GOOGLE ADS STRATEGY

I've inherited enough broken accounts to say this with some confidence: almost none of them were broken by bad execution. The bids were fine. The copy was fine. What was wrong had been decided in the first week and never revisited, and by the time anyone called me, six months of data had been generated on top of a decision nobody remembered making.

So this is the pillar on how I build a google ads strategy before touching a single campaign — the objective, the measurement, the structure, the priorities. It's the least glamorous part of any google ads marketing strategy and it determines nearly everything. Automation raised the stakes here rather than lowering them: when the model does the optimising, your inputs are the only lever you still control, and strategy is just the name for choosing those inputs deliberately.

What follows is the order I actually work in when a google ads strategy gets built from nothing, not a checklist reconstructed after the fact. Four layers, each one feeding the next, and a diagnostic move at the end for reading an account that's already failing. I'll be explicit about the inferences, because the value here isn't the steps — anyone can list steps — it's knowing which decision you're really making when you think you're making a small one.

The first decision: one objective, and the cost you accept for it

Everything starts with what this account is for. Not "growth" — an actual, single, nameable outcome. Google's own framing is unambiguous: the objective you select should align with the main thing you want to get from your campaign, and choosing the right one helps optimise for outcomes like generating online sales, encouraging sign-ups, or driving traffic Google Ads Help on campaign objectives. Read that as a warning, not a setup step. The objective isn't a label on a google ads campaign; it's the thing the machine will maximise, at the expense of everything you didn't name.

Why "both" is the most expensive answer

The conversation I have most often goes: we want leads and we want brand awareness. Sure — but not in the same campaign, and not with the same money, and one of them is going to lose when they conflict. They always conflict, usually around month three, usually in a budget meeting where somebody has a number to defend.

Here's the inference I'd want people to take: an objective is not an aspiration, it's a trade you're agreeing to. Choose Sales and you're accepting that the algorithm will find the cheapest conversions available, which frequently means the people who would have bought anyway. Choose reach and you're accepting a number that won't reconcile with revenue for a quarter. Neither is wrong. Choosing without knowing which cost you've accepted is what's wrong, and it's the single most common origin of the "the account stopped working" call.

The accounts that survive their own success are the ones where somebody said out loud, at the start, what they were giving up. That sentence is the beginning of a google ads strategy; everything else is elaboration. I've never seen that written in a brief. I've seen a hundred briefs that list four objectives and rank none of them, which is the same as having no objective while feeling like you've done the work.

The question I ask instead

Not "what's your goal" — everyone has a rehearsed answer to that, polished for a meeting that isn't this one. I ask: if this account doubled next quarter, what specifically doubles, and who in your company would notice? The answer names the metric, the owner, and usually the reporting line where the account actually gets judged. That's the objective. If nobody can answer, the account has no strategy yet and building campaigns will only postpone the discovery.

The follow-up is nastier and more useful: what would have to happen for someone to shut this off? That's where the real constraint lives. An account judged on cost per lead and an account judged on pipeline contribution are different accounts even when the campaigns look identical, and the difference will show up as a fight nine months in if you don't surface it on day one.

The objective is not the same thing as the channel

People collapse these constantly. "We want search ads" is not an objective — it's a guess at an answer, made before the question. Sometimes the guess is right, and it still matters that it was a guess, because when it stops working you won't know whether the channel was wrong or the objective was.

My habit is to write the objective in a form where no channel appears in the sentence. If I can't do that, I haven't understood the business yet. Whether the money ends up in Search, Shopping, or Demand Gen is a downstream decision that the objective, the demand, and the margin should make almost for me — and when it doesn't feel almost automatic, that's a signal the layer above is still vague.

Second: measurement, before anything else exists

This is where I break the order most people expect. Structure comes third, not second. Before I draw a single campaign box, I want to know exactly what the account will count as success — this is the layer where a google ads strategy quietly succeeds or fails — because in an automated account, measurement isn't reporting. It's steering.

The inference people miss

Smart bidding strategies don't optimise toward your business. They optimise toward your conversion configuration, which is a model of your business built by whoever set the tags up, possibly a developer who left in 2023. Those two things drift apart silently and the account gets more confidently wrong over time, not less. The algorithm is loyal to the definition, not the intention, and it has no mechanism for noticing the gap.

