Start with a service the business can deliver profitably, then build a focused seed list. Use Keyword Planner for your service area's demand and costs, and other free sources for language and seasonality.
Pick a launch plan the budget can afford, and test every forecast against real enquiries after launch. The research is finished when the search terms report says it is, not when the spreadsheet looks tidy.
The example that runs through this guide is an example plumbing business in Melbourne's inner north. It's licensed for domestic plumbing and gas fitting. It has two vans and a phone staffed from 7am to 6pm on weekdays.
Urgent repairs and planned hot water replacements are the two services it wants more of. Its service area is about 25 minutes' drive from the depot. Nothing about the business comes from a client. And none of its numbers is a benchmark.
Define the service and build a seed list
Research begins with a commercial scope, not a tool. Which service has the margin, the capacity and a page or phone route ready for enquiries this month? For the plumber it is urgent repairs during staffed hours and hot water replacement quotes.
Bathroom renovations stay out of scope. The plumber can do that work but can't staff it until March. A service can be valuable and still be the wrong one to buy demand for right now.
Write the scope down before anyone opens Planner. It comes back later as the gates every keyword theme has to pass.
Service worth buying demand for: [the job, in the customer's words]
Licence or authorisation: [what covers the work, or none needed]
Area and capacity: [where the team works and how many jobs a week it can take]
Contribution per job: [what a completed job leaves after costs, or Unknown]
Enquiry route: [the page or phone line, and the hours it is answered]
Monthly ad budget: [the amount available to test with]
Not this month: [services that wait, and why]
The customers' own language comes next. The best evidence of how buyers describe the work sits in four places:
The last three months of enquiry notes.
The phrases customers use on the phone.
The site search log, if the website has one.
The search terms report from any earlier campaign.
"Hot water not working" appears in those notes more often than "hot water system repair". That difference is a keyword. Existing organic rankings describe the audience the website already reaches. That's useful context, not the paid opportunity.
Then build the seeds. A seed is a short service, problem or place phrase you give Planner to start from. It isn't a keyword you've decided to buy.
Service seeds | Problem seeds | Location seeds | Commercial modifiers |
|---|---|---|---|
emergency plumber, hot water replacement | burst pipe, no hot water, blocked drain | brunswick, coburg, northern suburbs | quote, cost, near me, same day |
gas fitter, hot water installation | leaking tap, gas smell, water heater not heating | preston, thornbury, melbourne north | licensed, 24 hour, price |
plumber, hot water system | hot water leaking, pilot light out | inner north melbourne | replacement, repair, install |
The matrix produces seeds by combination. The point is a focused dozen or two, not every cell multiplied by every other cell. Brand terms, meaning the business's own name and any trading names, go on a separate list. That way they never mix with the service research.
Access and use Google Keyword Planner
Get access to the tool
Keyword Planner lives inside a Google Ads account. Google's Australian Keyword Planner page calls it a free resource. The page says you first need to complete the account setup and create a campaign. If the account uses Smart Mode, you also need to switch to Expert Mode.
Google's page on using Keyword Planner adds a billing step. You must enter your billing information to use basic features such as keyword ideas.
Creating a campaign isn't the same as running one. Nobody needs ads live to do the research. There's no universal workaround for the billing step worth promising. Complete the setup, then pause any campaign it created. Make sure nothing is spending by accident.
Choose the entry route
Planner has two doors and they answer different questions.
Discover new keywords expands a focused seed or a website into ideas. Starting with keywords, you enter the service terms. You can also add your own domain so Google tries to exclude unrelated ideas. Starting with a website, Google reads the site's content for related keywords. It doesn't use hyperlink content.
Ideas drawn from a competitor's website are ideas about its content. They aren't evidence of the keywords that business buys.
Get search volume and forecasts is the other door. Paste an existing list, or upload a CSV with a Keyword header. You can also upload a full plan template with campaign, location and ad group columns. You get the historical metrics and a forecast for that list.
Set the research market
The settings decide what the numbers mean. That's why they get recorded beside every figure. These are the controls to check before reading anything.
Research location: [the service area, or the labelled wider area]
Language: [English]
Network: [Google only, or Google and search partners]
Historical date range: [the months covered]
Currency: [the account currency, AUD]
Date pulled: [the date the figures were pulled]
Entry route and seeds: [Discover or volume and forecasts, and the seed list]
Google's page on refining keyword ideas lists the controls. Enter countries, regions or cities for the location. Use Nearby for ideas in surrounding areas, or remove every location to go global. Pick the language. Choose Google alone or Google and search partners for the network.
