Your ad platform’s scoreis not grading your marketing.

Two numbers sit near the top of the screen in Meta Ads Manager and Google Ads: the opportunity score and the optimization score. Both are out of 100, both go up when you do what the platform asks, and both read like a verdict on your marketing. Neither is one, and neither platform says it is.

What the two scores measure

Meta’s opportunity score runs on a 0–100 point scale and moves with how many of the recommendations in Ads Manager you apply, each ranked by the performance impact Meta estimates it would have. In Meta’s own example, fixing an audience fragmentation issue may add 35 points; applying an Advantage+ placements recommendation may add 5 ¹.

Google’s optimization score runs from 0% to 100% and is calculated in real time. It reads the statistics, settings and status of the account and its campaigns, adds the relevant impact of the recommendations available, and takes recent recommendations history into account. It is shown for active Search, Display, Video Action, App, Performance Max, Demand Gen and Shopping campaigns only ².

Both numbers answer one question, and answer it well: how closely does this account follow the platform’s current advice?

Meta is direct about the limit. Its help page says the score, high or low, tells you nothing about how the account is actually doing ¹.

The same page tells advertisers not to turn off campaigns with a low score, and to judge performance based on their business goals instead ¹. That is a fair thing for a platform to publish.

Who is holding the ruler

Here is the incentive problem, without the insinuation it usually comes with. Both scores are produced by the company selling you the media. That does not make them dishonest. Meta says the recommendations behind its score are experimentally proven to help improve performance ¹, and each platform publishes a plain description of what its number is measuring ².

It does mean the scale has a fixed centre of gravity: a recommendation exists because the platform believes it improves outcomes on that platform. So neither score can tell you that the channel is the wrong home for the budget, that two systems are claiming the same sale, or that the offer rather than the account is what is struggling.

A free score from another company can work the same way, agencies’ included. WordStream’s Google Ads grader is free, uses your Google Ads credentials to analyse the account, and grades it against 10 key metrics; the first question in its FAQ asks what to do if your grade is not great, and the answer routes to WordStream by LocaliQ’s own solutions ³. Troy gives a score away too, so apply the same test: ask what the scorer sells.

What an independent read checks instead

If the platform score answers “have you taken our advice”, the question left over is whether the money is working. Four things carry most of that answer, and no platform score can see them.

• Measurement. Whether the sales in your own records and the conversions in the ad account agree, and whether anything is counted twice.
• Creative. How many genuinely different ideas are live, rather than versions of one idea, and how long since anything new was tried.
• Targeting. What is spent on people who will never buy: overlapping audiences, places you do not serve, search terms you would never have bid on.
• Pacing. Whether the budget is spent when it works, and whether the account is edited so often that it never settles.

Three of the four need a second source — your own orders, your delivery area, your calendar — which is why no platform can supply them. All four are answerable at a score of 92 and at 41.

A check you can run this week

None of it needs a tool. Set aside an hour. Two steps lean on what the platforms document. Meta’s ad sets usually leave the learning phase after about 50 results in the week after the ad set’s last significant edit, and editing an ad, ad set or campaign during learning resets it; Meta advises against unnecessary edits and frequent budget changes ⁴. On Google, recommendations can be applied for you automatically: a setting available at account level only, which will not increase your budget, with the changes it applies recorded in the account’s history ⁵.

• Count last month’s orders or enquiries in your own records, and put that next to the conversions the ad account claims for the same month. Until you know why they differ, everything else sits on the gap.
• Count your edits in the change history for the past 30 days. If budgets are being changed most days, part of the budget is paying for restarts.
• On Google, check whether recommendations are being applied automatically, then read what was applied last month. If a change surprises you, that is where it came from.
• Group every live ad by idea, not by file. Four crops of one photograph is one idea, and one idea is not a test.
• Sort last month’s spend high to low and read the bottom half. Ads with real spend and no result are a decision you have not made yet.
• Check locations and search terms against where you actually sell. Then write down the one number the business needs this quarter: if neither score moves it, neither score is your dashboard.

When the recommendation is the right call

The opposite mistake is more expensive, and this article could cause it. Ignoring the platform because it sells you the media costs real money. Take the recommendation when it fixes something broken: a rejected ad, a policy error, a missing conversion event. Errors and warnings are part of what the opportunity score is built from ¹, and that part of the number is a list of faults.

Take it when it gives the system enough to learn from: fewer ads and ad sets rather than many, similar ad sets combined, and a realistic budget left alone long enough to settle ⁴. There your interests and the platform’s point the same way.

Be slower where the recommendation moves money or widens reach: broader targeting, new placements, larger budgets, campaign types that spend across channels you did not pick. Those can be right too, but only your own numbers can judge them — and if your measurement is not trustworthy, you cannot. If you use auto-apply, keep it to changes you would have approved by hand, and read the history once a month ⁵.

Where the Troy Score fits

The free Troy Score grades those four dimensions across every channel you connect, and shows the evidence behind every deduction. Troy sells no media, which removes one bias and not all of them: it sells help with the fixes, so the report carries the instructions to do them yourself. It does not predict results and never claims to. Read it next to your platform scores, not instead of them: one says whether you took the platform’s advice, the other whether that advice is worth taking.