Is Your Google Ads Agency Doing a Good Job? 6 Signals to Read

Six signals for judging a Google Ads agency, what evidence may require the manager, and when weak performance does not mean the account is broken.

By Gavin Sevastian · Updated August 11, 2026 · 10 min read

Money leaves your account every month. A report arrives. The report looks fine. And you still cannot answer the one question you actually care about, which is whether any of it is working.

That is an uncomfortable place to sit and it is a very common one. Google Ads is easy to buy and hard to inspect, so most owners end up trusting a relationship instead of checking the work.

Below are six signals you can understand without becoming a Google Ads specialist. They are not all independent of the manager: ownership and reporting can be checked from records you control, while conversion definitions, search-term work and some account history may require access or an export from the agency. That dependency is part of what the evidence tells you.

One promise before we start. Some of what looks like bad performance is completely normal, and this page will tell you which. Firing a competent manager over a seasonal dip is an expensive mistake, and it happens a lot.

Start Here: What the Six Signals Can Tell You

Direct access changes how much you can verify

Your own login. Not the agency's, not a shared one. An email address you control, on the account, at ads.google.com.

Without it, you can still inspect invoices, reconcile reported leads against business records and ask for exports. You simply cannot independently verify every account-level claim. Everything gets easier once the business can see the account with its own eyes.

There is one check that needs nothing at all, and you can run it right now on your phone.

The six signals

The useful categories are practical control of the account, the definition of a conversion, agreement between reported leads and the enquiries the business actually saw, evidence of ongoing search-term review, reporting that reaches cost per lead, and transparent separation of management fees from Google's media cost.

They are deliberately foundational. A clean invoice does not prove good keyword judgment, and a recent negative-keyword export does not prove the campaign is profitable. Together they establish whether the relationship is visible enough to evaluate at all.

What a failed check does and does not mean

One failed check is a conversation, not a termination. Most of these fail through carelessness at signup or through neglect six months later, and both are fixable by the person already doing the work.

Several weak signals together are different. That is a pattern, and a pattern plus no clear answer when you ask about it is the point where switching becomes reasonable.

Check 1: Do You Actually Own the Ad Account?

The business should have direct access through an email address it controls, with enough authority to retain practical control if the agency relationship ends. A linked manager account is normal. The concern is not that the manager exists, but that the advertiser cannot independently access the client account or its history.

Google's ownership default depends on how the account was created: a manager that creates a client account becomes its owner manager, while linking an existing client account does not grant that ownership by default. Unlinking can also disrupt shared conversion tracking, remarketing or invoicing arrangements, so access problems should be resolved before a transition rather than during one.

The detailed ownership rules, Admin implications, exit risks and Customer ID questions belong to the Google Ads agency pricing and ownership guide. For this agency-quality review, the signal is simpler: your business can see and retain the account it paid to build.

Check 2: Is a Conversion a Real Lead, or a Button Tap?

Ask for the conversion actions, one at a time

A conversion count is the sum of whatever somebody decided to count. That is not cynicism, it is how the product works, and it means the number on your report is only as good as the definitions behind it.

So ask one question: what are the conversion actions on this account, and what does each one record?

A healthy lead generation account has a small number of actions, each matching something a human actually did. Called and spoke to someone. Submitted a form. Booked a time. The shapes worth questioning are page views counted as conversions, "clicked the phone button," newsletter signups, and several near duplicate actions all marked as primary and quietly added into one headline number.

Anyone doing the work can answer this in a few minutes, because it is a screen they already look at.

A tap on a phone number is not a phone call

This is the distinction that changes how you read every report you have ever been sent.

Row two is the one to sit with. If your reported leads are mostly taps, your lead count is a measure of intent to call, not of conversations. That is a large gap in a trade where the phone ringing out is a real and ordinary problem, and it is worth reading alongside why your Google Ads are not getting calls before you conclude the ads are the issue.

The default counting setting inflates web lead counts

There is a second setting doing quiet work. Google's conversion counting options are "Every conversion" and "One conversion," and Every is the default for website actions, in-app actions, Analytics transactions, clicks on a number on a mobile website, and imported conversions.

Google's own description of the alternative is the whole argument: One conversion "is a good choice if you're not interested in the number of sales, but instead whether or not a certain kind of lead was generated."

That is a lead generation business, described exactly. An account left on the default counts one person's three form fills as three leads. Nobody lied to you. Nobody configured it either, and that distinction is the honest version of this entire page.

One caution, because it gets repeated wrongly all over the internet. Switching to One does not de-duplicate by person. It de-duplicates per ad click. A repeat caller who arrives through separate clicks still books separate conversions, and no Google setting prevents that. Anyone who tells you the setting solves repeat callers has not read the documentation.

Check 3: Do the Reported Leads Match the Ones Your Office Can Name?

Why the reconciliation matters

This signal relies mainly on business records rather than technical knowledge, although the reported conversion count may still come from the manager. A completed period should show reported leads in the same general neighbourhood as calls, forms and enquiries the office can identify after normal conversion lag is considered. The comparison does not need to reconcile perfectly to reveal a material gap.

A gap in either direction is the finding

Use a month, not a week

Small numbers move a lot. A week with two leads instead of five feels like a collapse and is usually nothing, because at that volume ordinary variation looks identical to a real change.

Comparing a single week against another single week is the most common way an owner talks himself into a wrong conclusion, and it goes both directions. A good week convinces him things are fixed. A bad one convinces him to make a call he regrets. Use a full month, and use one that has finished.

Check 4: Is Anyone Reading the Search Terms Report?

What the report is, in plain terms

Every account has a list of keywords someone chose to target. Separately, it has a list of what people actually typed into Google before your ad showed up. Those two lists are never identical, and the second one is where the money is either being spent well or thrown away.

The second list is called the search terms report, and it is the closest thing an ad account has to a conscience.

Google's own documentation explains it with an example that needs no translation: "if you sell eyeglasses, and you noticed that the search term 'wine glasses' is triggering your ads, you might want to add 'wine' as a negative keyword."

Your version of "wine glasses" is easy to guess. Searches for jobs you do not take. Parts and products people want to buy and install themselves. Warranty questions for a manufacturer you are not. And the reliable one across every trade: people looking for work rather than looking to hire, typing your service name followed by the word jobs or salary or apprentice.

None of those people are ever going to call you, and every one of them can cost you a click.

Evidence should show a pattern of review

You do not need to audit every query yourself. The useful evidence is that recent search terms were reviewed and that the manager can explain examples of what was excluded and why. An export produced on request is weaker evidence than a recurring pattern of dated decisions, but response time alone does not prove whether the work happened.

Two reasons junk terms are not automatically a failure

Here is where a lot of contractor-facing advice goes wrong. Someone opens the report, sees an irrelevant search, and concludes the account is being neglected. Sometimes it is. Often the finding is just how the product works.

So the signal is not whether junk shows up in the report. Junk always shows up. The signal is whether anyone is doing anything about it, month after month, without being asked.

That reframe matters because it changes what you are actually buying. You are not paying someone to produce a perfect list of keywords once. You are paying for the ongoing act of reading what came in and blocking what should not have. If you want the wider view of where accounts leak money, the common Google Ads mistakes contractors make covers the rest of the pattern.

Check 5: Does the Report Show Cost Per Lead, or Only Impressions and Clicks?

Google's minimum is clicks, impressions and cost, and that is not a report

Google publishes an advertiser guide that third parties focused on small budgets are required to hand to their clients. Most contractors have never seen it.

That is the fairest and strongest version of this argument, and it is worth holding onto. The monthly PDF that frustrates you is probably not hiding anything. It is meeting a standard that was never designed to answer your question.