Set the date range above the graph. The range changes both the trend and the suggestions. Note the account currency and the capture date too, or a figure can't be compared later.
The plumber researches its 25 minute service area, not all of Melbourne and not Australia. A business in Geelong or Townsville researches Geelong or Townsville rather than borrowing a Melbourne number.
The research location is a lens on the data. It doesn't change a live campaign's targeting. Targeting is set separately in the campaign's location settings, as our guide to Google Ads location targeting explains.
In a small market, a suburb level view returns little. So widen the research deliberately to learn the language. Label the wider geography in the record, and never present its volume as local demand.
Refine the ideas
The same page lists the filters above the results table:
keyword text
average monthly searches
top of page bid ranges
competition
organic impression share
ad impression share
There's also an option to exclude keywords already in your account. The categories panel groups the ideas, with all categories selected by default. So you can untick the irrelevant ones.
A Planner filter trims the research view and nothing else. A negative keyword in a campaign stops ads serving. They're different objects doing different jobs. The exclusions you notice here are written down as negative candidates for whoever manages the campaign's negatives. They aren't applied as if they were live.
Check the brand and non brand themes separately. Keep the synonyms that carry the same job. Resist the service by every suburb grid. It multiplies rows without multiplying customers.
Our research covers 30 single service local plans from October 2023 to September 2026. Across them, about 1 in 5 Planner ideas survived manual relevance review. That is the normal shape of the work rather than a sign the seeds were wrong.
Save the plan
Tick the keywords worth keeping. Choose a match type from the Broad match dropdown. Use Add keywords to put them in a new or existing plan. On the forecast page, you can switch between Conversions and Clicks. Conversions needs conversion tracking in the account.
You can change the average daily budget, the bid strategy and the date range. You can also download or share the forecast. Because that forecast is easy to misread, it's worth defining.
Google's page on organising keywords covers how Planner organises new keywords. Planner will try to match them to ad groups in your top campaigns unless you untick that option. The page suggests having at least 25 ideas before organising.
Planner bases suggested match types on the ad group's current keywords. You can accept, reject or drag keywords between groups. The page says plainly that keywords marked In plan aren't active until you implement the plan.
A saved plan is a research artefact. Review its groupings before anyone builds from them. Keep it distinct from an enabled campaign.
Read volumes, bids and forecasts correctly
Every column in Planner is a different measurement over a different window. The arithmetic goes wrong the moment they are mixed.
Metric | What it describes | Basis | Not to be used as |
|---|---|---|---|
Average monthly searches | Historical searches for the keyword and its close variants | 12 month average by default, rounded, exact match only | Clicks you can buy or reach under a broader match type |
Competition | Number of advertisers that showed on the keyword relative to all keywords | Low, Medium or High | SEO difficulty or a cost estimate |
Top of page bid, low and high | What advertisers historically paid for top of page positions | Last 30 days, about the 20th and 80th percentiles | An expected CPC when averaged together |
Forecast clicks, cost and CPC | Estimated daily performance for the saved plan | About a week of data, with bid, budget and seasonality applied | A promise, or a figure that survives a changed plan |
Google's page on forecasts and historical metrics is the source for every row. Historical stats are only shown for exact matches. So changing a keyword's match type in the plan changes the forecast, not the history.
Treat the volume as a rough guide, not a count. The same page says it's rounded. Volumes also move with the seasons and current events.
Averaging the low and high top of page bids is common advice. Calling the result your expected CPC is where plans go wrong.
The midpoint of two percentiles of other people's bids isn't a forecast of your click cost. Forecast CPC is, for a labelled scenario. That figure gets compared with actual CPC later.
The forecast and the history disagree by design. A saved forecast shows its assumptions on the page. So record them with the numbers.
Google's page says forecasts account for bid, budget, seasonality and other factors, unlike historical metrics. Forecasts are based on one week of data averaged to daily figures. The page on using Keyword Planner says they refresh daily. Each refresh uses the last 7 to 10 days, adjusted for seasonality. That page also recommends planning weekly rather than monthly when markets are moving.
Two consequences follow. Overlapping rows are not independent demand. The page says that for a list of broad match keywords any overlap is taken into account. Planner tries to predict how traffic divides between overlapping terms. Google says this makes the forecasts less accurate.
