Account structure that still makes sense in a year

Here is the only test of an account structure that we have found useful. Open the campaign list. For each campaign, say out loud what it is for and what number would make you change it. Not how it is performing. What it is for.

Most accounts fail inside the first four campaigns. We have inherited accounts with forty of them, half paused, carrying names like “Search 2 Copy Final”, where nobody still at the company could explain what any of them had been built to answer. The campaigns were not performing badly. They were unreadable, which turned out to be the more expensive problem, because every decision anybody made about them was a guess.

Structure is a reporting decision now

Automated bidding took away most of the old reasons for granular structure. You do not need one keyword per ad group to control bids. The system does not want your help splitting traffic by device or by hour, and in most accounts it does that better than a person would.

What structure still does is answer questions. Every boundary you draw is a line in a future report. If you need to know whether roof repair earns more than roof replacement, those are two campaigns. If you will never need to know, they are one. Spend should be divisible along the lines of decisions you might actually take.

Which has an uncomfortable corollary for anyone who likes a tidy account. A split you will never act on is not organisation. It is fragmentation, and you pay for it in conversion data thinned across every bucket you created.

The splits that earn their keep

You probably already separate some of these, so skim the list and look for the one you do not.

Brand versus non-brand. Always separate. Blending them makes cold acquisition look profitable when it is not, and it is the most common source of misleading reporting we come across.

Profit lines that genuinely differ. Two services with very different margins or very different sales cycles need separate budgets, or the bidding will favour whichever converts most easily, which is usually the one that earns least.

Geography you would ever budget differently. Separate countries, yes. Separate districts of Varna, almost never.

Languages. Always separate campaigns. Mixed-language ad groups produce ads nobody wants to click, and for Bulgarian businesses selling to both local and expat markets this is the split most often missed.

Campaign type. Search, Shopping, Demand Gen and Performance Max do different jobs. This list changed in 2026. Google is retiring standalone Display campaigns and folding the Display Network into Demand Gen, with the migration tool rolling out from June 2026 and the transition expected to complete by 2027. The inventory is not going anywhere. What changed is that the Display Network went from being the default surface of its own campaign type to an opt-in channel inside Demand Gen alongside YouTube, Discover, Gmail and Maps. If you had a Display campaign quietly working, check that the channel is still switched on after migration rather than assuming it carried over.

Prospecting versus remarketing. Different job, different economics, different definition of a good month.

Building something you can still read in a year

1. Write the questions before the campaigns. List the decisions you expect to face. Should we spend more on service A. Is the second city worth it. Is remarketing paying for itself. Each question that needs isolated spend becomes a boundary. The rest do not.
2. Pick a naming convention and never break it. Type, intent, product, market. “Search Brand BG”, “Search NonBrand Roofing BG”, “PMax Retail EU”. Boring and consistent beats clever, and it is free.
3. Keep ad groups few and tight. The test is whether you can write one honest headline that fits every keyword in the group. If you can, leave it alone. If you cannot, split it.
4. Write one page outside the platform. What each campaign is for, its target cost per lead, its primary conversion action, and why. Update it when something changes. That page has rescued more accounts for us than any clever optimisation, usually at the moment a staff member leaves.

Delete rather than pause, too. Old paused campaigns are noise in the list you are trying to read. Keep the notes, not the clutter.

Where the rules bend

Small budgets should be flatter than any of the above implies. On a modest monthly spend, one Search campaign for brand and one for everything else is often correct, and further division only starves the bidding.

Large accounts go the other way. Once a single product line spends serious money on its own, splitting by sub-category is legitimate, because each piece still has signal and somebody in the business genuinely owns each budget.

Ecommerce with thousands of SKUs runs on different logic again, organised around margin bands and product performance rather than themes. The principle holds: split where you would take a different action, not where the catalogue happens to have a category page. We covered the broader divide in ecommerce versus lead generation accounts.

The part we have not resolved

Everything above rests on a trade-off we are still not comfortable with, and it would be dishonest to present it as settled.

Readability and signal pull against each other. The structure that lets you answer “is service A worth more than service B” is the structure that divides your conversions into pools too small for the bidding to learn from. Google’s own guidance on evaluating a target is to judge it over a period containing at least thirty conversions, which a business closing six jobs a month will not reach inside a quarter in any single campaign, let alone two. So the split that makes the account legible is often the split that makes it perform worse. Merge them back and the account bids well and tells you nothing.

Our bias is towards readability, and we will defend it: an account you understand produces better decisions over years than an account that is theoretically optimal and opaque, because the decisions that matter most are made by owners rather than by algorithms. We also know that bias costs some clients measurable performance in the first few months, and we have had the argument with people who were right to raise it. We have not found a rule that resolves it, only a judgement we make case by case, usually on the question of who will be reading the account and how often. If that is a judgement you would rather make with someone who has made it before, that is what a consultation is for.

So take the test at the top with that caveat attached. Some of the campaigns you cannot explain should be merged. Some of them should be explained instead.