The chain you actually care about

What you care about runs in one direction: spend, then leads, then booked jobs, then revenue. Most reports stop at the first step and decorate it.

Be careful not to swing too far the other way. Impressions, clicks and click-through rate are legitimate diagnostic data. A good manager looks at them constantly, because they are how you tell a targeting problem from an ad copy problem from a budget problem. They are simply not a report to the person paying. They are the instrument panel, not the destination.

What to ask for if your report stops short

Cost per lead is arithmetic. Spend divided by leads. There is no tooling required and no sophistication involved, which is why its absence month after month is a choice rather than a limitation.

Ask for four things, one line per campaign: spend, lead count, cost per lead, and the definition of lead being used. That last one is not a technicality. After the previous section you already know that a lead can mean a conversation or a tap, and the number is uninterpretable until you know which.

Then ask the harder question: how many of those leads turned into quoted jobs?

Most managers cannot answer that from inside the ad account, because the ad account cannot see your calendar or your quotes. It needs your records. A manager who comes back and asks you for that number is doing the job properly, not passing the buck, and the relationship gets meaningfully better the moment you give it to him. If the answer changes what you want to spend, what a contractor should spend on Google Ads is the next thing to read.

What a Normal Cost Per Lead Looks Like in Your Trade

Published benchmarks by trade

You now have a cost per lead and probably no idea whether it is good. Here is published third-party data to hold it against.

Source: LocaliQ's home services search advertising benchmarks.

Look down the last column before you look up your own trade. Cost per lead runs from about $45 to about $228 inside one dataset, in one country, in the same year. That is a fivefold spread across trades that all fall under home services.

Which means anyone quoting you a single good cost per lead, across trades, is not looking at data. They are telling you about the accounts they happen to have seen.

A cost per lead is meaningless without the job value beside it

This is the habit worth building, and it cuts both ways.

A number that causes sticker shock in isolation can be trivially profitable against the job it produces. If a lead costs $200 and one in three becomes a job worth $9,000, nobody who has run the arithmetic is upset about the $200.

The reverse is just as real. A cheap-looking lead cost can be a disaster if the leads are the wrong kind of work: small callouts you make nothing on, tyre-kickers who wanted a price over the phone, jobs outside your area. Cheap leads for work you do not want are not a bargain.

So never quote one number without the other. Cost per lead against average job value, and eventually against your own close rate. And say the fair thing out loud, because it is true: a high cost per lead is not by itself evidence of a bad agency. It might be evidence of a competitive trade. For the trade-by-trade version of this comparison, cost per lead by trade goes deeper.

Local Services Ads are a different product with different numbers

One more number, because contractors compare these two constantly and the comparison is usually made badly.

SearchLight Digital's Local Services Ads benchmark reports an average cost per lead of $53 across $6.72M in LSA spend from 888 contractors and 126,650 leads in February 2026. Electrical came in at $39, HVAC at $51, plumbing at $57.

Set that against the table above and Local Services Ads look dramatically cheaper. Sometimes they are. But they are a different product, sold differently, with different lead quality and a different ceiling on volume, and comparing one number to the other without saying which is which is exactly how owners talk themselves into the wrong channel. Local Services Ads versus Google Ads is the honest side-by-side.

Check 6: Is the Fee Disclosed, and Is the Spend Figure Real?

Google's third-party policy requires a separately charged management fee to be disclosed in writing and on customer invoices. It also requires shared Google advertising cost data to show the exact amount Google charged, exclusive of the provider's fees. A blended total therefore obscures information Google expects the advertiser to receive.

That makes transparent billing a useful agency-quality signal, but it is only one. Tax documentation, customer-ID access and the consequences of a policy complaint are separate questions. The detailed rules, pricing structures and ownership implications are covered in what Google Ads agencies charge, where that material belongs.

What a Guarantee, a Badge, or a Ranking Claim Actually Proves

Three things get sold hard to contractors, and none of them mean what they are made to sound like. Handling them fairly is more useful than calling anyone a liar, so here is what each one actually proves.

Nobody can guarantee a position

If someone has promised you the top spot, you do not need my opinion on it. You can have Google's.

From the same advertiser guide linked in the reporting section above: it is not possible to guarantee a specific ad position, position is determined by an auction and changes dynamically with every new search, and third parties who guarantee a specific position "are misrepresenting how Google's advertising works."

That is Google using the word misrepresenting about a sales claim. Nothing this page adds is going to improve on it.

There is a Canadian layer too, and it is worth knowing about because most contractors assume this sort of thing is unregulated. Under paragraph 74.01(1)(b) of the Competition Act, you cannot make a claim about a product's performance or effectiveness "unless you can prove that the claim is based on an adequate and proper test," and the Competition Bureau is explicit that the test must have been done before the claim is made. Not after, when someone asks.

Two details make that relevant to you rather than to consumers. The Act reaches representations promoting "the supply or use of a product or any business interest," which is exactly what an agency selling management to a contractor is doing. And the courts assess the general impression a claim conveys, not only its literal meaning. A pitch built to leave you expecting the top spot is judged on that impression, not on the footnote.

Ads do not improve your organic rankings

This one is quick, and it kills a common upsell.

Google's own words, from that same guide: advertising on Google "has no impact on your organic or natural ranking in the search results; the two are completely separate."

So if anyone has told you that running ads will lift your position in the free results, or that pausing ads will drop it, that is not how the system is built. Both are worth doing for their own reasons and the case for each stands on its own, which is the actual comparison worth having in Google Ads versus SEO for contractors.

What the Google Partner badge measures, and what it does not

The badge is real, but its published criteria measure certification, activity and scale rather than the return on your account. It can establish that a provider participates in Google's programme. It cannot establish lead quality, booked-job economics or the judgment of the person assigned to you.

The pricing guide carries the current Partner and Premier Partner criteria and what each one does and does not imply. Do not confuse that agency badge with Google Guaranteed, a different programme attached to Local Services Ads and the advertiser's own screening. How to get Google Guaranteed covers that programme.

Things That Look Like Bad Performance and Usually Are Not

This is the section that keeps you from making an expensive mistake. Every item below is something owners routinely read as failure, and every one is usually the system working as designed.

Recent months can look worse before delayed conversions arrive

Start here, because it is the most common one by a distance.

This alone explains a large share of the panicked emails agencies get. The dashboard on the 3rd of the month is not a verdict on the month that just closed, and treating it as one produces a conversation that neither of you should be having. If you want the fuller picture of ramp-up timing, how long Google Ads takes to work covers the early months.

Seasonality moves demand by multiples, not percentages

Owners think in percentages. Search demand in the trades does not move in percentages.

WebFX's analysis of US home services search queries using Ahrefs found peak-to-valley swings of 594% on heating system repair, 609% on frozen pipe repair, 393% on emergency AC repair and 266% on AC repair. Not everything swings like that. Plumber near me moved 36%, roof repair near me 24%.

The instruction that falls out of this is simple and most people get it wrong. In a seasonal trade, month over month is usually the wrong comparison, because you are measuring the calendar rather than the work. Same month, year over year, is usually the right one.

If your leads dropped in a seasonally quiet month, that alone is not a finding. If they dropped against the same month last year, now you have something to talk about.

Roughly half of calls to home services businesses never reach a person

This one is uncomfortable and it needs saying plainly, because it is the single most common way an owner misdiagnoses his own situation.

Invoca, a call tracking vendor, published 2026 home services benchmarks reporting that 52% of callers to home services businesses speak with a person. That rises to 65% on calls lasting over 15 seconds and 73% over 30 seconds, and answer rates across sub-industries ranged from 32% to 74%. The same report found 55% of home services businesses do not ask the lead to buy or book the job.

