The Optimisation Score Is A Sales Target, Not A Health Check

Most small businesses don’t meet their optimisation score in the interface. They meet it on the phone.

Someone from Google rings, friendly and genuinely helpful in tone, offering a free review of the account. Your score comes up early. It’s sitting at 68 per cent, and there’s a set of recommendations that would take it to 94.

The suggestions sound reasonable. Add some broad match keywords. Raise the daily budget on the campaign that’s limited. Switch the bid strategy to something automated.

Nobody’s lying to you. The person on the call may well believe every word.

But it’s worth understanding what that number actually is before you agree to anything, because it isn’t the health check it sounds like.

Your optimisation score measures how closely you follow Google’s advice. It does not measure how well your account performs.

Those are very different things, and the gap between them is where a lot of small budgets quietly go.

WHAT THE SCORE ACTUALLY MEASURES

Google describes the optimisation score as an estimate of how well your account is set to perform. It runs from 0 to 100 per cent, and it’s available at both account and campaign level.

Read the mechanism rather than the description, and it gets clearer.

The score reflects how many of Google’s recommendations you’ve acted on, weighted by how impactful Google estimates each one to be.

That’s it. It’s a compliance measure.

What it doesn’t include

It doesn’t measure your conversions. It doesn’t measure your cost per acquisition. It doesn’t know your margins, your capacity, or whether the enquiries you’re getting are any good.

A business could double its cost per qualified enquiry and watch the score go up, provided it did so by following recommendations.

Equally, a well-run account with deliberate settings can sit in the seventies indefinitely and be in excellent shape.

The dismissal quirk

Here’s the detail that gives the game away.

Dismissing a recommendation also removes it from the calculation. So you can raise your score by declining suggestions as well as by accepting them.

A number that improves whether you act or decline isn’t measuring your account. It’s measuring whether you’ve responded to a queue.

That’s worth knowing next time someone presents the figure as a diagnosis.

WHERE RECOMMENDATIONS CAME FROM

This is the piece of context that reframes everything, and it’s not a conspiracy theory. It’s the documented history.

Recommendations began as an internal tool for Google’s own sales representatives. The purpose was to help reps spot upselling opportunities in the accounts they looked after.

Crucially, there was a human filter. A rep looked at the suggestion, decided whether it made sense for that particular advertiser, and only raised it if it did.

Then recommendations moved into the interface, surfacing automatically in every account.

The filter went with the move. What used to be a shortlist somebody thought about is now a queue generated by a system that doesn’t know your business.

Why this matters more than it sounds

I want to be fair here, because there’s a lazy version of this argument and it isn’t true.

Google doesn’t benefit from your account failing. A business that wastes its budget and stops advertising is worth nothing to them. Plenty of recommendations genuinely improve accounts.

But the incentives aren’t symmetrical, and that’s the part to hold onto.

A recommendation that increases your spend is good for Google whether or not it works for you. A recommendation that reduces your spend is rarely suggested, even when it’s the right answer.

And if a suggestion goes badly, the cost lands entirely on you.

The person recommending the change and the person carrying the risk are not the same. That’s the whole argument.

THE RECOMMENDATIONS THAT COST THE MOST

Four come up repeatedly, and all four move money rather than fixing faults.

Adding broad match keywords

This is the most expensive one, and it arrives dressed as growth.

Google offers to add broad match versions of keywords you already run on exact or phrase match. The pitch is that you’re missing relevant searches.

What actually happens is that your ads start appearing for queries you never chose. Spend rises, and conversion rate usually falls, because a chunk of that new traffic was never going to buy.

Broad match isn’t inherently bad, and in accounts with strong conversion data and tight negative lists it can work well. Bolted onto a small account with a thin negative list, it’s the fastest way to spend more for less.

Raising the budget on a limited campaign

You’ll see this whenever a campaign is limited by budget, which for most small businesses is a permanent and deliberate state.

The recommendation isn’t wrong that more budget would produce more volume. It usually would.

What it can’t tell you is whether the extra volume comes at an acceptable cost. The cheapest, most relevant clicks tend to be bought first. The additional ones are often less good.

That’s a question to test with a small increase and a month of measurement, not a box to accept on a call.

Changing the bid strategy

Suggestions to move from manual bidding to Maximise Conversions or Target CPA come up constantly.

Automated bidding is often the right choice. It’s also entirely dependent on your conversion tracking being accurate.

If your tracking counts form views rather than submissions, or double-counts, or fires on the wrong page, you’re handing the algorithm a broken signal and asking it to optimise hard towards it.

Fix the tracking first. Then the recommendation might well be worth taking.

Expanding targeting

Widening locations, adding audiences, opting into partner networks. Each individually small.

The pattern is the same as the others. More reach, more spend, and no assessment of whether the extra reach contains anyone who’ll buy.

THE RECOMMENDATIONS WORTH TAKING

I don’t want to leave the impression that the whole feature is noise, because that’s not true and it would cost you.

Some recommendations are genuinely useful, and they share a characteristic. They fix something broken rather than spending more money.

Tracking and measurement fixes

If Google tells you a conversion action isn’t recording properly, or a tag is missing, take that seriously.

Broken tracking undermines everything else in the account, and these alerts are often the first sign of it.

Ad extensions and assets

Suggestions to add sitelinks, callouts, structured snippets or an image asset are usually worth doing.

They improve how your ad looks, they cost nothing extra, and the downside risk is close to zero.

Do write them yourself rather than accepting auto-generated versions, so the wording is yours.

Clear technical errors

Disapproved ads, broken final URLs, conflicting negative keywords blocking your own traffic. These are faults, and fixing faults is a free improvement.

The conflicting negatives one is genuinely valuable and hard to spot manually.

The sorting rule

Here’s the test that works on a recommendation you’ve never seen before.

Ask whether it fixes something broken or spends more money.

Fixes are usually worth taking, quickly. Spending changes are worth testing, slowly, on your own timetable.

That single question sorts most of the queue without needing to know the specifics.

THE RECOMMENDATION YOU WILL NEVER RECEIVE

Here’s an exercise worth doing once. Scroll your recommendations and look for anything that would reduce your spend.

You won’t find one.

You’ll never be told to pause a campaign that isn’t working. Or to narrow your locations because half your enquiries come from outside your service area. Or that your budget would do more in a different campaign entirely. Or that Search is capped, and the money should go somewhere else.

Every one of those is a real recommendation a competent person might make about your account. None of them appears.

Why the absence matters

This isn’t evidence of bad faith. A system built to find opportunities in an advertising platform will find opportunities to advertise. That’s what it was designed to do.

But it does tell you what the queue is and isn’t.

It’s a list of things you could add. It’s not a review of your account, because a review includes subtraction.

So the recommendations can be a useful input, and they can’t be the whole conversation. Everything on the list points in the same direction, and no honest assessment of an advertising account points in only one direction.

What a real review includes

Somebody genuinely assessing your account would ask which campaigns to stop, not just which to expand.

They’d look at whether the enquiries are any good, which no automated system can judge, because only you know which ones turned into work.

They’d check whether your tracking is measuring the right thing before trusting any number built on top of it.

And they’d be willing to tell you to spend less. That last one is the clearest test of whose interests are being served.

WHAT THE SCORE IS ACTUALLY USEFUL FOR

I’ve been hard on the score, so let me give it its due, because dismissing it entirely costs you something too.

It’s a decent error-surfacing mechanism

The score is a poor diagnosis and a reasonable alarm.

Broken conversion tracking, disapproved ads, conflicting negative keywords blocking your own traffic, a dead final URL. These are genuine faults, and the recommendations queue often catches them before you do.

For a business without someone checking the account weekly, that early warning has real value.

It’s a prompt to look

A score that drops sharply usually means something changed. That’s worth investigating, even if the specific recommendation behind it isn’t worth taking.

Use the movement as a trigger to look, rather than the number as a verdict.

It’s a shared reference point

If you work with an agency or a freelancer, the score gives you something concrete to ask about.

Not “why isn’t it 100”, which is the wrong question. Better: “Which of these have you declined, and why?”. A good answer to that tells you a great deal about how carefully your account is being managed.

Someone who has thought about each suggestion will have reasons. Someone who has ignored the account entirely won’t.

So the honest position is neither “follow it” nor “ignore it”. Read the queue, take the fixes, treat the spending suggestions as questions.

AUTO-APPLY, AND WHAT TO SWITCH OFF

Auto-apply lets Google implement recommendations without asking you. It’s controlled by category in the Recommendations section, and each type can be toggled independently.

Some accounts have categories switched on that nobody remembers enabling. It’s worth a look regardless of your view on the feature.

Leave off anything that changes strategy

Keep auto-apply off for anything affecting keywords, match types, targeting, bidding strategy, ad copy and campaign creation.

Those are decisions about how your money gets spent, and they should be made deliberately by somebody who knows your business.

The specific risk is that a change happens quietly, performance shifts, and three weeks later you’re trying to work out what altered and when.

Low-risk categories are defensible

Maintenance-style items, removing genuinely redundant elements, minor quality fixes. Reasonable people enable those, and it saves time.

The test is the same one as before. Is this fixing something or spending something?

Check what’s on right now

Take two minutes and look at your current auto-apply settings.

If anything in the keyword, bidding or targeting categories is enabled and you didn’t turn it on deliberately, switch it off and then look at whether anything changed recently.

HOW TO HANDLE THE CALL

This is the part nobody writes about, and it’s where most small businesses actually encounter all of this.

Be straightforward about it. The person calling is doing their job; they often know the platform well, and the call can be useful. You don’t need to be defensive or rude.

You just need to not decide anything on the call.

Three sentences that solve it

Say you’d like the recommendations sent by email so you can review them properly.

Ask which ones would increase your spend and which would reduce it. That single question reframes the whole conversation, politely.

Then say you’ll come back after checking against your own numbers.

Nobody reasonable objects to any of that. If somebody does object, that tells you something worth knowing.

The question to bring to any suggestion

For each recommendation, ask what it would do to your cost per qualified enquiry.

Not clicks. Not impressions. Not the score. The cost of an enquiry that was actually worth having.

If the answer is that nobody knows, that’s fine, but it means the suggestion is a test rather than an improvement. Test it with a small budget and a month of data.

Never let a change be applied during the call itself. There’s no recommendation so urgent that it can’t survive a week of thought.

WHY 100 PER CENT IS A BAD GOAL

Chasing the number is where this turns expensive.

An account at 100 per cent has accepted or dismissed everything Google suggested. That tells you the queue is empty. It tells you nothing about the business.

Meanwhile, plenty of the best decisions in an account will actively lower the score. Declining broad match. Keeping a budget deliberately capped. Running a narrow, profitable campaign that Google would rather see widened.

Every one of those is a deliberate choice, and every one shows up as a missed opportunity in the score.

A well-run account often sits in the seventies. That’s a sign of judgement, not neglect.

If you want a number to aim at, use one from your own business. Cost per qualified enquiry. Enquiries per month. Revenue against ad spend.

Those numbers care whether you make money. The optimisation score doesn’t know whether you do.

BUILDING A SCORECARD THAT MEANS SOMETHING

If you’re going to stop trusting one number, you need something to replace it with. Otherwise you’ve swapped a flawed measure for no measure, which is worse.

Here are the four I’d track. All of them are things the platform can’t judge for you.

Cost per qualified enquiry

The most important number in most small accounts, and the one almost nobody records.

Not cost per conversion. Cost per enquiry that was genuinely worth having.

Count them by hand if you have to. Take the month’s spend, divide by the number of enquiries you’d have been glad to receive, and write the figure down.

The gap between that and your reported cost per conversion is usually instructive on its own.

Enquiries per month, tracked over time

A simple count, month by month, in a spreadsheet.

Advertising accounts move slowly, and monthly numbers are noisy. A twelve-month line tells you far more than any single month’s dashboard.

It also protects you from the most common mistake in this whole area, which is reacting to a bad fortnight.

Share of enquiries you can actually serve

This one gets forgotten, and it matters more than people expect.

If you’re turning work away, more volume isn’t the goal, and any recommendation to increase spend is answering a question you don’t have.

Growth in advertising only helps if the business can absorb it.

Branded search volume

A slow indicator of whether people know who you are.

Check it quarterly in Search Console. If it’s rising, your marketing is doing something beyond the last click, and that’s worth knowing when you judge channels that don’t convert directly.

How to use them

Look at all four once a month. Ten minutes.

Then when a recommendation arrives, you have something to check it against. Would this improve cost per qualified enquiry? Do I even want more volume this quarter?

Those questions are answerable when you have your own numbers, and unanswerable when the only figure in front of you is a percentage on a dashboard.

A number you chose beats a number you were given, every time.

A WORKED EXAMPLE, START TO FINISH

Here’s a hypothetical to show the shape of it. A specialist installer on eight hundred pounds a month, running two Search campaigns.

They get the call. Score at 71 per cent, with three recommendations that would take it to 96.

Recommendation one: add broad match versions of eleven existing keywords.
Recommendation two: raise the daily budget on the limited campaign by forty per cent.
Recommendation three: switch to Maximise Conversions.

They ask for it by email and check each against the sorting rule. Does it fix something or spend something? All three spend something.

So they treat all three as tests rather than fixes.

The bid strategy one goes first, because it depends on tracking. They check and find the conversion action is counting anyone who reaches the contact page rather than anyone who submits the form. The recommendation would have optimised hard towards a meaningless signal.

They fix the tracking. That takes a fortnight of clean data before anything else is worth doing.

The budget increase they test properly. Fifteen per cent for one month, not forty. Enquiries rise slightly. Cost per qualified enquiry rises by more. They keep the smaller increase and decline the rest.

