Open a Google Ads account managed "the way it used to be done" and you almost always find the same architecture: a campaign per match type, one per region, one per service, and often the full cross-product of all three. It made sense when a person set bids keyword by keyword. Today a model sets them, and that model needs one thing above all to work: concentrated conversion data.
Why Smart Bidding rewards consolidation
Smart Bidding estimates, auction by auction, the probability that a click becomes a conversion and bids accordingly. It uses real-time signals — device, location, time, audience, query — but weighs them against the conversion history of the bidding entity: the campaign, or the set of campaigns sharing a portfolio strategy.
Every time you split a campaign in two, you split that history too. If an account generates 120 conversions a month across 20 campaigns, no single campaign has enough to tell a signal from a coincidence. The same 120 conversions across four campaigns give each one a base it can genuinely estimate from.
In early 2026, on Google's Ads Decoded podcast, the product lead for Search ads set out two markers. First: an indicative benchmark of around 15 conversions over 30 days per campaign. Second, and more important: consolidation is not the goal; the goal is equal or better performance with less granularity. Structure that serves real budget decisions, reporting needs or operational differences belongs. Structure that exists out of habit does not.
Our working threshold. Fifteen conversions in 30 days is a floor, not a comfort zone. Below 30-50 monthly conversions per bidding entity, a CPA or ROAS target swings too much week to week to manage with confidence. It is the same reasoning we apply to minimum budgets in how much budget Google Ads actually needs.
What no longer needs splitting
- Match types. Separate campaigns for exact, phrase and broad made sense for manual bid control. Under Smart Bidding, keywords of the same intent sit in the same ad group: the match type decides where you can appear, the model decides the bid.
- Regions of one country with the same economics. With Smart Bidding strategies, location bid adjustments are not applied: the system already uses location as a signal. Splitting North and South without a different target or budget just gives you two campaigns with half the data each.
- Devices. Same logic: device is an auction signal, not a reason to duplicate a campaign.
- Single-keyword ad groups. Ad groups are built around a theme and a landing page, not around a keyword.
When splitting is still mandatory
Our rule is simple: split when a decision changes, not when a label changes. Five decisions justify a campaign of their own.
| Reason | Why it needs its own campaign |
|---|---|
| Budget control | A business line needs guaranteed spend or a hard cap, regardless of where the model would find the cheapest conversions. |
| Different targets | A service with a sustainable CPA of £40 and one that can bear £150 cannot share a blended target without one of them being sacrificed. |
| Different margins | In ecommerce, categories with very different margins need different ROAS targets, or conversion values that reflect margin. |
| Country and language | Different markets mean different auctions, currencies, pages and ads. Mixing them makes both the target and the reporting meaningless. |
| Brand and non-brand | Brand searches convert very cheaply: inside a non-brand campaign they flatter the results and push the model to buy demand you would have had anyway. |
There is a sixth, less common case: different conversion goals. If one campaign should optimise towards quote requests and another towards webinar sign-ups, campaign-level goals need to stay separate, and so do the campaigns.
Splitting without fragmenting
When you need several campaigns for budget or reporting but volume is thin, there is a middle way: a portfolio bid strategy. Campaigns sharing it pool their conversion data for bidding while keeping budgets and reports separate. It only works if they genuinely share the same goal and target: a portfolio mixing different intents recreates the blended-target problem.
Search, Performance Max and Demand Gen in one account
Consolidation is not only about Search campaigns. In 2026 a typical account runs three families of campaigns, and the risk is less fragmentation than overlap.
Search: explicit intent
Search remains the place to hold the demand you know and want to control. The prioritisation rules matter: if the user's query is identical to an eligible exact match keyword in a Search campaign, that campaign takes priority over Performance Max. Phrase and broad match keywords, including those extended by AI Max, share priority with Performance Max search themes instead, and Ad Rank decides. If you want Search to own your most important queries, put them on exact match and do not let that campaign run budget-limited. We covered AI Max and its controls in AI Max for Search campaigns.
Performance Max: the catalogue and the full inventory
Performance Max earns its place mainly where there is a product feed, or an objective that benefits from Google's whole inventory. Two rules: set brand exclusions from day one, otherwise PMax harvests brand traffic and credits itself with conversions Search would have taken anyway; and split asset groups by real category or audience, not to multiply campaigns. The channel reporting that arrived this year finally shows where it spends: see Performance Max in 2026.
Demand Gen: creating demand, with a different yardstick
Demand Gen works across YouTube, Discover and Gmail to reach people who are not yet searching. It deserves its own campaign and budget precisely because it must be judged by a different yardstick: fewer direct conversions, more effect on brand demand and later searches. Lumping it in with last-click-optimised campaigns means switching it off after three weeks for the wrong reasons. When it actually pays is covered in Demand Gen in Google Ads.
A restructuring, worked through
Take an account built for this example, but with a very common shape: a services business selling three services in one country, 120 conversions a month in total. The inherited structure:
- 18 non-brand Search campaigns: 3 services × 3 match types × 2 regions;
- 1 brand campaign;
- 1 Performance Max campaign with no brand exclusions.
The 18 non-brand campaigns share around 85 conversions a month: under 5 each on average, many below 2. None reaches the minimum threshold. Every CPA target is a gamble.
Services A and B have similar economics (sustainable CPA around £60); service C is worth three times as much. The two regions have no different prices or targets. The new structure:
| Campaign | Contents | Expected conversions/month |
|---|---|---|
| Brand | Brand queries only, defensive target | ~25 |
| Non-brand A+B | Two services, ad groups by theme, all match types together, whole country | ~55 |
| Non-brand C | High-value service, own target, protected budget | ~30 |
| Performance Max | With brand exclusions, one asset group per service | remainder |
From 20 campaigns to 4. Both non-brand campaigns clear the minimum threshold, service C keeps its own target and budget because its economics differ, and brand stops subsidising the non-brand numbers.
How to execute it without burning a month
- Measurement first. Consolidating on the wrong conversions concentrates the error. Check goals and values before anything else.
- One family at a time. Start with the highest-volume service, wait for it to stabilise, then move on.
- Pause old campaigns, do not remove them. They stay as a historical reference and a way back.
- Realistic targets at launch. Start from the actual CPA of recent weeks, not the one you wish for, and tighten afterwards.
- No verdict for two or three weeks. A new setup goes through a learning phase: judging it after a few days means judging noise.
Warning signs of an over-consolidated account
The opposite mistake is just as expensive, and harder to spot because the aggregate numbers often look fine. The symptoms:
- One service eats the budget. The model goes where conversions are cheapest: if one line absorbs 80-90% of spend and the others vanish, the structure is setting your commercial priorities for you.
- Blended CPA is fine, per-line CPA is not. A blended target that meets the overall goal can hide a loss-making service propped up by a very profitable one.
- Brand inside non-brand. Excellent ROAS, zero growth: the system is harvesting demand that already existed.
- Generic ads and pages. Ad groups covering too many intents end up with copy that fits everything and convinces no one.
- The report stops answering. If you cannot say what you spend and earn on a line the business manages separately, you have consolidated too far.
The fix is not to go back to the old structure, but to pull out only what has a business reason: one more campaign for the starved line, a separate target for the loss-making service, brand out of non-brand.
If you want to know which structure your account's volumes can support, the free audit starts right here: conversions per bidding entity, overlap between Search and Performance Max, campaigns that exist only out of habit. You can also see the services that cover this work and our FAQ.