An account with 140 campaigns, most spending under a thousand rupees a week, is not a well-organised account. It is a machine learning system starved of data, spread so thin that no campaign ever accumulates enough conversions to optimise against. The structure that felt organised on a whiteboard is the reason performance is flat.
Account structure has changed direction in recent years. The old instinct was to segment relentlessly. The current reality is that automated bidding needs volume, & over-segmentation destroys volume. But consolidation has its own failure mode: you lose the ability to control budget where it matters. The job is finding the line.
Structure serves three purposes only
Every structural decision should trace back to one of these. If it does not, it is organisation for its own sake.
- Budget control. Budgets are set at campaign level. Anything you need to fund separately must be a separate campaign. That is the single hardest constraint in the system.
- Bidding strategy. Different goals need different strategies. A lead generation campaign & a brand defence campaign should not share one.
- Reporting clarity. You need to see performance by the dimensions you make decisions on: service line, location, device behaviour, funnel stage.
If splitting a campaign does not change how you allocate budget, set bids, or make decisions, do not split it.
Segment on the axes that matter
In practice, four dimensions justify separate campaigns most of the time.
Brand versus non-brand. Always separate. Brand traffic converts at a completely different rate & will distort every metric it is mixed into. Mixed reporting makes non-brand look far healthier than it is.
Service line or product category. Where the services have genuinely different margins, sales cycles or target audiences. If your margin on one service is triple another, they cannot share a budget sensibly.
Geography. Only where you need different budgets or bids by area. A business serving Kathmandu, Pokhara & Biratnagar with different competitive intensity & different capacity in each has a real reason to separate. A business serving one valley does not.
Intent or funnel stage. High-intent commercial terms & broader research terms behave differently enough to warrant separate control, particularly when budget is tight & you want to protect the commercial terms first.
Notice what is missing: match type. Splitting campaigns by match type made sense when match types behaved predictably. It no longer does, & it fragments data for no benefit.
Ad groups: tight enough to be relevant, loose enough to learn
The old advice of one keyword per ad group produced technically perfect relevance & statistically useless data. The modern approach groups keywords by shared intent, so that a single set of ads is genuinely appropriate for every query in the group.
The test: could you write three ads that make sense for every keyword in this group? If yes, the group is coherent. If one keyword forces an awkward compromise in the headline, it belongs elsewhere.
In practice this means roughly five to twenty closely related keywords per ad group, with three responsive search ads & a landing page that matches the group's intent specifically. A group of keywords pointing at a generic homepage will underperform regardless of how well the ads are written.
Match types in a world of loose matching
Exact match is no longer exact & broad match is no longer reckless. Both statements matter.
A workable default: build on phrase & exact match for your known, proven terms where you want control. Use broad match deliberately, in its own ad group or campaign, paired with a smart bidding strategy & watched closely, as a discovery mechanism. Broad match without conversion-based bidding & without disciplined negatives is where budgets disappear.
The critical companion to all of this is the search terms report, reviewed weekly without exception. Loose matching means the queries you actually pay for drift constantly, & the report is your only visibility into that drift.
Negative keywords are structural, not tactical
Treat negatives as part of the architecture. Build shared negative lists that apply across campaigns for the universal exclusions: free, jobs, careers, salary, DIY, complaints, & the competitor terms you deliberately avoid.
Then use campaign-level negatives to keep campaigns out of each other's territory. Add your brand terms as negatives in non-brand campaigns, or brand traffic will leak in & inflate the results. If you run separate service line campaigns, cross-negate the terms that clearly belong to one service.
Review new search terms weekly & add negatives from the report rather than from imagination. Guessing at negatives blocks queries that would have converted.
When to consolidate instead
If your campaigns are individually generating fewer than roughly 30 conversions a month, automated bidding has too little to work with. Consolidate. Merge related campaigns, accept slightly less granular control, & let the system optimise against a meaningful dataset.
This is a genuine trade-off & it is uncomfortable for anyone used to managing at keyword level. The compensation is that reporting segments & audience signals can recover much of the visibility you lose from structure, without fragmenting the data itself.
Naming conventions, because you will forget
Adopt a consistent convention on day one & apply it without exception. Something like Brand or NonBrand, then service, then location, then match strategy. It looks bureaucratic & it saves hours every month once the account has forty campaigns & two managers.
The same applies to conversion actions. Name them clearly, mark only the primary business outcome as a primary conversion, & keep secondary actions as observed. Bidding towards a soft conversion such as a page view will optimise you towards people who read & never enquire.
Auditing an inherited account
Work in this order when you take over an account:
- Check conversion tracking first. Roughly half of underperforming accounts have a tracking fault, & every other conclusion depends on it being correct.
- Look at spend concentration. Find where the money actually goes rather than where the structure suggests it goes.
- Pull ninety days of search terms & read them. This tells you more about the account than any dashboard.
- Identify campaigns with insufficient conversion volume & plan consolidation.
- Check that landing pages match ad group intent & load quickly on mobile.
Change one structural thing at a time & allow two to three weeks between changes. Automated bidding relearns after every significant edit, & stacking changes makes attribution of the improvement impossible.
Structure is not glamorous work & it is the foundation everything else sits on. If you want an account reviewed properly, look at our paid search service, or share your account structure & we will tell you where it is leaking.