Take the monthly searches for "hot water replacement", "hot water system replacement" and "replace hot water system". Adding them up and calling the total demand counts the same searches three times. The combined plan forecast is the figure for the combined plan. The historical counts are per row.
And searches aren't clicks. A keyword's 1,000 monthly searches don't offer 1,000 clicks to anyone, let alone one advertiser with one budget.
Missing data is the other half of reading Planner properly. The page on using Keyword Planner says keywords with very low search volume aren't discoverable or forecastable. They return an error asking for different keywords or URLs. Google's personalised advertising and data privacy policies often prevent results for sensitive categories such as health or finance.
A blank isn't proof that nobody searches. Keep the term in the plan with Unknown against its demand. Corroborate it with customer language and a deliberately wider research geography, labelled as such. Then let the launch produce the evidence.
Never type in a volume. Never substitute a national figure without relabelling it. A Planner result also isn't a policy eligibility check or legal approval for advertising the service. That's a separate review.
Check seasonality and customer language
Each free source answers one question, and none of them answers Planner's.
Google Trends supplies relative interest, not volume. Google's Trends data FAQ says the data is a sample of real searches. It's scaled from 0 to 100. The scale is based on a topic's proportion to all searches in the chosen geography and time range. So 100 marks the highest point in your comparison, not 100 searches.
Different regions showing the same interest do not always have the same total search volume. Low volume terms appear as 0. Trends should be one data point among others.
Google's page on comparing terms explains that a term matches the words you enter in your chosen language. A topic groups terms that share a concept in any language. Trends doesn't return results for misspellings, spelling variations, synonyms or singular and plural versions. So the choice between term and topic changes the line.
Compare like with like:
the same geography
the same time window
the same search type
a consistent term or topic selection
Check several years rather than one before calling a bump seasonal.
For the plumber, compare five years of "hot water system" interest in Victoria with its own booking lead time. That comparison answers one question. Should the planned work campaign be built before the cold months rather than during them?
The graph belongs in the record only if it comes from a dated export. A seasonal chart drawn from a hunch is fiction.
Autocomplete and People Also Ask supply phrasing and objections. Google's page on Autocomplete says the predictions reflect real searches. They're shaped by:
the language of the query
the location it comes from
trending interest
past searches, for signed in users
Policies filter some predictions out. The same page says autocomplete doesn't simply show the most common queries. That means the order of the suggestions isn't a ranking of demand. Record "is it cheaper to repair or replace a hot water system" as a question for the ad copy, the landing page, the content plan or further research.
Don't buy it as a keyword just because Google suggested it.
Competitor ads are observations, not intelligence. Google launched the Ads Transparency Center in March 2023 as a searchable hub of ads from verified advertisers. You can search by advertiser and by region across Search, YouTube and Display. You can also see the last date an ad ran and its format.
The hub shows the creative. It doesn't show:
the keywords
the bids
the spend
the lead quality
the profit behind it
A competitor running an ad proves only that the competitor is running an ad.
Don't click live competitor ads to inspect their destinations. That spends their money for nothing and tells you little. Visit the site directly and note the offers and proof they lead with. Write your own claims from your own evidence. That's where our guide to writing relevant Google Ads takes over.
Prioritise keywords for a focused launch
Every theme faces two tests: four gates it must pass, then five scores that rank whatever survives. Buyer intent is one of the scores. It uses the five groups in our guide to high intent vs low intent keywords:
immediate action
provider selection
commercial investigation
problem aware
learning
The groups are a way of reading a search, not a value ranking.
The four gates come straight from the commercial scope: service, delivery, destination and budget. A theme that fails any one of them is out of the launch. No score rescues it. A 5 for location fit cannot save a job seeker query in a customer campaign. A renovation theme the business cannot staff until March fails the delivery gate, however profitable the work.
Themes that pass the gates are ranked on five criteria. Each is scored 1, 3 or 5 against the anchors in the table. Unknown is allowed wherever the evidence is missing.
Criterion | Score 1 | Score 3 | Score 5 |
|---|---|---|---|
Buyer intent | Learning, with no sign of hiring | Commercial investigation, or problem aware with the need unclear | Immediate action or provider selection |
Service relevance | Adjacent or excluded work | The service with a scope question attached | Exactly the work the business wants more of |
Likely contribution | Low value jobs or unknown margin | Moderate contribution after job costs | High contribution with a known close rate |
Location fit | Mostly outside the service area | Partly inside, or the job location is unclear | Inside the area the crew can reach |
CPC affordability | Forecast CPC above the most you can pay | Close to that ceiling, needs a tighter plan | Comfortably below the ceiling |
Buyer intent scores how close the searcher is to contacting a provider, not how valuable the search is. A problem aware search like "hot water leaking" can earn a 5 when the person plainly wants it fixed today.
