Google Ads Optimization Score and What It Ignores

Open any Google Ads account and the number is there, top right of the Recommendations page: 78%. 84%. 61%. Below it, a stack of cards. Add 14 keywords. Raise your budget by $30/day. Switch these phrase match keywords to broad match. Apply all and watch the number climb.

Now scroll the whole list and try to find a single card that asks you to stop paying for something. There isn't one.

What the optimization score actually measures

Optimization score is Google's estimate of how completely your account uses the settings and features Google recommends. It runs from 0% to 100%, and Google states that 100% means the account is set up to perform at its full potential according to Google's recommendations. That last clause is the whole article.

The score is not derived from your profit, your margin, or your cost per acquisition against a target you set. It is a weighted rollup of recommendation cards, where each card carries an "uplift" percentage. Apply a card, the score rises by that uplift. Dismiss a card, the score also rises — the card leaves the pool. Two completely opposite actions produce the same number.

That alone should tell you what kind of instrument you are holding. A metric where "do the thing" and "refuse to do the thing" both improve the reading is not measuring account health. It is measuring how many open questions Google has about your account.

The score is also used for things beyond your dashboard. Agency partner status has historically leaned on maintaining a minimum average optimization score across managed accounts. So somewhere in the chain, someone has an incentive to apply cards that you would not have applied yourself. If you work with an agency, that is worth an honest conversation — and worth reading alongside the honest maths of agency versus in-house.

The recommendation categories, and their direction

Google groups recommendations into buckets: keywords and targeting, bidding, ads and extensions, repairs, measurement, budget. Look at what each bucket asks for.

• Keywords and targeting — add keywords, add broad match variants, expand to new locations, remove conflicting negative keywords.
• Bidding — switch to Maximize conversions, raise your target CPA, lower your target ROAS.
• Budget — your campaign is limited by budget, raise it.
• Ads — add more headlines, add more assets, turn on automatically created assets.
• Campaign types — add a Performance Max campaign, upgrade Smart Shopping.

Every one of those is an expansion. The only recommendation that regularly asks you to remove something is "remove conflicting negative keywords" — which is a recommendation to un-block traffic you deliberately blocked. It is technically a removal and directionally an expansion.

Why there is no "add negative keywords" recommendation

Google has all the data it would need to build one. It knows which search terms you paid for, how much you spent on each, and how many conversions each produced. campaign_search_term_view in the API returns exactly that for Search, Shopping and Performance Max. Building a card that says "these 40 terms cost you $612 in 90 days with zero conversions, add them as negatives" is not a technical problem.

It doesn't exist because the score is built around what Google believes will grow your results, and the mechanism by which Google grows results is reach. A negative keyword is subtraction. It removes auctions you were entering. In the short run and in Google's own metrics, it makes impressions go down, clicks go down and spend go down. Conversions, in the best case, stay flat.

There is a defensible version of Google's position here. Broad match and Smart Bidding are designed to learn from a wide signal pool. Cutting queries out of that pool does reduce what the system has to learn from. Google is not lying when it says aggressive negative keyword lists can constrain automated bidding — and we have written about the real signs that your negatives are hurting you, because that case is real.

But "some negatives hurt" is a long way from "no recommendation to add any, ever." The honest position is that the two effects need weighing against each other in your account, with your data. The score does not weigh them. It only ever counts one side.

The "remove conflicting negative keywords" card deserves a look

This one is worth handling properly rather than dismissing on reflex. Google flags it when a negative keyword is blocking a keyword you are actively bidding on. Sometimes that is a genuine mistake — someone added free as a broad negative and it is now blocking free shipping returns policy searches that convert fine.

More often it is a match type accident. Broad negatives behave unpredictably across match types, which is why we tell people to default to phrase and exact. If you want the mechanics, negative keyword match types don't work like positives covers what actually blocks what.

So: read the card, check which negative it names, check whether that negative has genuinely been costing you conversions. Then apply or dismiss with a reason. What you should not do is bulk-apply it to get 4% of score back, because that silently reopens traffic somebody blocked on purpose.

What the score misses in a real account

Here is the gap made concrete. Take a mid-size account, roughly $8,000 a month across Search, Shopping and one Performance Max campaign. Optimization score sits at 82%.

The open recommendations are: add 22 keywords, raise the Shopping budget by $20/day, add 3 headlines to two responsive search ads, and switch four phrase match keywords to broad. Apply all four and the score goes to 97%.

Now open the search terms report for the last 90 days. Typical findings in an account like that:

• [brand] reviews reddit — $180 spent, 0 conversions.
• how to fix [product] yourself — $95 spent, 0 conversions.
• [competitor] pricing — $240 spent, 1 conversion, a 20-minute demo no-show.
• [product] jobs and [product] salary — $60 between them, 0 conversions.
• free [product] template — $310 spent, 0 conversions.

That's $885 across 90 days, or roughly $295 a month, about 3.7% of spend, on terms that were never going to buy. The score says nothing about any of it. It doesn't score it down. It doesn't flag it. And if you apply the "switch to broad match" card, several of those terms get more traffic, not less.

We see the share of spend on never-converting terms land somewhere between a few percent and the high teens in accounts we audit, depending mostly on how much broad match and Performance Max is running and how long it's been since anyone looked. The point is not the percentage. The point is that the number displayed at the top of your Recommendations page is completely blind to it. If you want a method for putting a defensible figure on your own account, how to calculate wasted ad spend honestly walks through it without the inflated arithmetic most agencies use in pitch decks.

Job seekers, researchers and DIYers are invisible to the score

Three categories of query cost money in almost every account and appear in no recommendation:

People looking for work. [your company] careers, [your industry] salary, [job title] apprenticeship. They click, they read, they bounce. They never convert and they never will. Blocking job seeker terms is one of the highest-certainty cleanups available and the score has never once mentioned it.

People doing research who will buy nothing. what is [product], [product] vs [product] comparison chart, is [product] worth it. Some of these belong in your funnel. Many are students and tyre-kickers. That requires judgement, which is precisely what a generic recommendation engine cannot supply.

People trying to avoid paying. free, diy, template, crack, torrent, cheap, second hand. If you sell a $4,000 service, cheap [service] is not a lead. The score doesn't know your price point.

Where the blind spot costs most: Performance Max

Performance Max is the campaign type the optimization score pushes hardest, and the one where the missing recommendation hurts most.

The score will suggest you create a Performance Max campaign, add more asset groups, turn on final URL expansion, broaden your audience signals. It will not suggest a single negative keyword, even though the campaign type now accepts up to 10,000 campaign-level negative keywords and reports its search terms in the same campaign_search_term_view as everything else.

Performance Max also can't use shared negative keyword lists. So if you maintain one good shared list of junk terms across your account, that list simply does not apply to your Performance Max campaigns. The traffic flows straight past it. Nothing in the optimization score tells you this. You have to know it, and then push the same negatives into each Performance Max campaign individually.

That combination — hardest-pushed campaign type, no shared lists, no recommendation to add negatives — is why Performance Max is usually where the money goes missing. We've mapped the specific hiding places in where Performance Max wasted spend hides, and the practical ceiling question in negative keywords in Performance Max and the 10,000 cap.

Display, Video and App report nothing at all

One more limit, and it is Google's, not ours. campaign_search_term_view covers Search, Shopping and Performance Max. Display, Video and App campaigns report no search terms, because in most cases there is no search query behind the impression. There is nothing to clean there with negative keywords. If your score is being lifted by recommendations to expand into Display, understand that you are moving spend into a channel where the search-term cleanup lever does not exist.

How to read the score without letting it drive

The score is useful as a list of things Google noticed. It is not useful as a grade. Here is how to use it without giving it the wheel.

