Structure follows decisions, not platforms
The most common structural fault is an account built to mirror the platform's menus rather than the business. A campaign per ad format, an ad set per interest, a campaign per keyword theme because that was how the account was first set up. None of these splits corresponds to a question the business asks, so the reports they produce answer nothing useful, and the budget cannot be moved between them on any principled basis.
Start instead from the decisions. Which budgets need to be set independently? Which results are worth different amounts, and so need different cost per acquisition (CPACost of each conversion.) or return (ROASRevenue generated per unit of ad spend.) targets? Which reports will a review actually use to move money? A split earns its place only when it serves one of those three purposes. The targets come first, from the performance economics guide, because the granularity of the targets decides the granularity of the structure: if two product lines share a margin and a target, they can share a campaign; if they do not, they cannot.
This principle survives platform changes. Interfaces rename campaign types, add automation and retire options regularly, as the platforms' help centres record. An account built around decisions can absorb those changes; an account built around a menu has to be rebuilt each time the menu changes.
Splitting by objective, stage, product and market
Four dimensions justify a split in most accounts. Each comes with a test: split only if the answer to the question in the second column is yes.
| Dimension | Split when… | Typical reason | Cost of splitting |
|---|---|---|---|
| Objective | The result optimised for is different (sales, leads, installs, reach) | Each objective needs its own conversion event and bidding signal | Low; mixing objectives confuses bidding |
| Funnel stage | Prospecting and RemarketingAdvertising to previous visitors or engagers. need different caps, targets or judgement | Remarketing reports flattering returns and must be capped and tested separately | Moderate; smaller remarketing pools learn slowly |
| Product or category | Margins, prices or targets differ materially | A single target would over-fund low-margin lines and starve high-margin ones | Moderate to high; splits the conversion data |
| Market or language | Currency, price, competition, regulation or creative differ | Budgets are set per market and reports must show each | High in small markets; low data per campaign |
The marketing funnel framework supplies the stage vocabulary, and it is useful for deciding which stages need their own campaigns. In practice most performance accounts need at most two stage layers: one that finds new people and one that re-engages people who have already shown interest. Finer stage splits rarely change a budget decision and usually just divide the data.
Splits can be nested, and the order matters. Objective is almost always the top split, because mixing objectives in one campaign is the surest way to confuse bidding. Market usually comes next where budgets are set per country. Product and stage then sit inside, and where the platform allows, they are better expressed as groupings within a campaign than as separate campaigns.
Consolidation vs control
Automated bidding on every major platform learns from conversions. Google's and Meta's help centres both describe a learning period after a campaign launches or changes materially, and both advise that the system needs a steady flow of conversion data to stabilise. Each split divides that flow. An account with many small campaigns gives each one too little data, keeps them in learning, and makes their results noisy enough that nobody can tell good from bad.
Consolidation is the response: fewer campaigns, broader targeting, more creative per campaign, and let the system allocate within it. It usually improves efficiency. It also hands decisions to the platform that the business may want to keep, such as how much goes to existing customers, which products get promoted and which markets get priority.
The balance is struck campaign by campaign:
- Consolidate where the results are worth the same, the target is the same and the business is content to let the platform allocate within the group.
- Keep control where the business needs a different target, a hard budget, a cap or a clean report, and accept the data cost knowingly.
- Use reporting dimensions instead of campaigns where the platform can break results down by product, placement or audience without a structural split. Many questions that seem to need a campaign only need a report.
When an account is stuck in learning, the fix is nearly always consolidation. When a review cannot answer which market or product is profitable, the fix is nearly always a split, or a reporting dimension that serves the same purpose. On paid social, a paid social campaign build sheet records each consolidation or split with its reason before launch, so the trade-off is decided on paper rather than discovered in the account.
Naming conventions and tracking parameters
A structure only works if people can read it. A naming conventionAn agreed pattern for naming campaigns, ads, events and tracking parameters so that reports can group them reliably. encodes the splits in each campaign's name in a fixed order, so that anyone, and any report filter, can tell what a campaign is without opening it. The pattern is usually the dimensions from the table above in the order they nest: objective, market, stage, product, then a free-text descriptor. Separators and abbreviations are agreed once and written down.
