattern in more detail in why Google Ads month two costs more.

Our reading is that a fresh account serves the most urgent, ready-to-book searches first, before the bidding has widened its reach. After a few weeks it starts buying harder searches — more research, more comparison, more people not ready yet — and cost per enquiry rises towards its true level. The month-two rise is the end of a honeymoon, not a sign that something has broken.

The launch is also when the business side has to be ready. Most trade enquiries come from phones: mobile clicks delivered 76% of enquiries across the sample, and 58% of mobile enquiries were phone calls. Demand does not keep office hours either: about 26% of calls from ads arrive before 7am, after 6pm or on a weekend. Weekends pay their way: for advertisers that run ads all week, a weekend enquiry costs only 4% more than a weekday one.

So the schedule, the call roster and the landing page need to fit that reality before the first dollar is spent: a tap-to-call button that works, a form that fits on one screen, someone answering in every hour the ads run, and copy that promises only response times the business can keep. If nobody can answer at night, schedule the ads to match — knowing that this misses real demand rather than saving waste.

Every keyword list is a guess, written before launch in the way the trade talks about its work: "hot water system replacement", "switchboard upgrade", "end of lease cleaning". Customers do not talk like that. They describe symptoms ("no hot water", "power keeps tripping", "water under the sink"), they add a place or an urgency, they misspell, and they phrase things once and never again. SoudCoh's research sample shows how large that gap is.

For the median trade advertiser, 59% of enquiries came from searches that were not on the keyword list — even allowing for plurals and word order. Those discovered searches took 64% of search-term spend. Discovery is not a side budget; it is the larger part of the account, and a launch keyword list is a hypothesis, not a map of demand. Trade enquiries come from searches you never chose looks at the pattern in more detail.

The share holds across trades — from 52% of enquiries for the median cleaning business to 64% in plumbing and 63% across other trades and home services. Plumbing searches are especially varied because customers describe what they see — a leak, a smell, a noise — rather than the job that will fix it.

Discovered demand costs a little more. An enquiry from a search that was not on the list cost 1.19× as much as one from a search worded like an existing keyword. That premium is the price of learning, and it is modest. The mistake is to see it and switch discovery off; the better response is to harvest it — promote the repeat winners to keywords of their own and keep the net open for the next ones.

Much of that demand is rare. For the median advertiser, 54% of enquiries came from searches that received two clicks or fewer in total, and 33% came from searches that appeared in only one calendar month. Trade demand is episodic: a burst pipe, a failed inspection, a lease ending, a storm. No keyword tool would suggest most of those phrases. They can only be reached by keywords with room to match, which is why an exact-only build leaves customers on the table.

Nor does discovery finish. From the third month of running onwards, 25% of each month's enquiries still came from searches that had not appeared in any earlier month. Search-term review is a standing job, not a launch task, and the keyword list should keep growing from evidence.

Some kinds of search deserve a closer look because they behave differently from the rest. "Near me" is how many customers ask for a tradie without naming a suburb. Those searches produced 17% of enquiries for the median trade advertiser and converted clicks to enquiries at 1.3× the rate of other searches — the customer has usually decided to hire and is choosing who. The search says nothing about where the customer is, so it only works when location targeting reaches people physically in the service area and the page shows the area and a phone number in the first screen (why near-me searches convert better). Searches naming a suburb or local town are a smaller slice — 7% of enquiries — that converts slightly better but costs more per click, so the case for splitting them out is relevance and control, not cheaper leads.

Emergency searches are the story trades tell about paid search, and they are a thin slice in the data. Even for plumbers and electricians, emergency-worded searches ("emergency", "24/7", "urgent", "after hours") took 1.9% of search-term spend for the median advertiser, and produced 1.9% of paid-search enquiries. They convert well but cost more per click. Run a separate emergency ad group only if someone genuinely answers the phone at night, and bid it on its own numbers so it cannot quietly take the budget (emergency searches are a small share of trade demand).

Price researchers are not waste either. Searches containing "cost", "price", "how much" or "quote" converted less often but cost less per click, so an enquiry from them cost 0.95× what other enquiries cost. A blanket negative on "price" throws away enquiries at a normal cost; answer the price question in the ad and on the page instead. Length matters too: three-word searches — typically a service plus a place or a problem — converted at 1.24× the rate of searches of five or more words, which tend to carry questions and research. Long-tail room adds volume, but long searches need the tightest cleaning.

Match types set how much room each keyword has: exact is meant to catch the keyword and little else, phrase lets the customer add words around it, and broad lets Google decide what is related. In trade campaigns they behave differently from the way they are often described.

