Our Google Rep Only Recommends Higher Budgets and Targets. What Evidence Should We Ask For?

Two kinds of advice arrive on that call, and they need two answers

One comes with a test you can run in your own account. The other comes with an estimate you cannot recompute. Profit Forensics holds every recommendation to one bar: take the first, and park the second until somebody shows you how the number was built.

The advice splits by type: budget, broad match, Performance Max, AI Max, Search Ads 360, and the claim that consent mode explains lower-than-expected conversion counts. Each merits a different check against data from your own account. This article is filed under Account Audits & Agency Reviews, because the decision is whether the account supports the change.

The score grades your settings, and dismissing advice lifts it

A score near the ceiling does not tell you the account makes money. Google’s page on optimisation score calls it “an estimate of how well your Google Ads account is set to perform.” Scores run from 0% to 100%, “with 100% meaning that your account can perform at its full potential.”

The page calculates the score in real time from statistics, settings, status, the impact of available recommendations, and recent recommendation history. Applying or dismissing recommendations changes it, and the account “can reach an optimization score of 100% by applying or dismissing all recommendations.”

Google also says the score “is not used by your Quality Score.” The English page checked for this article does not say the score is excluded from ad rank, and we do not claim that exclusion. The page does not say the score measures profit, lead quality, incremental sales, or which campaign brings in the money.

Inside one account the link to results was flat

In our published comparison, the median rank correlation between the score and click-through rate, inside an account, was 0.05. That is no measurable link. For conversion rate the median was -0.05. The middle of the click-through correlations ran from -0.22 to 0.21. Remove any single account from that analysis and the median stays at 0.05, so the result is not driven by a single outlier account. That comparison is published as what the optimisation score actually ranks inside an account. It is not a profit test, and it is not a reading of the period after 17 August 2026.

The score also moves when you dismiss a recommendation, because dismissal takes it out of the pool. A number that rises either way is a configuration reading, not a report card.

A separate review finding points the same way. The score can hit 99% while an account still wastes about 19% of spend on zero-conversion terms and pays a 14% cost-per-click premium over the vertical median: an account scoring 49% in the same niche wastes about half as much (9%, with a cost per click 64% below that median). Hold the score where the business result is acceptable, and ask for the screen behind it.

A higher budget is a request for more spend

Read a recommendation to raise the budget as a request for more spend, until your own account shows the return. The figure in the interface often looks like a stretch from current reach towards full reach. Ask which current reach the recommended budget was scaled from, and ask to see that reach. A forecast is not that screen.

From 17 August 2026 eligible campaigns move towards your target

On an eligible target-based campaign (one aimed at a cost per acquisition or a return on ad spend) that is limited by budget, the extra spend moves towards the target you set. Google’s page on changes to target based bid strategies says that starting on 17 August 2026, bidding was updated “to deliver more consistent and predictable performance based on the targets you set, even when you make budget adjustments.” The same page says “Google won’t automatically adjust your bidding targets or budgets.” The documented next step is the forecast of additional conversions at your target. The Eligibility section covers Search, Shopping, Performance Max, Demand Gen and Travel; Hotel and Display already use this behaviour. App, Video reach and Video view campaigns keep their previous bidding behaviour.

That forecast is an estimate from historical data. Google’s page on the Results tab says actual results can move with the market, with competition in the auction, and with other changes you make. Google analyses performance a week after you apply a bid or budget recommendation, then compares it with Google’s own estimate of the alternative. Impact is “a 7-day rolling average over a 28-day period.” Impact results for an eligible recommendation are available after it is applied; they are not available beforehand. You cannot recompute Google’s estimate of what would have happened without the change. Keep your own before-and-after on the metric the business uses.

If an eligible budget-limited campaign is already beating its target, the extra budget moves delivery towards the target rather than preserving the over-performance. Whether a limited-by-budget label is a reason to spend more is a separate question, answered in our note on that status.

The current auto-apply set does not raise the budget

In the current set, budget raising is absent. Google’s page on applying recommendations automatically says “Auto-applying recommendations won’t increase your budget”, and “Budget raising isn’t included in this iteration of auto-apply recommendations.”

Those auto-apply settings can still change a cost target and add broad match. The page lists “Adjust your CPA targets”, “Adjust your ROAS targets”, “Set a target CPA”, “Set a target ROAS”, and “Add broad match keywords.” Those switches are separate from the 17 August 2026 bidding change, which does not itself move your target or your budget. Read together, the pages allow a sequence: a target moves under a setting you turned on, and a later budget increase spends towards the target now on the campaign. We have not counted how often.

In our window, rank took far more impressions than budget did

Budget was the smaller loss overall. From January 2024 to July 2026, snapshot 10 July 2026, budget cost more of the lost impression share than rank did in only 14.7% of Search campaign-months. The median campaign-month lost 43.18% of eligible time to rank and 1.24% to budget. Those are separate impression-share medians, not one campaign, not hours or money, and not a measurement after 17 August 2026. The published comparison is where lost impression share actually goes. Our study does not test whether a larger budget recovers impressions lost to rank. Check the budget column and the rank column separately.

Broad match: ask for the cost, not the win rate

Ask for cost per recorded conversion in this account. A win rate on a slide is a different metric. Adding broad match keywords sits on the auto-apply list, so the change can land without a meeting and without a budget increase.

In the original sample of our published comparison, broad match cost 6.4% more per recorded conversion than exact match at the median account, and broad was more expensive in two thirds of those accounts. The other accounts in that same sample went the other way. That 6.4% is the original window of the study, not a rule for yours, and not a result from after 17 August 2026. Later samples on the same page, including a recheck in October 2026, put the median cost at 1.16 times the exact-match cost and at 1.03 times that cost, and in one slice the conversion rate at 0.667 times. Those ratios are further measurements, not an estimate of the August change. The published comparison is how much more broad match costs per recorded conversion.

Ask for the new search terms and the cost per recorded conversion on the same products. Act on an extreme outlier, not on a small average gain that would put already-converting queries at risk.

Performance Max: one window is not a direction

Quote both windows, or quote neither. In the original sample, Performance Max came in more than 10% above its target cost per acquisition in 71.3% of campaigns with an explicit target, and Search in 26.3%. A later, smaller sample, January 2026 to May 2026, reversed the shares more than 10% above target CPA to 27.7% and 45.5%. The later slice is a second measurement, not a trend. Advertisers chose the campaign types, so the gap may reflect the accounts, not the product. Do not treat either pair as the rule for your account, and do not read 71.3% as a result from after 17 August 2026. The published comparison is whether Performance Max hits its target cost per acquisition.

Ask for this account’s target cost, this account’s actual cost, and the date you will judge it on. One anonymous marketer told Digiday there were bonuses for getting clients onto Performance Max: “There were bonuses for the staff who could get their clients on PMax.” That is Digiday’s report of 13 August 2025, one unnamed source. We have not verified that claim independently. Help pages do not describe that pay. A bonus is a reason to ask for the test, not a finding that the campaign type fails.

AI Max: ask for a test before committing a live budget

Do not turn on a change you have not tested only because a rep recommends it. My caution comes from practitioner experience, not a portfolio measurement. I want to see evidence from the account before committing a live budget, especially where the proposal could disturb traffic that already converts.

The help pages checked for this article do not give AI Max a pass or a ban. Ask which queries it will add, how brand queries stay out, and which result turns it off. If the change would put traffic that already converts at risk, for a small average gain, wait until the outlier is extreme enough to justify the risk.

Search Ads 360: ask which job the current account cannot do

Ask for the job the current Google Ads account cannot do. Nothing on the help pages checked for this article sets a spend level at which Search Ads 360 becomes necessary. A recommendation to move is a recommendation to buy a different product, and it takes the same test.

Ask, in writing: which jobs the current account cannot do, which of those you actually use, who pays, what happens to bidding and to conversion actions on the day of the move, and how you leave. A price conversation answers none of those.

Consent mode: check the tag before you touch the budget

Check the tag in Tag Assistant before you raise the budget to fill a gap you have not seen. Google’s page on troubleshooting consent mode says consent mode “needs to be implemented for at least 7 full days and meet certain data thresholds until you see impact results in Google Ads.” The check is Tag Assistant. An empty Consent tab means the mode is not implemented, or the tag is blocked.

That page does not establish consent mode as the cause when organic and paid fall together. “Missing conversions” stays an estimate until someone names the earlier conversion count, the conversion action, and the cause: a blocked tag, a missing consent signal, or a model. Ask for the Tag Assistant check, and leave the budget alone until that trace exists.

What Google has not said

The help pages checked here do not say what representatives are paid, and they do not say a recommendation is wrong because a salesperson brought it. Search Engine Land quoted a Google spokesperson, who said: “The employees making those calls are members of the sales team, not support.” The sales team “has competitive targets and is rewarded with performance-related commissions.” Support, also called gTech, troubleshoots product issues. The report’s date in search is 23 January 2024. The article body did not load on 12 November 2026, so the wording comes from search results for that address on the same day. We do not state a commission formula.

Change history is not a complete record of what the caller changed. Google’s page on change history keeps changes “during the past 2 years“, and sometimes omits account-level settings and “changes made by Google representatives during consultations.” An empty log does not prove the rep changed nothing. A full log is not a full list of those edits. Auto-applied items leave a different trail: Google’s auto-apply page sends you to change history for the list of what was applied, and the switches themselves live on the auto-apply management page.

The reply you can send back

One paragraph covers every advice type above. Send it before the meeting rather than after it, and ask for the answers in writing.

“Before we apply anything, please show it in our own account: the campaign, the target, the actual cost, the date we will judge it on, and the reach figure behind any budget number. For broad match, the cost per recorded conversion for broad and for exact on the same products. For AI Max, the queries it will add and the result that turns it off. For measurement, the conversion action, the earlier conversion count, and the cause.”

Then check the answers against three numbers you own: your target against your actual cost, impression share lost to budget against lost to rank, and cost per recorded conversion on broad against exact. A recommendation that cannot be checked against the metric that applies to it stays a proposal. If the representative and your agency read the same screen differently, ask both to cite that screen before the budget moves.

What to check in your account

1. Write the recommendation down before anyone applies it. Open Recommendations. Record the type: budget, a cost or return-on-ad-spend target, broad match, AI Max, a product move, or a measurement warning.
2. Put the target next to the actual. Open the campaign. Note whether its type is eligible under the linked Eligibility section, whether it is target-based and limited by budget, and whether today is after 17 August 2026. For an eligible campaign, if results are already better than the target, a larger budget spends towards the target.
3. Split lost impression share. In Campaigns, add the columns for impression share lost to budget and lost to rank. The published 14.7% is a reason to look, not your result.
4. Open auto-apply from Recommendations. Confirm budget raising is absent. Then check cost-per-acquisition targets, return-on-ad-spend targets, and broad match keywords. The management page is where those switches live.
5. Read the impact results after an eligible bid or budget recommendation has been applied. The comparison is Google’s estimate, a 7-day rolling average over 28 days.
6. For a consent claim, open Tag Assistant and verify consent mode. Read impact only after at least 7 full days, and only if the thresholds are met. An empty Consent tab means the mode is missing or the tag is blocked.
7. Open change history from the Campaigns menu. It covers the past 2 years. An empty stretch does not prove the rep changed nothing. The auto-apply page sends you here for the list of what was applied.

When this does not apply

The call is about a policy violation, not a growth idea. Ask Google to remove the violation before the working conversation. A budget discussion while the flag is still on the account mixes two decisions. This sequence is for a policy call only.

The campaign is an eligible type, target-based, limited by budget, already at a target you accept, and lost impression share says budget is the constraint. From 17 August 2026, a budget increase on such a campaign moves delivery towards the target you set. If you have decided to buy that demand, refusing the increase leaves it unmet. The increase may be justified when the actual result is acceptable and lost impression share shows budget, not rank, as the constraint.

Tag Assistant already shows a missing or blocked consent implementation, and the only change on the table is the tag. A confirmed tag fault is a fix you can see. A request to raise the budget is still separate and still needs the budget evidence above.

When to call an independent check

Use a third reading when the recommendation conflicts with the figures in the account, or when the two readings cannot both be right. When the rep’s estimate and your agency’s view disagree, a Profit Forensics verdict is a third opinion with no stake in either. Bring the recommendation text, target against actual, the two lost-impression-share columns, the auto-apply switches, and the Tag Assistant view if measurement is in dispute.

The same call fits when a board is about to apply a stack of recommendations to move a score. The score moves when you apply advice and when you dismiss it. The review names the screen that would have to move before the spend does.

Researcher’s take

I treat the caller as a salesperson with good data, and I still take the meeting when a wrong target would be expensive. In the accounts I examine, the recommendation that survives is the one a colleague can rerun without the caller in the room. A score that rises when you dismiss the advice does not survive that test. Neither does a forecast you cannot recompute. If the estimate and your agency disagree, get a third reading before the budget moves.

Igor Ivitskiy, Doctor Ads Profit Forensics

FAQ

Should we follow our Google rep’s recommendations?

Follow a recommendation that arrives as a test you can run in your account. Decline an estimate you cannot recompute. Inside an account, the score’s median link with click-through rate was 0.05, and dismissing a recommendation lifts the score too.

Our chief executive asks: can’t we just call our Google rep?

A call does not produce a documented figure the board can audit. The forecast is a historical estimate. Impact results are available after you apply an eligible bid or budget recommendation. Change history sometimes omits consultation edits.

The rep says we are missing conversions. Is that estimate real?

Treat it as an estimate until someone names the conversion action, the baseline, and the cause. Consent mode shows impact only after at least 7 full days and once data thresholds are met. Check it in Tag Assistant. A drop in organic and paid together is not that page’s diagnosis.

Are Google reps paid on the recommendations we apply?

The help pages checked for this article do not say how representatives are paid, and publish no commission formula. Search Engine Land quoted a Google spokesperson describing a sales team with competitive targets and performance-related commissions, not support. That report does not make the recommendation false. Ask for the test either way.

Why does our rep keep pushing Performance Max, broad match and higher budgets?

They are separate proposals. After 17 August 2026, a higher budget spends an eligible budget-limited target-based campaign towards the target you set. Broad match needs your own cost comparison. Performance Max was more than 10% above its target CPA in 71.3% of campaigns in one sample and 27.7% in a later one.

How we know

Google Ads Help, checked 12 November 2026: About optimization score; About the Results tab in Google Ads Recommendations; About applying recommendations automatically; Manage auto-apply recommendations; Changes to target based bid strategies; Troubleshoot consent mode; About change history.

Secondary reports, not help pages. Search Engine Land, sales and support clarification, 23 January 2024, wording from search results for that address because the body did not load on 12 November 2026. Digiday, Confessions of a marketer on pulling every client off Google’s PMax, 13 August 2025, one anonymous marketer. Reader questions in forums and search are the wording behind the questions above, and the threads are retold without names. They show that people ask. They are not the answer.

No new portfolio data in this article: it combines Google’s documentation (checked 12 November 2026) with what our audits look for. Where a figure comes from our published studies, the study is linked.

The 99% score beside about 19% wasted spend is one review finding, not a second study and not a profit test. The judgement is mine: accept a test you can rerun, treat a budget recommendation as a request for more spend, and leave an untested proposal alone until the account supports it. Profit Forensics reads those screens when the rep and the agency disagree.