The same structure needs to reach the analytics tool. Tracking parameters such as UTMURL tags identifying traffic source. tags carry source, medium and campaign into analytics, and the campaign value should repeat the platform name exactly, or a deterministic short form of it, so that platform and analytics reports join cleanly. Platforms with auto-tagging pass their own identifiers, which is helpful inside their ecosystem and does not replace a consistent convention across all of them. The naming conventions and UTMs guide in the analytics root sets out the full rules; the performance marketer's job is to apply them to every campaign before launch, not to clean them up afterwards.
A naming convention is also a record of intent. When a campaign's name says it is a prospecting campaign for one market and one category, a report showing it spending on existing customers in another market is an anomaly that somebody will notice.
Budgets and targets per structure
Each structural unit should carry three things: a budget, a target and a key performance indicator (KPIA metric tied directly to an objective.) that the review will read. If a campaign cannot be given all three, it probably should not be a separate campaign.
Budgets follow the decisions that created the split. Where markets are separate, each market gets its own budget set by the plan, and the platform is not allowed to move money between them. Where products are split by margin, the higher-margin group can be given more room on cost per result rather than more budget, letting bidding find the volume. Shared budgets and portfolio strategies, which the help centres document for several platforms, are useful within a group of campaigns that share a target, and harmful across groups that should not.
Targets come from the economics, adjusted for the role of each unit. A remarketing campaign should carry a stricter target than prospecting, because much of what it reports would have happened anyway. A new-market campaign may run looser for a defined learning period. Write the target next to the campaign in the plan so that changes to it are deliberate and visible.
A bike parts retailer collapses forty campaigns into nine
Sprocket Supply sells bicycle components in three countries and had 41 campaigns across Google and Meta, split by
ad format, keyword theme and interest group. Most saw fewer than 10 conversions a week and sat permanently in
learning. The team mapped decisions first: budgets were set per country, and two margin tiers (drivetrain parts at
38% margin, accessories at 21%) needed different ROAS targets. The new structure had nine campaigns: per country,
one prospecting campaign per margin tier and one capped remarketing campaign, with names following
objective_country_stage_tier. Ad formats and keyword themes became groupings inside campaigns. Within two months
every campaign was out of learning, weekly reports showed profit by country and tier, and the review moved budget
from accessories in one market to drivetrain parts in another.
What differs by platform
The principles above hold everywhere. The vocabulary, the hierarchy and the degree of automation differ, and each platform root owns its own structure guide.
| Platform | Hierarchy in brief | Where the difference bites | Platform guide |
|---|---|---|---|
| Google Ads | Account, campaign, ad group, with campaign types that bundle inventory | Campaign types decide inventory and automation; search keyword grouping matters, other types are largely automated | Google Ads account structure |
| Meta Ads | Campaign, ad set, ad, with budgets at campaign or ad set level | Audience and placement decisions sit at ad set level; consolidation and broad audiences are favoured | Meta Ads campaign structure |
Programmatic (DV360) | Advertiser, campaign, insertion order, line item | Budgets and pacing live on insertion orders; line items carry targeting and bidding | Display & Video account structure |
Use the table to translate. A split that this guide calls a campaign may be an ad set on Meta or an insertion order
in DV360; the question of whether to split is the same on all three. The
fundamentals hub sets structure alongside economics and tracking.
Common mistakes
- Copying the platform template. A structure built around formats or menus produces reports nobody can act on; start from budgets, targets and reports.
- Over-splitting for control. Each split divides conversion data; an account permanently in learning is usually an account with too many campaigns.
- Mixing objectives in one campaign. Bidding cannot optimise for sales and leads at once; objective is the first split, not the last.
- Leaving remarketing blended and unreported. If it cannot be capped or reported separately, it cannot be tested, and it will claim results it did not cause.
- Splitting where a report would do. Many questions about product, placement or audience can be answered with a breakdown rather than a new campaign.
- Inconsistent names and tags. Campaign names that do not match tracking parameters make platform and analytics reports impossible to join.
- Restructuring everything at once. A full rebuild resets learning everywhere and destroys comparability; change in stages and read each one.
