Every Google Ads account has an optimisation score. It sits in the top-left corner of the Recommendations tab, a percentage from 0 to 100. Google calls it a measure of how well your account is set up to perform. Some agencies chase 100% religiously. Others ignore it. Both camps are wrong.
The score is a useful signal. It is not a performance metric. Knowing what it actually measures, and which recommendations are worth accepting versus quietly serving Google's revenue, is one of the more practical skills a media buyer picks up. Here is how I think about it.
What the Optimisation Score Actually Measures
The optimisation score is a real-time estimate of how closely your account follows Google's recommended best practices. It is calculated per campaign and aggregated to an account-level percentage. Each recommendation in the list carries a weight (a potential score uplift) that reflects how much Google thinks the change would move performance.
Google generates recommendations from your current settings, historical performance, auction dynamics, and machine-learning models that predict the likely outcome of each change. The score also accounts for whether you have already dismissed a recommendation. Dismissing one removes it from the calculation, which bumps your displayed score even though nothing in the account actually changed.
That last bit is worth sitting with. The optimisation score is not a fixed assessment of account health. It is a dynamic measure of how many pending suggestions you have either applied or dismissed. Two accounts with identical performance can have wildly different scores simply because one has more outstanding recommendations. An account hitting target CPA at 65% is outperforming an account at 95% that accepted every budget bump and blew past its monthly cap.
The score also updates in real time. New recommendations appear when Google's models detect opportunities (or what Google calls opportunities), and existing ones expire when they are no longer relevant. Your score can drop overnight without you touching the account. Google just generated new suggestions you have not acted on yet.
How the Score Is Calculated
Google does not publish the exact formula, but the mechanics are observable. Each recommendation in the Recommendations tab shows a percentage uplift, for example, "Apply this recommendation to improve your optimisation score by up to 3.2%." Add up the pending uplifts, subtract from 100%, and you have your current score.
Recommendations fall into several categories:
- Bidding and budgets. Suggestions to switch bid strategies, raise budgets, or adjust target CPA and ROAS values. These tend to carry the heaviest weights.
- Keywords and targeting. Adding new keywords, removing redundant ones, broadening match types, or expanding audience targeting.
- Ads and assets. Creating new responsive search ad variations, adding sitelinks, callouts, or structured snippets, and improving ad strength.
- Repairs. Fixing disapproved ads, resolving conversion tracking issues, or addressing policy violations. These are almost always worth acting on.
The weighting is not equal across categories. Budget and bidding recommendations carry disproportionate weight because Google's models estimate they will move conversion volume the most. This is technically true. Spending more money does produce more conversions. It also conflates volume with efficiency. An agency working to a fixed monthly budget gets nothing from a recommendation to raise that budget by 40%, no matter how good it would make the score look.
You can also dismiss recommendations you disagree with. Dismissing pulls the recommendation out of the calculation for a window (usually 30 days), which increases your score. This is a legitimate way to manage the metric without making changes you think will hurt the account. Some agencies run a weekly review specifically to dismiss irrelevant suggestions and keep the visible score in a range that does not panic clients.
What Is a Good Optimisation Score?
Every client asks this. The honest answer is: it depends on what you accepted and what you dismissed.
Somewhere between 70% and 80% is typically healthy for a well-managed account. It tells you that you have applied the recommendations that make sense and dismissed the ones that do not. Accounts in that range have usually handled the high-value stuff (broken tracking, missing ad variations, the right bid strategy for the goal) and declined the spend-increasing suggestions that did not match the brief.
Chasing 100% is almost always counterproductive. To get a perfect score, you would need to accept every single recommendation Google surfaces, including ones that broaden targeting beyond your audience, raise budgets the client never approved, and switch bid strategies on campaigns that were doing fine. Chasing the perfect score can actively hurt the account.
A score below 50% usually means there are real problems. At that level you are probably ignoring recommendations that would genuinely help: disapproved ads sitting unfixed, missing extensions, conversion tracking that has been broken for weeks. The repair-style recommendations at the bottom of the list are almost always worth doing.
Treat the score as a to-do list, not a report card. Review it regularly, act on what makes sense, dismiss what does not, and do not let the number drive strategy. A 75% score with strong performance metrics beats a 95% score with blown budgets and a deteriorating CPA every time.
Recommendations That Are Usually Safe to Accept
Not every recommendation is a trap. A lot of them are genuinely useful, especially on accounts that have been left alone for a while. These are the categories I almost always apply.
Repair recommendations. These flag broken stuff: disapproved ads, conversion tracking errors, policy violations, broken sitelinks, missing final URLs. There is no strategic reason to leave any of it unfixed. Free wins for both the score and the actual account. If you keep running into the same errors and want a deeper checklist, our Google Ads troubleshooting guide walks through the most common failure modes and how to clear them.
