Every Google Ads rep opens the call the same way: "Your optimisation score is 78% — let's get that up." It sounds like a health check. It behaves like a sales target.
The optimisation score is a 0–100% number Google puts on your account, with a queue of recommendations that each carry a "points uplift". Apply them all and you hit 100%. The unspoken promise is that 100% means a better-performing account.
It doesn't. The score measures how closely your account matches Google's default preferences: broader matching, higher budgets, automated bidding, fewer of your own controls. Some of those are good ideas. Others hand decisions to the auction that you'd usually want to keep for yourself. The whole skill is telling the two apart before you click.
The TL;DR: optimisation score is a compliance metric, not a performance metric. It rises as your account looks more like Google's defaults, not as it makes more money. Treat the recommendation feed as a to-review list, never a to-apply list. The three that most often wreck accounts on "apply all": adding broad match keywords, letting automated recommendations run without review, and accepting a lower target CPA/ROAS that starves volume. Turn auto-apply off, review the feed weekly, and judge every recommendation against your own conversion data before you touch it.
Why a higher score can mean a worse account
The score is built from recommendations, and recommendations lean toward what Google's models want to serve. That's not a conspiracy. It's just what the number is built to reward. Google itself describes the score as an estimate of how well your account is set to perform, generated by its own systems.
The trouble is the gap in incentives. "Well set to perform" in Google's world means more automation and more reach. Your goal is profit at a target efficiency. Those two things agree maybe 60% of the time. The other 40% is where the score tells you to do something that lifts the number and drops your return.
Somewhere along the way 100% became the goal. It was never meant to be one. A tightly-run account sitting at 82%, because someone has actually dismissed the recommendations that don't fit the strategy, will usually beat the same account bulldozed up to 100%.
The recommendations that cost you
1. "Add these keywords" (usually broad match)
The most common uplift Google offers is adding keywords, and they nearly always arrive as broad match. Click "apply all" and you inherit a pile of loosely-matched terms overnight. Your search terms report fills up with queries you'd never have bid on.
Broad match isn't the enemy. Paired with strong Smart Bidding and clean conversion data it can genuinely work. But you can't audit what you didn't choose. Before you accept any keyword recommendation, look at what it will actually match against your history. The BigQuery n-gram workflow I use for search terms at scale exists because "apply all" makes that waste invisible one term at a time.
2. "Raise your budget" on a campaign that isn't ready
"Limited by budget" recommendations assume more spend buys more good conversions at the same efficiency. That only holds while the next pound of spend is still profitable, and most accounts hit that ceiling well before Google's estimate says they will.
Scaling budget is a signal-quality call, not a slider you drag. I've written before about why accounts stop responding to more budget at higher spends. The score has no idea where you sit on that curve, because it can't see your margins.
3. "Lower your target CPA / raise your target ROAS"
Tempting, because the score rewards it and it looks like efficiency. But tightening a Smart Bidding target throttles the auctions the algorithm will enter, and if your conversions take days to land, you're tuning against numbers that haven't finished reporting. That's the trap I dig into in the piece on conversion lag and Smart Bidding: a target that looks safe on day 3 is often strangling volume you'd have kept by day 10.
4. Auto-applied recommendations, the silent one
This is the one that actually hurts, because you never watch it happen. Auto-apply lets Google implement whole categories of recommendations for you. Overnight, keywords get added, targets shift, "conflicting" negatives get removed. You open the account to numbers that moved and nothing in your own change log to explain why.
If a change can happen to your account overnight and you didn't decide it, that isn't optimisation. It's Google optimising for Google.
A case study (composite, based on patterns I see over and over)
A composite of the lead-gen accounts I've audited, built from patterns rather than one real client. Roughly £40k a month, sitting at 94% optimisation score, rep delighted. Under the bonnet: auto-apply switched on for "keywords and targeting" and "automated bidding", broad match auto-added across three campaigns, and target CPA auto-lowered twice in a single quarter.
The symptoms over the previous 90 days were textbook. Search terms bloated with junk queries, cost per lead up 34%, and lead quality (the thing the score can't see) clearly down in the client's own CRM.
What we changed: turned off every auto-apply setting, dismissed the broad match and target recommendations, kept the two structural ones that were genuinely sound (a sitelink gap and an unused conversion action), and rebuilt the bid target from lag-adjusted data.
Over the next 90 days, same budget: cost per qualified lead down 28–35%, wasted spend on junk terms down by around 40%, and the optimisation score settled at 81%. Lower, and healthier. The rep flagged the "low" score on the next call. That reaction is the tell.
Common mistakes
- Treating the score as a KPI. It's a diagnostic feed, not a target. Report it to a client as a win and you train everyone to chase the wrong number.
- Leaving auto-apply on to save time. It saves time by removing your judgement. Switch it off and spend ten minutes a week instead.
- Dismissing everything on reflex. The opposite mistake. Some recommendations, like a missing sitelink or an untracked conversion action, are free wins. Read them properly.
- Never dismissing anything. Dismissing is supposed to happen. It tells Google that path doesn't fit your strategy and tidies the feed. A steady 78% you understand beats a 100% you don't.
Bottom line
- Turn auto-apply off for every category. Decisions get made by you, when you're looking.
- Work through the recommendation feed weekly. Read every item, apply the real structural wins, dismiss the rest.
- Never accept a keyword, budget, or target recommendation without checking it against your own conversion and margin data first.
- Stop reporting optimisation score as a performance number. Report cost per qualified outcome instead.
The accounts I see pulling ahead aren't the ones at 100%. They're run by someone who reads every recommendation, takes the three that help, and happily ignores the number when it argues with the data. The score is Google's opinion of your account. It's allowed to have one, and you're allowed to disagree. If yours is sitting near 100% and nobody's questioned it lately, the free audit is the quickest way to see what "apply all" has been doing while you weren't looking.
And if you'd rather have someone go through the feed with you, line by line, that's what the Google Ads audit service is for.
Sources and further reading: