Google Ads account structure: a complete guide

Most people treat account structure as an organisational preference — a question of where things live, like deciding which drawer the invoices go in. It is not. Structure is the mechanism that decides which searches receive your money, and almost every account I am asked to look at has a structure problem wearing a bidding problem's clothes.

Here is the mechanism, stated plainly. A campaign holds a budget. Within that campaign, every keyword competes for the same pot. Google will spend that pot on whichever auctions it can win most cheaply, subject to whatever the bid strategy is trying to achieve. So if a single campaign contains both "what is a banquet hall" and "book banquet hall rohini 300 guests", the cheap informational clicks will consume the budget, because there are far more of them and they cost far less.

That is not a failure of the algorithm. It is doing exactly what it was told. The failure is upstream: somebody put two completely different commercial propositions in the same box and then asked the machine to optimise across them.

The diagnostic

Before restructuring anything, run this. It takes ten minutes and it tells you whether structure is your problem.

1. Open the search terms report for the last 90 days.
2. Sort by cost, descending.
3. For the top 20 rows, write down which campaign and ad group each sits in.
4. Now sort by conversions, descending, and do the same for the top 20.

If those two lists live in the same places, your structure is broadly working and you should be reading a different guide. If your converting queries are scattered across ad groups built around something other than buying intent — product ranges, service categories, the way your website navigation happens to be organised — then your best traffic is sharing a budget with your worst, and no amount of bid tuning will separate them.

There are many ways to organise an account. Almost all of them are ways of organising it around you — your product catalogue, your service lines, your internal team structure. There is one that organises it around the buyer, and it is the only one that keeps working when budgets get tight.

Split by distance from purchase.

Tier 1 — ready to buy. The query contains a transactional signal: "book", "hire", "buy", "quote", "price", "near me", a specific model number, a specific location plus a specific service. These people have decided. They are choosing between suppliers. Bid aggressively, use tight match types, send them to a page that lets them act.

Tier 2 — comparing. "Best", "vs", "alternatives", "reviews", "which", plural category terms. They have accepted they need the thing and are working out who from. Moderate bids. Send them to a page that makes the comparison for them and does not pretend the decision is already made.

Tier 3 — researching. "What is", "how to", "why", "guide", broad category terms with no commercial modifier. They may buy in three months or never. This is where budget goes to die in most accounts.

The point of separating them is not that tier 3 has no value. It is that you cannot make a rational decision about tier 3's value while it is invisibly consuming tier 1's budget. Separate them and the question becomes answerable: over 90 days, what did tier 3 cost, and what did it eventually produce? Sometimes the answer justifies the spend. More often it does not, and the honest conclusion is that paid search is an expensive way to buy awareness that content and organic search will produce for free.

What this looks like in practice

For a business with three services and three intent tiers, you do not need nine campaigns. You need enough separation that budgets cannot leak across tiers, and enough consolidation that each campaign gets sufficient conversion volume for bidding to learn from.

That tension is real and it has no universal answer. A rough rule: consolidate until each campaign is producing at least 15–30 conversions a month, then stop consolidating. Below that, automated bidding has too little signal to work with and you are trading structural clarity for statistical noise.

If your entire account produces fewer than 30 conversions a month, do not build intent tiers yet. Run one well-negatived campaign, get the volume up, and split later. Structure is a tool for allocating budget between things that both work. It cannot manufacture volume that is not there.

A campaign is a budget plus a bid strategy plus a set of settings that apply to everything inside it. Those three things are the only reasons to create a new one.

Create a separate campaign when:

• The budget must be ring-fenced. Tier 1 traffic should not be able to lose to tier 3 traffic in an internal auction.
• The bid strategy differs. A brand campaign and a generic prospecting campaign want different things.
• A campaign-level setting differs. Geography, ad schedule, language, network, device modifiers.

Do not create a separate campaign because two services feel conceptually different to you. Conceptual tidiness costs you conversion volume, and conversion volume is what bidding runs on.

Ad groups

An ad group exists to make the ad and the landing page match the query. That is its entire job. The old single-keyword-ad-group orthodoxy is largely obsolete — close variants mean an exact match keyword now matches a range of queries anyway, and splitting into dozens of one-keyword groups fragments data without improving relevance.

The workable modern rule: one ad group per theme that deserves its own ad copy and its own landing page. If you would write the same headline and send the traffic to the same page, they belong together.

Settings that quietly cost money

Three campaign settings are wrong by default often enough to be worth checking on every account:

Location targeting. The default is "Presence or interest" — people in your area, plus people merely showing interest in it. For a local service business this is almost never what you want. Change it to "Presence: People in or regularly in your targeted locations". This one setting has probably wasted more money than any other in Google Ads.

Search partners and Display expansion. Both are on by default on new Search campaigns. Both should be a deliberate choice with their own performance data behind it, not something you inherited.

Ad rotation and scheduling. If nobody answers the phone at 2am, and your conversions are phone calls, consider whether you want to be bidding at 2am.

Match types have changed more than most practitioners' mental model of them. Modified broad match was retired in 2021. Phrase match absorbed its behaviour. Exact match has matched close variants — plurals, misspellings, reordering, same-meaning paraphrases — since 2018, and the definition of "same meaning" keeps widening.

The practical position in 2026:

Exact match is not exact. It is "same meaning as". It still gives you the tightest control available, and it is where your proven, high-value queries belong.

Phrase match covers the meaning of the phrase with additional words either side. It is the workhorse for most accounts.