No lecture attached. But whoever is spending the money is entitled to ask what happens to the calls after they ring, and a contractor who fires a competent manager over an unanswered-phone problem has paid a switching cost to keep the actual problem.

Four more that look wrong and are not

Row one is worth internalising because it generates so many false alarms. Google's budget pacing documentation sets the day and month limits explicitly, so a heavy Tuesday is not overspending, it is the system buying volume when volume is there.

Row two is the one to be patient with. A new or reactivated strategy, a strategy setting change, or campaigns, ad groups or keywords being added to or removed from a strategy can trigger Learning. Target changes do not reset prior learning, although large ones can create short-term volatility. If you want the mechanics, Google Ads bid strategy explains what is actually happening during that window.

What These Checks Cannot Tell You

Everything above verifies that the basics exist

Ownership. Measurement. Search terms. Reporting. Fees.

Each is binary or close to it, which is why a non-technical owner can understand the evidence. Some can be verified from the advertiser's own records; others still depend on account access or material the manager supplies. That boundary is the design of this page and also its limit.

Passing all six means the foundation is sound. It does not mean the account is good. Those are different claims, and conflating them is how owners end up satisfied with an account that is competently maintained and quietly mediocre.

The judgment calls are the part you cannot audit from outside

Consider three decisions that no checklist can settle.

Whether the right keywords were bought. Two search terms can look nearly identical and behave completely differently. One brings people ready to book, the other brings people pricing a job they intend to do themselves next spring. You cannot tell which is which by reading the words. You find out by spending money on both and watching what comes back, which takes months and costs real money to learn.

Whether the bidding approach fits the budget. Google permits some automated strategies with little or no campaign conversion history because its systems can use query-level and cross-campaign signals. Permission is not the same as a stable evaluation sample. Lower volume can make targets harder to derive and results harder to interpret, so the choice remains a judgment about the budget, the trade and the season.

Whether a change should be made this week or waited out. This is the hardest one and the least visible. A bad fortnight can be noise, a seasonal turn, or the start of a real problem, and those three look identical on a Tuesday. Acting too early resets calibration and destroys the evidence you needed. Acting too late costs a month.

None of those produces a checkbox. All of them need someone in the account regularly, with enough accounts behind them to know what a normal bad fortnight looks like.

Which is the honest reason this is full-time work

The checks on this page take an afternoon. The judgment takes recurring attention and enough comparable history to recognize when a weak fortnight is ordinary variation rather than a structural problem.

That is why an owner running crews may not want this job. It is different work from the one already filling the week. This is how I manage it for contractors, if you want a comparison point.

What to Do With What You Found

Fix ownership and measurement before you decide anything else

Whatever you concluded about your current arrangement, do these two things first. They are worth doing even if you intend to stay exactly where you are.

If the measurement is wrong, a new agency inherits polluted history and starts from a false picture. If your business does not hold Admin access, a transition can leave you locked out of the history your money paid to build. The account itself retains its campaign history when a manager is unlinked, which is why securing your own access before the move matters.

So ownership and measurement are not the first two checks by accident. They are what makes any later decision survivable, including the decision to change nothing. And if the real question turns out to be whether to run ads at all, are Google Ads worth it for contractors is the better place to answer it.

The response is evidence too

A direct conversation can resolve gaps caused by an old signup, a weak report template or work that was done but never explained. A manager who can provide the relevant evidence and correct a real weakness may be worth keeping. The concerning pattern is several foundational gaps combined with evasive or unsupported answers, not one imperfect report.

When the pattern is the finding

Account access and measurement should be made durable whether you stay or leave. A transition can disrupt shared tracking, remarketing or invoicing, while polluted conversion history follows a new manager into the next engagement. That makes an angry unlink a poor substitute for a planned handoff.

The final judgment is pattern-based. Practical control, meaningful conversions, business-record reconciliation, ongoing query review, outcome-oriented reporting and transparent billing are six different windows into the same relationship. Several of them going dark at once matters more than a single bad month, badge or dashboard score.