Broad match they decline outright, because their negative keyword list is thin and their budget is too small to fund the learning.

Their score stays in the seventies. Their cost per qualified enquiry falls, mostly because of the tracking fix that no recommendation had flagged as urgent.

That’s the whole method. Sort, test the spending ones slowly, take the fixes quickly, and let the score land wherever it lands.

THE MISTAKES THAT WASTE THE MOST MONEY

Treating the score as a health check. It measures compliance with a queue of suggestions, not account performance.
Applying recommendations during the call, before checking anything against your own numbers.
Accepting broad match additions with a thin negative keyword list and a small budget.
Switching to automated bidding before verifying that conversion tracking is accurate.
Leaving auto-apply enabled on keyword, bidding or targeting categories without knowing it’s on.
Chasing 100 per cent, which usually means accepting changes that widen spend.
Ignoring the genuinely useful recommendations, particularly tracking errors and conflicting negatives, because you’ve decided the whole feature is noise.
Never asking which recommendations would reduce spend. The answer is almost always none, and noticing that is the point.

FREQUENTLY ASKED QUESTIONS

What is a good Google Ads optimisation score?

There isn’t one, in the sense people mean.

The score reflects how many of Google’s recommendations you’ve acted on, weighted by Google’s estimate of their impact. It doesn’t measure conversions, cost per acquisition or profit.

A carefully run account can sit in the seventies. A wasteful one can reach 100 per cent. Use your own numbers instead.

Should I apply Google Ads recommendations?

Some of them. Sort them with one question: does this fix something broken, or does it spend more money?

Tracking fixes, extension suggestions and technical errors are usually worth taking straight away.

Broad match additions, budget increases, targeting expansion and bid strategy changes are tests. Run them on your own timetable with your own measurement.

How do I turn off auto-apply recommendations?

The settings live in the Recommendations section of your account, where each category can be switched on or off separately.

Leave it off for keywords, match types, targeting, bidding strategy, ad copy and campaign creation.

It’s worth checking what’s currently enabled, because these sometimes get switched on and forgotten.

Does optimisation score affect my ad performance?

The score itself doesn’t. It’s not part of the auction, and it doesn’t influence where your ads appear.

Applying the recommendations behind it can change performance, in either direction depending on the recommendation.

Raising the number for its own sake achieves nothing at all.

Why does Google keep telling me to increase my budget?

Because a campaign is limited by budget, which is a normal and often intentional state for a small business.

More budget usually does produce more volume. Whether it produces enquiries at a cost you’d accept is a separate question, and it’s yours to answer.

Test a small increase for a month and compare cost per qualified enquiry before and after.

Are Google Ads recommendations a sales tool?

They started as one. Recommendations were originally internal, helping Google’s sales representatives find upselling opportunities in accounts they managed.

A rep decided which suggestions were worth raising with which advertiser. That human filter disappeared when recommendations moved into every account automatically.

None of which makes them worthless. It means the judgement about relevance is now yours.

Is my agency’s optimisation score a good way to judge them?

On its own, no, and judging them on it will push them towards decisions that suit the score rather than your business.

A better question is which recommendations they’ve declined and why. A considered answer, with reasons specific to your account, is a good sign.

An account nobody has looked at will show a queue of unaddressed suggestions, including genuine faults like tracking errors. That’s the thing to be concerned about, rather than the percentage itself.

Do dismissed recommendations affect the score?

Yes, which is the quirk that reveals what the number is.

Dismissing a recommendation removes it from the calculation, so declining suggestions raises your score much as accepting them does.

A measure that improves whether you act or decline isn’t assessing your account. It’s tracking whether you’ve cleared a queue.

Should I tell my Google rep I’m not applying their recommendations?

You can simply say you’re reviewing them against your own numbers and will come back.

That’s a normal, professional answer and it ends the pressure without any confrontation.

You’re not obliged to justify a decision about your own budget to anybody selling advertising.

WHERE TO START THIS WEEK

Two things, and neither takes long.

First, check your auto-apply settings. If anything in the keyword, bidding or targeting categories is switched on and you don’t remember enabling it, turn it off. Then look at whether anything shifted in the account recently.

Second, write down the number you actually care about. Cost per qualified enquiry is the one I’d use. Work out what it was last month.

That figure is your real optimisation score. It shows whether the business made money, which is more than the one on the dashboard can say.

Then next time the call comes, you’ll have something to check the suggestions against. That’s the whole difference between a useful conversation and an expensive one.

Did You Enjoy This Blog Post?

I hope you enjoyed this blog post, and thank you so much for being here. We also upload videos to our YouTube channel every weekday. Please subscribe so you are one of the first to be notified.

If you enjoyed this blog, you may also like: