Account structure

Updated November 12, 2026

How AdsGuru treats your structure

AdsGuru does not force a rebuild. It optimizes within the structure you have, and raises structural proposals only when the structure itself is measurably costing money: ad sets starved below learning volume, campaigns bidding against each other, Performance Max eating brand Search.

This is a deliberate stance. Restructuring resets learning, interrupts delivery and rewrites history baselines, so it has real costs that only clear evidence justifies. Tools that begin by demanding their preferred structure are optimizing for their templates; the account's job is the business's results, and plenty of imperfect structures perform fine.

When a structural proposal does arrive, it carries the evidence, the expected gain, the transition costs stated plainly, and a migration plan. It waits for your approval, always; structure is the one category that never runs on autopilot.

The practical consequence for onboarding: nothing about connecting requires preparing the account first. Connect what exists, let the audit describe it, and decide on structure from evidence rather than from a migration checklist written for a different business.

Why structure matters at all

Structure decides two things: where budget decisions get made, and how concentrated your data is. Modern bidding on both platforms is machine learning, and the machine learns per optimization unit, per ad set on Meta, per campaign or bid strategy on Google. Spread your conversions across twenty units and each one learns from a trickle; concentrate them in four and each learns from a stream.

This is why the platforms themselves have pushed consolidation for years: fewer, bigger units calibrate faster and perform more stably. Meta restarts an ad set's learning phase after significant edits, and ad sets that never accumulate enough conversion volume live in learning limbo indefinitely, with the erratic delivery that implies.

Structure also decides what you can control separately: separate campaigns exist so things can have separate budgets, targets or creative. Every split should answer the question 'what decision will I make differently because this is separate?' If no answer, the split just dilutes data.

Google Ads structure in brief

A Google account splits into campaigns (budget, bid strategy, settings level) and ad groups (keyword and ad grouping). The canonical shape: brand Search kept separate from non-brand, because brand clicks are cheap, convert at your site's natural rate, and inflate any campaign they share; Shopping or Performance Max for product feeds; and non-brand Search organized by intent theme, tightly enough that ads match queries, loosely enough that data pools.

The modern failure modes are over-segmentation left over from older eras (single-keyword ad groups fragmenting everything) and PMax overlap: Performance Max can serve on queries your Search campaigns bid on, and will serve on brand unless excluded, booking demand that existed anyway. Brand exclusions and clean separation of duties are the fix; the negative keywords page covers the fencing tools.

The audit checks these specifically: fragmented ad groups below data thresholds, overlapping keyword coverage, PMax-versus-Search cannibalization, and campaigns optimizing toward the wrong conversion action.

Meta structure in brief

A Meta account splits into campaigns (objective and, with CBO, budget), ad sets (audience, placement, optimization event) and ads. The modern consolidated shape is few ad sets with broad audiences, letting the delivery system do targeting work through creative, with budgets at whichever level your CBO/ABO choice puts them.

The era of many narrowly-sliced interest ad sets is over for most accounts: slices fragment conversion data, audiences overlap anyway (your 'fitness interests' and 'wellness interests' ad sets bid on the same humans), and each slice sits in learning forever. Advantage+ shopping campaigns take consolidation to the extreme, one campaign absorbing most prospecting.

What still deserves separation: prospecting versus retargeting (different message, different economics), genuinely distinct products or geographies with different targets, and structured creative testing where a controlled split is the point.

Signs your structure is costing money

• Ad sets stuck in learning for weeks: each is starved below the conversion volume that lets delivery calibrate.
• Many units, tiny budgets: twenty campaigns sharing what four could concentrate; nothing ever accumulates evidence.
• Auction overlap: multiple campaigns or ad sets eligible for the same query or audience, bidding against yourself.
• PMax booking brand: Performance Max conversions concentrated on brand queries, while brand Search impression share falls.
• Mixed intents in one unit: brand with non-brand, prospecting with retargeting; the average hides two different businesses.
• Splits with no decisions attached: segments that exist for reporting curiosity, paid for with fragmented learning.
• Orphans and leftovers: empty ad groups, paused experiments half-cleaned, campaigns optimizing to deprecated conversion actions.

Any of these can be fine in a specific context, which is why they are flagged as evidence-backed findings rather than auto-fixed.

Consolidate or segment: the decision

The test compresses to one sentence: separate things that need separate decisions, pool everything else. Most accounts err on the side of too many splits, because splits feel like control and their cost, diluted learning, is invisible on any single day.

What a restructuring proposal looks like

When the evidence clears the bar, the proposal lays out: the finding (these six ad sets averaged too few conversions each to exit learning over the window), the evidence (volumes, overlap data, impression share), the proposed end state (consolidate to two ad sets, budgets preserved), the transition costs stated honestly (learning will reset on the consolidated sets; expect erratic days before stabilization), and the staged plan, since moving everything at once maximizes disruption.

You approve, edit or reject; approving applies the stage and logs it. The honest note here: structural wins are real but not instant, and the week after a consolidation often looks worse before it looks better, because learning restarts. The proposal says that out loud so nobody judges the surgery by the first bandage.

Staging also creates the measurement: each stage gets its own before and after in the change log, so the restructure's value is evaluated from data rather than argued from memory at the next review.

Review your structure manually

1. Draw the account on one page: every active campaign, its budget, target, and one-line purpose. Difficulty producing the one-liner is itself a finding.
2. On Meta, check each ad set's learning status and weekly conversion volume; list everything delivering erratically or stuck in learning.
3. On Google, check brand and non-brand separation, and pull a search terms overlap check: the same queries appearing under multiple campaigns means auction overlap.
4. If you run PMax, compare its converting queries (as exposed) against brand terms, and check brand Search impression share trend.
5. For every split, write the decision it enables. No decision: candidate for pooling.
6. Sequence any changes in stages, one structural change at a time, with a stabilization window between.

This is also roughly the checklist behind the audit's structure pass, so an audit on connect gives you the findings without the afternoon.

Common questions

Related