Read every card, apply few. Some cards are genuinely good. "Fix your conversion tracking" is a repair card and you should absolutely act on it — an account with broken tracking makes every other decision guesswork. Ad strength and missing sitelinks are usually cheap wins. Budget and match type cards are strategy decisions disguised as maintenance.

Dismiss with a reason and move on. Dismissing raises the score. If you have decided broad match is wrong for a campaign, dismiss the card. Don't leave it sitting there dragging the number down and tempting whoever logs in next.

Turn off auto-apply. Google offers to apply certain recommendation types automatically. There are auto-apply options that add broad match keywords and raise budgets. If auto-apply is on and nobody checks the search terms report, the account expands on a schedule and contracts never. That is the worst possible configuration.

Never chase 100%. An account at 100% has applied or dismissed every card, which tells you nothing about whether the account makes money. We regularly see profitable accounts in the 70s and loss-making accounts in the 90s.

Run cleanup on its own schedule. The score has no cleanup function, so cleanup needs its own slot in your week. How often you should check search terms depends mostly on spend and how much broad match you run, but the answer is never "when Google reminds me," because Google will not.

The order of operations matters

If an account is spending and not converting, the recommendation cards are close to the last place to look. Tracking first, then search terms, then structure, then bids. We laid out the full sequence in the order to check things when Google Ads spends with no conversions. Applying eleven recommendation cards to an account with a broken conversion tag just gets you to the wrong place faster.

For a periodic sanity pass, the one-hour audit checklist includes the search-term work that the score leaves out. Run it quarterly at minimum.

Building the recommendation Google won't build

The card that should exist is simple to specify: show me every search term from the last 90 days with more than X in spend and zero conversions, grouped so I can block them in batches.

You can build it yourself. Open the search terms report, set the date range to 90 days, add the Conversions column, sort by cost descending, filter to conversions = 0. Read down the list. Everything obviously wrong gets added as a phrase or exact negative. This takes 20 to 40 minutes the first time in a mid-size account and 10 minutes a week after that.

Two things to know before you start. First, the report doesn't show you everything — Google withholds terms below a privacy threshold, so some of your spend sits in a bucket you will never see itemised. What the search terms report hides covers how much and why. Second, the Match Source column tells you whether a term came from a keyword or from Performance Max or dynamic targeting, which changes where the negative has to go — with the caveats set out here.

Some calls are genuinely hard. Competitor terms are the classic one: expensive, low conversion rate, occasionally the source of your best customers. That decision is yours and depends on your margin and your sales cycle — block or keep competitor searches lays out both sides rather than pretending there's one answer.

This is the job PPCleaner does nightly. It reads the search terms your account paid for across Search, Shopping and Performance Max, judges which were never going to convert, and adds them as negative keywords. It never touches budgets, bids, ads or keywords — so it will never fight with a recommendation you decided to apply. It will never block a term that converted in the last 90 days. And every block can be undone in one click for 90 days, which matters, because some of the calls are judgement calls and you should be able to reverse ours.

It will not raise your optimization score. Blocking a bad search term isn't a thing Google scores. If your agency reports on optimization score, expect that number to stay exactly where it is while your cost per conversion moves.

The one number worth watching instead

Replace optimization score with a number you compute yourself: the share of last month's spend that went to search terms with zero conversions in the last 90 days.

Pull it monthly. In a clean, mostly-exact-match Search account it should be low single digits. In an account leaning on broad match and Performance Max it will be higher, and some of that is legitimate prospecting you want to keep paying for. The absolute figure matters less than the direction. If it climbed from 6% to 11% over a quarter, something expanded and nothing contracted — which is exactly what happens when the only guidance an account receives comes from the Recommendations page.

Then go find where it went. How to find wasted ad spend in Google Ads is the practical version, and if you're about to add a few hundred negatives in one sitting, check how many negative keywords are too many first — there is such a thing as overdoing it, and the fix for an unguarded account is not an over-guarded one.