Broad match is now a genuinely different tool than it was five years ago. Paired with Smart Bidding, it is not "match anything vaguely related and hope" — the bid strategy evaluates each individual auction and bids according to how likely that specific query is to convert. Google's own guidance is that broad match plus Smart Bidding needs no match-type segmentation to optimise.

That guidance is correct conditionally, and the condition is the thing most accounts fail. Broad match plus Smart Bidding works when your conversion data is clean and your conversion volume is sufficient. It works because the bidding algorithm can tell good queries from bad ones. If your conversion actions are counting newsletter signups and duplicate form fills, broad match will confidently find you enormous volumes of traffic that produces more of them.

So: broad match is a measurement-quality bet, not a keyword strategy. If you have clean conversions, meaningful volume and a maintained negative list, use it — it will find queries you would never have thought of. If you do not, it will find you a very expensive education.

My default on a new or newly-inherited account is to start on phrase and exact, get measurement verifiably correct, build the negative list, and only then introduce broad match into a campaign that has enough conversion history to bid on it properly.

A negative keyword list built at launch and never revisited is a list of last year's mistakes. Search behaviour changes, Google's matching widens, and new irrelevant queries arrive continuously.

Building the first list

Open the search terms report, last 90 days, sorted by cost. Read every line that has meaningful spend and no conversions. For each, decide: is this query one I would ever want? If no, it becomes a negative. Work down until the rows get too small to matter.

This is the single most profitable hour available in most paid search accounts, it requires nobody's permission, and it is almost always the first thing that gets skipped when the week fills up.

The categories that recur

Across most accounts, wasted spend clusters into recognisable groups. It is worth building these deliberately rather than waiting to discover them one query at a time:

• Job seekers — "jobs", "salary", "career", "vacancy", "internship", "hiring"
• Free and DIY — "free", "diy", "how to make", "template", "sample", "cheap"
• Education — "course", "training", "certification", "tutorial", "pdf", "book"
• Competitor and comparison — deliberate, not accidental; decide whether you want these
• Wrong geography — city and country names outside your service area
• Wrong industry sense — homonyms of your product term in other fields

Build these as shared negative lists at account level and apply them to every relevant campaign. Campaign-level negatives should be reserved for keeping your intent tiers apart — tier 1 terms negatived out of tier 3 campaigns and vice versa, so the tiers cannot poach each other's traffic.

Match types on negatives

Negative keywords do not match close variants. A negative for "free" will not block "freely" and will not block a misspelling. This asymmetry catches people out constantly: your positive keywords match generously, your negatives match literally. Add plurals and common misspellings explicitly.

Cadence

Weekly while spend is significant; monthly at minimum. Put it in a calendar. The work is unglamorous, it compounds when done, and it decays when skipped.

If your brand terms sit in the same campaign as your generic terms, your account reporting is lying to you in a specific and flattering way.

Brand queries convert at several times the rate of generic queries and cost a fraction as much, because nobody is competing hard for your name and the searcher has already decided. Mixed into a generic campaign, those cheap conversions drag the campaign's blended cost per acquisition down and make the generic keywords look far better than they are. You will keep funding generic terms that are, on their own, losing money.

Split brand into its own campaign. Then you can answer the question that actually matters: what does new customer acquisition cost, excluding people who already knew our name?

Should you bid on your own brand at all?

The honest answer is that it depends on whether anyone else is.

Check the auction insights report for your brand campaign. If competitors are showing on your brand terms, you are choosing between paying a small amount to defend the click and letting them have it — and for a considered purchase, letting a competitor intercept someone who typed your name is expensive in a way that does not show up in your ad account.

If nobody is bidding on your brand, and your organic result already occupies the top of the page with sitelinks, brand bidding is largely paying for clicks you would have received free. Pause it for two weeks and watch total conversions, not paid conversions. If the total does not move, you have your answer.

Both of those are legitimate outcomes. What is not legitimate is running brand traffic mixed into generic and never knowing which of the two situations you are in.

Rebuilding an account is disruptive. It resets learning periods, it loses the campaign-level conversion history that bidding relies on, and it takes weeks to stabilise. It is worth it less often than agencies proposing it suggest, and more often than incumbents defending their own work admit.

Patch when: the structure is broadly organised around intent, conversion tracking is sound, and the problems are identifiable — a few campaigns misallocated, a negative list neglected, a bid strategy pointed at the wrong target.

Rebuild when:

• Conversion tracking has been wrong for long enough that the historical data is actively misleading. Bidding will learn from that history. Sometimes the fastest route to clean signal is a clean campaign.
• Campaigns are organised around something that cannot be fixed incrementally — one campaign containing every keyword the business has ever wanted, or a structure mirroring an org chart.
• The account has accumulated years of sediment: paused campaigns nobody understands, overlapping keywords competing with each other, conflicting negative lists.

Migrating without losing everything

If you do rebuild, do not delete the old account. Pause it.

Move in stages rather than all at once. Take your highest-value intent tier first, build it properly, run it alongside the old structure with a small budget, and compare. When the new version is demonstrably better, shift budget across and move on to the next tier. This costs more elapsed time than a big-bang rebuild and it is worth every day of it, because a staged migration gives you a control group and a big-bang one gives you a mystery.

Keep the conversion actions themselves untouched throughout. Rebuilding campaigns while simultaneously changing what counts as a conversion means you will never know which change caused what — and that, more than any structural mistake, is how a rescue engagement wastes its first two months.