Google Ads account strategy review

How to review a Google Ads account as a whole, once the campaigns have had their own reviews. The companion Google Ads Campaign Review & Audit Guide (“the campaign guide”) owns everything inside one campaign: workstreams W1–W12, thresholds R1–R11, cross-checks X1–X21, the measurement gate (§7.1) and the list of proposable changes (§8.3). This guide does not repeat them. It covers the decisions that only show up when campaigns are put side by side.

BigData Boutique defaults: B2B lead gen, long sales cycle, low conversion volume, currency ILS. Rules marked [H] are internal rules of thumb; adjust them to the account’s own history. Facts marked [G] are Google’s (§5).

1. Principles

1. The account’s outcome beats any campaign’s. A change that improves one campaign’s CPA and lowers the account’s qualified conversions is a loss.
2. Compare like with like. Campaigns are comparable only on the same conversion definition (S6), the same window and the same maturity (lag-complete).
3. Decide on the margin. Budget goes where the next shekel earns most, not where the average CPA is lowest. Brand always has the best average and usually no room.
4. Role before efficiency. Brand defense, high-intent capture, competitor conquesting and demand creation are judged on different metrics (campaign guide W10).
5. Few, large, reversible moves. One structural change per campaign per conversion cycle (X12, X19). Uncertain direction means an experiment, not a switch (campaign guide §8.6).
6. Leave tactics to the campaign reviews. A negative keyword, an ad, a keyword bid: not here, unless the same fix is needed in several campaigns (S8, S9).

2. Inputs

The window is a rolling 30 days against the previous 30, lag-complete. When a decision needs more conversions than 30 days hold (R1, R7), pull 90 days with the read tools before deciding.

3. Checks

Severity as in the campaign guide: 🔴 critical · 🟠 material · 🟡 hygiene.

S1 Goals & unit economics

Question: is the account pointed at the business goals, and does every campaign have a role and a target that follows from them?

• Each goal maps to at least one campaign, and each campaign with material spend maps to a goal. Spend with no goal behind it 🟠.
• Max allowable CPA (LTV × close rate) is known per service line and the campaign targets sit under it. No economics on record → an open question for a person, and every allocation finding drops one confidence level.
• Account pacing vs the monthly envelope and vs the goal’s due date: at the current run rate, is the goal reachable? If not, say what would have to change (budget, CVR, CPA) and by how much.
• Campaign KPIs that contradict each other or the goal (one campaign held to CPL, another with the same audience to clicks) 🟡.

S2 Budget allocation across campaigns

Question: would moving budget between campaigns produce more qualified conversions for the same total?

• Read the shares. Spend share far above conversion share marks a donor candidate; the reverse marks a receiver candidate. Use qualified conversions (W12) when CRM data exists.
• Receiver only if all hold: CPA ≤ target with acceptable lead quality (X2, X16), lost impression share to budget ≥ 10–20%, not in learning (X12), marginal CPA ≤ max allowable.
• Donor only if all hold: enough data (R7), CPA > target after its own review’s fixes were given a cycle, no open change on it that could explain the weakness (X20), and it is not a role judged on something else (demand creation, brand defense).
• Marginal check: weekly spend against weekly conversions over 8–12 weeks, or the bid / budget simulators (campaign_simulation). A receiver whose conversions flattened as spend rose has no headroom.
• Brand is not a receiver once brand impression share is at R5: there is no more demand to buy. Its low CPA is not evidence that more budget works.
• Step size: ≤ 20% per campaign per conversion cycle (R4). Increases are funded by the cuts or fit the envelope (X18).
• Shared budgets [G]: campaigns on one shared budget already reallocate between themselves; check for a large campaign starving a small, efficient one, and for campaigns with different goals sharing one budget 🟠. Performance Max cannot use a shared budget.
• Budget-limited tCPA / tROAS campaigns bid to target since Aug 2026 (campaign guide W3, W4): re-baseline before calling one a donor.

Typical recommendation: a paired move — update_campaign_budget down on the donor, up on the receiver (two proposals, each stating its pair). Changing which campaigns share a budget is manual.

S3 Cannibalization & query routing

Question: are campaigns buying the same demand twice, or taking credit for each other’s?

• Brand leakage: brand terms with spend in non-brand, Performance Max or AI Max traffic. Non-brand campaigns need brand negatives; Performance Max needs brand exclusions (X10). A Performance Max CPA that looks excellent while brand impression share fell 🔴/🟠.
• Same term, several campaigns (overlaps): decide the owner by intent match and landing page, then by CPA / CVR with enough data (R7). Route with exact negatives in the others (X6: never negate a term that is an active keyword there; X5: never where it converts better).
• How Google picks [G]: a query identical to an eligible exact-match keyword in a Search campaign wins over broad or phrase keywords and over Performance Max. A query identical to a phrase or broad keyword shares priority with an identical Performance Max search theme. Ties are broken by Ad Rank. So overlap is expected wherever exact coverage is missing: the fix is often an exact keyword in the owner, not a negative.
• Structural duplicates: two campaigns with the same keywords and overlapping locations, languages or schedules. One should own each market 🟠.
• Act only on material overlap: combined overlapping spend above the materiality floor, or a measured CPA difference > 30% (R6) between owner and non-owner.

Typical recommendation: exact keyword in the owner + routing negatives (manage_ad_group_criterion, add_campaign_criterion); brand exclusions and brand lists are manual.

S4 Structure & consolidation

Question: does the structure give each bidding strategy enough data and each decision a clean unit?

• Fragmentation: several campaigns with the same goal, market and economics, each under R1 conversions per 30 days. Smart Bidding learns per strategy, so pool them: consolidate, or put them on one portfolio bid strategy (S7) 🟠.
• Over-consolidation: one campaign mixing brand and non-brand, markets with different economics, or two goals. Its average hides both a winner and a loser 🟠.
• Dead weight: enabled campaigns with under ~1% of spend and no conversions in 90 days [H] → pause or merge 🟡.
• Untracked campaigns (campaign_id null) with material spend: they get no weekly review. Recommend tracking them (Campaigns page) before judging the account 🟠.
• Naming and one owner per intent (role × service line × market) 🟡.
• A restructure resets learning. Recommend it only when the expected gain clearly exceeds a conversion cycle of disruption, staged or as an experiment.

S5 Campaign-type & funnel coverage

Question: does every goal run on the right campaign type, and is a role missing or doubled? Use the campaign guide’s W10 table (goal → type, prerequisites, red flags), at account level:

• Missing roles: no brand-only Search campaign while brand queries exist; no remarketing or demand creation while high-intent capture is not budget-limited and is at its ceiling; a high-intent service line with no coverage (google_ads_generate_keyword_ideas).
• Doubled roles: Performance Max and Search on the same intent without brand exclusions and exact coverage (S3); Display campaigns still running next to Demand Gen (migration, campaign guide Appendix B).
• Order of investment [H]: fix measurement → capture existing demand (brand, high intent) until it is rank- or demand-limited → only then create demand. Demand creation funded while capture campaigns lose impression share to budget 🟠.
• Spend by role vs the plan: the share going to capture, conquesting and creation, and its trend.

Typical recommendation: a manual task or an experiment; only a paused Search campaign on an existing shared budget is proposable (create_campaign, campaign guide W10).

S6 Measurement architecture

Question: do the campaigns count the same thing, so their numbers can be compared and their bidding learns from the same signal?

• Which goals each campaign bids to: account-default or campaign-specific (campaign_conversion_goal, customer_conversion_goal; check field names with google_ads_describe_gaql_resource). Google recommends account-default goals so campaigns learn from one another [G]. Campaign-specific goals need a written reason.
• Mixed definitions: one campaign optimizing to form fills and another to qualified leads makes their CPAs incomparable. Until fixed, compare them on one shared action (segments.conversion_action_name by campaign) and block S2 moves between them 🟠.
• Account gate: if the Campaigns carrying most of the spend report low measurement confidence (W1), the account’s is low: recommend only the measurement fix (campaign guide §7.1).
• Offline values and attribution (model, lookback) consistent across the actions used for bidding 🟡. Offline import feeding only some campaigns 🟠.