The scores are your judgement, written down so the team can argue about it. They aren't a prediction of conversion rates. There's no pass mark that suits every business. A promising theme with Unknown against its cost is Test later or Further research. It's neither a launch nor an exclusion.
The four decisions are Launch, Test later, Content candidate and Negative candidate. The gap between the last two matters. A useful topic deferred from paid launch is not automatically a negative keyword.
"Is it cheaper to repair or replace a hot water system" is a content candidate and a landing page paragraph. It only becomes a negative if the search terms report shows it eating budget without producing enquiries.
Example query | Gates and scores | Decision | Reason |
|---|---|---|---|
emergency plumber brunswick | Passes all four. Intent 5, relevance 5, contribution 3, location 5, cost 3 | Launch | Urgent work in the core area, the phone is staffed, and the ceiling covers the forecast CPC |
hot water replacement quote | Passes all four. Intent 5, relevance 5, contribution 5, location 3, cost 5 | Launch | Planned work with high contribution, a quote page and a callback route for after hours |
plumber melbourne | Passes all four. Intent 5, relevance 3, contribution Unknown, location 1, cost 3 | Test later | Ready to hire, but too general and mostly outside the area. Test the phrase version with a tight location setting |
plumber jobs melbourne | Fails the service gate | Negative candidate | Employment, not a customer. Recorded for whoever manages negatives, not scored |
The three contrasts below show the same method in other businesses without building their campaigns.
Business | Launch theme | Deferred or excluded | Why |
|---|---|---|---|
Accountant | small business tax accountant plus the suburb | "how to do my own BAS" | A content candidate for the firm's blog, not a client enquiry this month |
Family law firm | family lawyer plus the city, consultation terms | "legal aid family law" | Fails the service gate for a private firm that doesn't take legal aid work, so it is a negative candidate |
Web design agency | website redesign for small business | "free website builder" | Different buyer and different budget. Excluded from paid, kept as an objection for the sales page |
There is no requirement for 20 to 50 keywords or two to four ad groups. Start with enough distinct, commercially useful themes to represent the affordable scope. Do not clone every variant into every match type. Stop when every shortlisted theme has a justified decision, a destination and an owner.
Two decisions follow the shortlist. Our guide to Google Ads match types covers how far each theme should reach. Our guide to Google Ads account structure covers how the themes become campaigns and ad groups.
Check whether the plan fits your budget
Keyword research and budget planning are one conversation. The arithmetic is simple enough to show in full.
Change one assumption and the plan changes shape. At $20 a click with the same 7% click to lead rate, $3,000 buys 150 clicks. That's 10.5 expected leads at about $286 each. The half lead is a long run expectation rather than half a person.
At $15 a click but a 4% rate, the budget buys the same 200 clicks. That's eight expected leads at $375 each.
None of these inputs is a Planner output or an Australian benchmark. They're transparent assumptions. The sensible plan is written at the pessimistic end as well as the hopeful one. For what clicks cost in the market, see our guide to how much Google Ads costs.
Then cap the whole thing by demand. Say the plan's forecast shows 120 clicks a month for the saved keywords in the saved area. In that case, a $3,000 budget divided by $15 doesn't become 200 clicks. It becomes unspent money and a decision about widening the theme or the area deliberately.
Whether $214, $286 or $375 is a cost the business can carry is a contribution question. Our guide to what a good cost per lead looks like works that out. It uses the profit on a job and the close rate.
Save a plan the campaign builder can use
The person who builds the campaign is often not the person who did the research. Six weeks later the researcher cannot remember why "plumber melbourne" was deferred either. Record the decision, not just the list.
Research date and source: [date, Planner route and any other source]
Market, language, network, date range and currency: [research location, English, network, months covered, AUD]
Service theme and customer need: [theme and the job behind it]
Demand and cost evidence or Unknown: [monthly searches, bid range, forecast CPC, or Unknown]
Gate checks and priority reasoning: [which gates passed and the scores]
Launch, test later, content or negative candidate: [the decision]
Proposed match type and destination: [match type and the page or phone route]
Review owner and next evidence needed: [person and what would change the decision]
Export or save the shortlist with that context, including Unknown where it applies.