Exact is not exact. Google now treats exact match as "same meaning", so it also matches plurals, misspellings, reordering and rewrites it judges equivalent. Across the sample, 47% of the enquiries that exact-match keywords produced came from searches worded differently from the keyword. That is mostly helpful — customers rarely type the textbook phrase — but it means an exact-only account still needs a search-term review and a negative list. Exact match alone is not control (exact match is not exact).

Phrase does the work. Phrase match carries 69% of keyword spend across the sample, with exact at 30% and broad at 2%. Phrase keeps the service words fixed and lets the customer add the rest — the suburb, the urgency, the symptom. It is the balance most trade campaigns need. For the median advertiser, phrase match and its close variants produced 34% of enquiries.

Exact is not cheaper per enquiry. Phrase-match clicks cost 17% less than exact-match clicks in the same account — and are 18% less likely to convert, so the two cancel out. Within the same account, a phrase-match conversion cost 5% less than an exact-match one. The price of a click and the value of a click move together, so match types should be judged on cost per enquiry, never on cost per click.

Broad has to earn its place. Where broad match was used, its enquiries cost 1.09× as much as exact-match enquiries for the same advertiser — not materially more. Broad match learns from conversion data which searches are worth buying; if the Conversions column is full of phone-number taps, it learns to buy taps. It is a tool for well-tracked, well-cleaned campaigns, not a default for an account nobody reviews.

The search-terms report is where an account learns. It is also incomplete. Google withholds low-volume searches from the report for privacy, and in the typical trade Search campaign the report accounts for only 52% of the money spent. The same is true of results: the report names the search behind roughly 51% of a typical advertiser's enquiries. The rest is spent on, and won from, searches that are never named (the search-terms report hides half your spend).

That has two consequences. A negative list can only police the visible half, so control has to come from structure too: tight themes, the right match types and the right location settings. And the visible half is a sample — assume the hidden half behaves like the long tail you can see. Review the report every week, but judge a campaign on enquiries and cost per enquiry, not on how tidy the visible terms look.

A good review sorts every search term with meaningful spend into one of three outcomes. Negate it if the intent is clearly not a customer — a job seeker, a DIY tutorial, a free download, a retail purchase, a service the business genuinely does not offer — and it has never converted. Review it if it looks relevant but has not converted yet; these go to a person, because the fix is usually a bid, an ad or a landing page rather than a negative. Promote it if it has produced an enquiry and is not yet a keyword in its own right: add it as exact match in the right ad group when its cost per enquiry is at or below target, and as phrase match when it is above.

Most of what a review surfaces is not junk. Across the sample, about 75% of the spend a daily search-term review flags is on searches that look relevant but have not converted yet. Negating them on a cost threshold would shrink the account. Some are simply early: 7% of relevant-but-not-converting searches held back from negation went on to produce an enquiry in a later month. Many others are one-offs: 53% never triggered an ad again. Adding those as negatives does nothing except lengthen the list. The review effort belongs on patterns that recur.

Promotion is the step most advertisers skip. Only 11% of the searches that produced an enquiry through phrase or broad match were ever added as keywords in their own right, and even among searches that produced three or more enquiries, only 33% had been promoted. Leaving a winner to be caught by a broader keyword works, but it gives up control: you cannot bid a search up, write an ad in its exact words or protect it from a stray negative unless it is a keyword. A monthly promotion pass is cheap, and it compounds. Where a cluster of converting searches fits no existing ad group, it becomes an ad group of its own — which is how structure grows from evidence.

Cleaning up search terms is the part of paid search that is easiest to sell, and the data suggests it is usually oversold. For the median trade advertiser, only about 4.4% of reviewed search-term spend was flagged as clear-cut waste to negate. The level differs by trade — lowest in cleaning, highest in plumbing, where the words overlap more with DIY and retail searches — but in a well-structured account junk is a small leak, not a flood. An agency that promises big savings from cutting wasted spend alone is describing an account nobody has looked after.

Where the waste comes from matters more than how much there is. The usual targets of template negative lists — job seekers and DIY searches — were only 12% of the spend flagged as waste. The biggest single source, 44% of flagged spend, was searches for a different service, or for something unrelated that happens to share a word. Those look relevant enough to click and are specific to each business, so they need a list built from the owner's "we don't do that" list and checked against converting searches. Template lists are cheap insurance on day one; they do not solve the problem.

The good news is that control compounds. Each negative removes a pattern for good, so the junk that reaches the account shrinks as the list matures. Two months into daily cleaning, the share of reviewed spend that needed negating had fallen by 43% for the median advertiser. The first month is when cleaning matters most; after that it becomes maintenance. Once cleaning is regular, the negatives that can still be added safely cover just 0.6% of monthly spend for the median advertiser. By then the money is in cost per click, conversion rate and budget allocation. Cleaning is hygiene, not a growth strategy.

The bigger risk runs the other way. Negative lists accumulate, are rarely audited, and block more than anyone intended. In 52% of the sample, the advertiser's own negative keywords block at least one search that produced an enquiry during the year. For the median advertiser the cost is small — searches now blocked had produced 0.4% of the year's search-term enquiries — but it is concentrated: in the worst case in the sample, searches blocked by the advertiser's own list had produced 6% of its paid-search enquiries. It is the cheapest fix in paid search, because those enquiries were already proven (negative keywords that block customers).

Rules alone make it worse. Word-pattern rules ("price", "cheap", "buy", "review") catch a lot of junk and a small but real share of customers: 1.5% of the searches a keyword-rule cleaner flagged had converted in the very window it was judging, and 2% of searches flagged by simple waste rules had produced a genuine enquiry at some point in the year. A term can look like waste in a seven-day window and still be a customer over twelve months. That is why every automated negative in SoudCoh's process goes through a person and a simulation before it is added.

Two details catch out careful people. Negative keywords do not match close variants, plurals or synonyms the way positive keywords do, so a negative written for one spelling can leave another open. And competitor names are rarely the saving they appear to be: for 72% of advertisers that paid for competitor-name clicks, those searches produced at least one real enquiry. People search a known name, find it busy or out of area, and call the next credible option. Negate a rival's name only after real spend with no enquiries over a long window, using the full business name, and give competitor traffic you keep its own low-bid campaign.

A useful health check is proportion. The typical trade advertiser carries about one negative keyword for every 5 keywords it bids on. Most control comes from structure — tight ad groups, sensible match types, correct location settings — and negatives are the finishing layer. A negative list growing much faster than the keyword list is a sign the structure is too loose, and every negative added is one more thing that can block a customer by accident.

By the third month an account has enough history for the conversation to turn to growth, and the usual first request is more budget. The data suggests checking something else first. The typical trade Search campaign in the sample appears in only 21% of the searches it is eligible for, and for 92% of advertisers, more impressions are lost to ad rank than to budget.

Raising the budget does nothing for impressions lost to rank. Under automated bidding, "lost to rank" also includes auctions the algorithm chose not to chase because they looked too expensive for the conversion value, so the levers are relevance and a realistic cost-per-enquiry target, not a bigger daily budget. Relevance is made of three things Google scores on each keyword: expected click-through rate, ad relevance and landing-page experience.

Expected click-through rate is judged against every other advertiser on the same search, so in trades the ad has to beat the directories and the large brands on the first line — price, speed and proof beat slogans. Landing-page experience comes next: a slow page, a page that does not mention the service searched, or no phone number in the first screen. Ad relevance is rarely the problem when each ad group has its own ads. For the median trade advertiser, 54% of scored keyword spend goes to keywords with a Quality Score of 4 or less. Some high-intent, high-competition searches never score well and are still worth buying, but each should be a deliberate choice, reviewed monthly.

Attention should follow the money: the few keywords that carry most of the spend deserve close attention, while the long tail is managed in bulk and pruned quarterly. Device and day adjustments come last and should be small. An enquiry from a mobile click costs 9% more than one from a desktop click in the same account, a small premium on three-quarters of the volume, so a blanket mobile bid-down saves a little per enquiry and gives away a lot of enquiries. Fix the mobile page first.

Everything in this paper so far — the launch, the discovery, the control — rests on one assumption: that the Conversions column counts real enquiries. Often it does not. In 32% of the sample, taps on a phone number, email or WhatsApp link were counted as conversions, and where they were, the Conversions column overstated real enquiries by close to double (+92%). A search-term review run on those numbers promotes the searches that produce taps, not the ones that produce jobs.

An enquiry is not yet revenue, either. 77% of trade advertisers leave Google's default value of one dollar on every enquiry, so the "conversion value" and return-on-ad-spend columns simply count enquiries in dollars; only 9% tell Google Ads what a job was actually worth. Until jobs are traced back to keywords, budget is steered by enquiry volume alone.

That is the subject of the companion paper, From keyword to revenue: measuring what trade marketing is really worth, which sets out the measurement stack and a tracking audit any business can run itself. For the wider picture — demand, structure and revenue across home services — see Marketing for home services in Australia, and for published cost-per-enquiry figures by trade, the Google Ads benchmarks for Australian trades. How we build and run Search campaigns for trades is described on our Search campaigns page.

The first 90 days set the pattern for everything after. An account launched with control, given room to discover and reviewed on a fixed routine gets easier to run every month: the negative list matures, the winners become keywords, and the numbers start to mean what they say. If you would like a second opinion on a new or inherited account, talk to us.