Ad extensions and assets. Adding sitelinks, callouts, structured snippets, and image extensions costs nothing and gives your ads more real estate on the SERP. The CTR lift from a complete extension set is well documented. If Google flags missing extensions, add them.
Responsive search ad improvements. If you have ad groups running a single RSA, or ads stuck on "Poor" ad strength, the recommendation to add headlines and descriptions is fair. More variations give Google more combinations to test, and "Good" or "Excellent" strength ads tend to enter more auctions. The trick is to write genuinely distinct variants that cover different angles (feature, benefit, social proof) instead of rewording the same line five times.
Negative keyword additions. When Google flags search terms that have been triggering your ads and converting badly, the recommendation to add them as negatives is nearly always right. Cleaning up irrelevant queries is one of the highest-impact optimisation strategies at any account size.
Conversion tracking upgrades. Recommendations to turn on enhanced conversions, move to Google Tag Manager, or migrate from older tracking are worth doing first. Better conversion data feeds better automated bidding, which lifts performance across the account.
Recommendations You Should Think Twice About
These are the suggestions that tend to serve Google's revenue more than your client's results. They are not always bad, but they need a real look, not a one-click accept.
"Raise your budget" recommendations. This is the most common source of tension between the score and how agencies actually run accounts. Google's models notice a campaign is hitting its daily cap and missing impressions, so they suggest a higher budget. If the campaign is performing and there is approved budget sitting unused, fine, take it. In most agencies the budget is fixed by the client. Raising one campaign means pulling from another, and the score does not see that constraint. Dismiss these unless you have explicit approval to spend more.
"Broaden your match types" recommendations. Google often suggests moving exact match keywords to broad, on the theory that Smart Bidding will handle relevance at the auction. In theory, that is true. Google's matching has gotten genuinely better. In practice, broadening match types on campaigns with tight CPA targets usually triggers a painful exploration period where CPA inflates while the algorithm wanders. If you broaden, do it gradually, campaign by campaign, with close monitoring, not as a bulk change.
"Enable auto-apply" recommendations. Google offers to apply certain recommendation types automatically, no human review. For low-risk things (removing redundant keywords, say) this can save time. For higher-impact ones like bid adjustments, budget changes, or targeting changes, auto-apply removes the review step that prevents costly mistakes. An auto-applied targeting change on a Friday afternoon can burn through a weekend's budget before anyone notices on Monday. Keep auto-apply tightly scoped to categories you have specifically vetted as safe.
"Use Maximize Conversions" or "Switch to broad match" on low-volume campaigns. These show up on campaigns with fewer than 30 conversions a month. The catch is that Smart Bidding needs conversion data to learn from. Move a low-volume campaign to Maximize Conversions and you often get erratic spending while the algorithm tries to find a signal that is not really there. Keep low-volume campaigns under tighter manual or semi-automated control until conversion volume justifies full automation.
"Add audience segments" expansion recommendations. Google sometimes suggests adding in-market or affinity audiences to search campaigns. Audience layering can give useful bid modifiers, but adding audiences for observation is very different from adding them for targeting. Check whether the recommendation is observe (harmless) or target (can dilute your keyword targeting) before accepting.
How to Discuss the Score with Clients
The score creates a communication problem for agencies. Clients see it in their Google Ads UI and reasonably want to know why it is not 100%. Google presents the number like a performance indicator, so 72% looks like a failing grade to anyone without PPC context.
The simplest fix is to bring it up before the client does. Add a short optimisation score section to your regular reporting with the current score, the recommendations you accepted (and why), and the ones you dismissed (and why). That reframes the metric from "how good is my account" to "here is how we are handling Google's suggestions on your behalf."
Be specific about dismissals. "We dismissed the recommendation to raise your Search campaign budget by 35% because your monthly budget is fixed at $15,000 and the campaign is already pacing correctly" lands a lot better than "we do not accept all of Google's suggestions." Clients respect specificity. What worries them is a low number with no explanation.
It also helps to be honest about the conflict of interest. Google makes money when you spend more. Some recommendations (raising budgets, broadening targeting, enabling extra ad formats) increase the cheque you write to Google. That does not automatically make them wrong, but it does mean they need to be weighed against the client's actual goals rather than accepted on autopilot. Most clients get this once you say it out loud.
If a client really insists on a higher score, the dismiss workflow is the tool. Review pending recommendations weekly, accept the useful ones, dismiss the rest with documented reasoning. The visible score stays in a healthy range and the account does not absorb changes that hurt performance. Over time, once the client sees that a 75% score lines up with strong results, the fixation on the number tends to fade.
For a broader framework on evaluating account health beyond Google's built-in score, our 30-minute audit guide covers the dimensions that actually predict performance. And if you want automated, cross-platform monitoring that goes beyond what the optimisation score can tell you, try Pace free and see how continuous analysis replaces poking at the score every Monday.