Use the Share button on the forecast page only with the people meant to review it. Or send the export with the same context attached. Saving or sharing research isn't approval to launch a campaign.
Keep the forecast export as a dated baseline. It's what the launch gets measured against. Map each launch theme to a destination before the handoff so the builder is not inventing pages. Our guide to setting up Google Ads for a small business takes the plan from there.
Review the plan before and after launch
A first pass can be timeboxed. Here is the hour.
The hour is a planning aid. It isn't a promise that a regulated or complex campaign can be researched in sixty minutes. A law firm, a clinic or a multi region business will need more passes. They'll also need a compliance review that this guide doesn't provide.
What the hour does is produce a shortlist with decisions attached instead of a spreadsheet with volumes attached.
After launch, the research moves into the account. Google's search terms report shows the searches that actually triggered the ads. Some terms are omitted for privacy where they lack enough activity. Because of that, the visible rows are evidence, not the whole population.
Read it against lead quality. Revisit the Test later and Content candidate decisions with real numbers. The report steps for exclusions are in our guide to Google Ads negative keywords.
Compare the dated forecast export with actual spend, clicks, CPC and outcomes. Keep the market and period comparable. Count only outcomes that have had time to mature. Record every budget, bid, destination or goal change before anyone discusses forecast error. Rule out the changes nobody wrote down before blaming the tool.
We've compared 20 launches from October 2023 to September 2026 with their saved forecasts. In those launches, first month actual CPC landed no more than 40% above the forecast CPC. That is why the forecast is a baseline and not a budget. Our guide to the first 30 days of Google Ads sets the review schedule.
Commercial scope
The launch service has margin, capacity and a staffed response route this month
Settings recorded
Location, language, network, date range, currency and capture date sit beside every figure
Metrics read correctly
No averaged bid ranges, no summed duplicate rows, no searches treated as clicks
Gates before scores
Every launch theme passed the service, delivery, destination and budget gates
Decisions recorded
Each theme is Launch, Test later, Content candidate or Negative candidate, with a reason and an owner
Baseline saved
The forecast export is dated and the first review date is in the diary
What we recommend at Elev8d
We'd start with the scope, every time. A plan built on the wrong service can't be rescued by better keywords. So the commercial scope gets written before anyone opens Keyword Planner.
If the account has run before, we'd mine its search terms report next. Searches that already cost real money beat any estimate. They also come in the customers' own words. Planner comes after that, to fill the gaps and check costs.
Then we'd launch small and treat the first month as part of the research. We'd never budget from the top of page bid range. We'd take the plan's forecast CPC as a labelled starting assumption. Once the first month's clicks are in, we'd replace it with real CPC.
We'd change the order in two cases:
The budget can't test even one theme properly. Fix that before researching more keywords. Our guide to the minimum budget for Google Ads shows what a small budget can and can't answer.
The service is regulated, such as legal, health or financial work. We'd get the advertising rules checked before the keyword list is written, not after. Those rules can take whole themes off the table.
We'll check your keywords and search terms against the services you want to sell. Then we'll say what we'd launch, test later or cut.
FAQ: questions that remain
Do I need a paid keyword tool?
Not for a first paid Search plan for a local or single service business. This guide's research runs on Planner, Trends, Autocomplete and People Also Ask. It also uses the Ads Transparency Center and the business's own enquiry records.
Paid tools earn their fee at scale, across many markets or for organic work. They don't fix a plan built on the wrong service.
How many keywords should I launch with?
Fewer than you might think. The plumber launches two themes and about a dozen keywords. That's what its budget can test properly in a month.
If you can't explain why a keyword is in the plan, it isn't ready for the plan.
Next steps
Write the scope before opening Planner
The scope is one service, one area and one response route. It also includes the contribution a completed job produces. If any of those is unknown, that's the research task, not the keyword list.
Run the hour and record the decisions
Seeds, Planner with the settings written down, the free checks, then the four gates and the scores. Every theme leaves the hour as Launch, Test later, Content candidate or Negative candidate, with a reason.
Save the baseline and book the review
Export the forecast. Map each launch theme to its page or phone route. Then hand the plan to whoever builds the campaign, with the first review date attached. Would you rather have the research, build and review run by people who do it every week? See how our Google Ads team works.
Sources
Google, Alejandro Borgia, 29 March 2023. What the hub shows about verified advertisers and their ads, by advertiser, region, last date shown and format.