Conversion actions and goals are manual changes.

S7 Bidding architecture

Question: is each campaign on a strategy its volume supports, with targets that agree with each other and with the economics?

• Ladder fit per campaign (campaign guide W4) and across the account: campaigns too small for their target strategy that could pool data.
• Portfolio bid strategies [G] group campaigns under one strategy and target; available for Search, Display and Shopping, not Performance Max. Candidates: same goal, same max allowable CPA, each under R1 alone. Not for campaigns with different economics.
• Inconsistent targets: the same service and market with targets far apart (> 30%, R6) and no economic reason 🟠. Targets above max allowable CPA (S1) 🟠.
• Leftovers: Maximize Clicks or Manual CPC on lead-gen campaigns that have conversion data 🟠.
• Concurrent learning [H]: when campaigns carrying more than about a third of spend are learning or were changed within one conversion cycle, recommend a freeze instead of new moves.
• Experiments need an individual daily budget and a standard (non-portfolio) strategy [G]: plan tests before moving a campaign into a shared budget or portfolio.

Strategy switches and portfolios are manual; a target step on one campaign is proposable (update_campaign).

S8 Account-level controls

Question: are the protections that should apply everywhere applied everywhere, once?

• Negatives at the right level: the same irrelevant terms (campaign guide W6 tier T4) negated campaign by campaign, or leaking in some campaigns only → one shared negative keyword list attached to every Search campaign, or account-level negative keywords [G].
• Brand list / brand exclusions defined once and used by every campaign that needs them (S3).
• Audience exclusions (customers, job seekers) and placement exclusions applied consistently; Customer Match lists fresh.
• Settings decided once: location option, Search partners, Display expansion, auto-apply recommendations. The same question answered differently across similar campaigns 🟡/🟠.
• Google’s recommendations that several campaign reviews dismissed or applied in opposite ways: settle the policy.

Almost all of these are manual tasks (campaign guide §8.4).

S9 Recurring problems & execution health

Question: what do the campaign reviews, read together, say that none of them says alone?

• patterns.recurring_workstreams: the same workstream with findings in several Campaigns usually has one cause: a shared landing page (W8), a shared conversion action (W1), a missing shared negative list (W6), one bidding policy (W4). Recommend the root fix once.
• patterns.stalled_findings: open for three or more reviews. Find out why: waiting on a person, blocked by another fix, or not worth doing. Recommend the unblocking decision, or closing it.
• Approval backlog: proposals waiting more than two weeks, and their combined expected impact. If material, recommend a decision session; do not re-propose them.
• Hit rate: across the Campaigns’ follow-up verdicts, which kinds of change delivered and which did not. A kind that keeps failing should stop being proposed.
• Conflicts between campaign-level actions (one review scaling a campaign while another routes its queries away).

S10 Test & learning agenda

Question: what is the account trying to learn next, and is it set up to learn it?

• A short queue of tests ranked by expected value: brand incrementality, Performance Max vs Search, AI Max, broad match, a bidding strategy, a landing page.
• At most one experiment per campaign, and for low-volume accounts one or two at a time [H], each with a success metric agreed up front (ideally cost per SQL) and ≥ 4 weeks or one conversion cycle.
• Seasonality and known events ahead (launches, holidays) that should move budgets or pause tests.

Experiments are manual tasks.

4. Validating a candidate

5. Facts and rules of thumb used above

Google [G] (Google Ads Help):

Internal [H]: dead-weight campaign = under ~1% of spend and no conversions in 90 days · concurrent-learning freeze at about a third of spend · one or two experiments at a time on low-volume accounts · capture demand before creating it. Everything else is a campaign-guide threshold (R1–R11).