So the audit I run first is uncomfortable and short: what's counted as a conversion, is it actually worth something, and are two things with wildly different values sharing one goal? An account counting newsletter signups and €4,000 purchases in the same bucket isn't mismeasured — it's being actively pointed at the cheap thing. Which is why how conversion tracking is configured is a strategic decision wearing an implementation costume, and it belongs on the whiteboard next to the objective rather than in a ticket somebody picks up later.

I've watched this play out with a client whose conversion rates looked spectacular for a year. They were counting a modal open. Nothing was wrong with the campaigns; everything was wrong with what the campaigns had been taught to want, and every optimisation made in that year had made the problem more efficient rather than smaller.

Values, not just events

If the objective involves money, the account needs to know what things are worth — otherwise every conversion is worth exactly one and the model optimises for volume of anything. Lead gen accounts are the worst offenders here, because the value lives in a CRM the account can't see. Getting that back in, even crudely, changes what the account learns more than any bidding tweak. Feed it a bad definition and no amount of clever choosing between bid strategies will save you: the strategy is a solver, and you've handed it the wrong equation.

"Crudely" is doing real work in that sentence. People stall for months waiting for perfect CRM integration when a three-tier value estimate — bad lead, normal lead, good lead — would have moved the account inside a fortnight. Directionally right beats precisely absent, every time, because the model isn't reading your values as truth. It's reading them as a ranking.

What measurement is for, and what it isn't

Two jobs, and they pull in different directions. Steering the algorithm wants few, clean, valuable signals. Understanding the business wants many signals, including ones you'd never bid toward. Conflate them and you get the classic failure: every micro-action promoted to a primary conversion because somebody wanted to see it in a report, and now pay per click money is chasing PDF downloads.

Keep the two separate on purpose. A small set of things the account optimises toward, a larger set it merely observes. That single distinction resolves most measurement arguments before they start, and it's a strategy decision — not a tagging one — because it decides what your ad spending is allowed to pursue.

Third: structure as a set of decisions, not a taxonomy

Now the whiteboard. And the mistake here is treating structure as organisation — a tidy way to file things. It isn't. In a google ads strategy, structure is the set of levers you'll have later. Every division you create is a place where you can put different money, read different numbers, and make a different call. Every division you don't create is a decision you've permanently delegated to the model.

The test for splitting

One question, and it's saved me from a hundred bloated accounts: would I ever want to spend differently on these two things? If yes, split them. If no, don't — you've just created a box that fragments data and buys you nothing. Not "are they different products." Not "are they in different site categories." Would the money ever move independently.

Geography usually passes. Margin tiers almost always pass, and almost nobody splits on them, which remains the most reliable free money I find in inherited accounts. Product colour never passes, though I've seen accounts built on it. A good google ads marketing strategy at this layer produces the fewest boxes that still let you make every decision you'll actually need to make — and the count is nearly always lower than people expect.

The cost nobody prices in

Every split divides the conversion data too, and automated bidding is a learning system that gets stupider on thin data. So structure is a genuine trade: control versus learning speed. Ten campaigns give you ten levers and ten poorly-informed models. One campaign gives you one confident model and no levers when it goes wrong. There's no right answer in the abstract — there's only the objective from step one, which tells you which levers you actually need, and everything else gets merged.

This is the exact place where old habits do the most damage. Anyone who learned this discipline a decade ago learned that granularity was virtue — one tightly themed box per intent, full control, everything separated. That was correct when you set every bid by hand. It's now often a way of starving your own bidding models while feeling rigorous, and the tell is an account where every box is individually defensible and the whole thing performs like a shrug.

Brand is a structural decision, not a tactic

One I'll die on. Brand traffic converts at a rate that has nothing to do with your advertising, and if it's mixed into acquisition, every number you look at for the next year is an average of two unrelated businesses. You'll conclude the account is healthy while acquisition quietly dies inside it. That's why the way brand keywords are handled gets decided at the strategy stage — it's about whether your reporting can tell you the truth, not about incremental clicks.

The same logic scales down. Any segment whose conversion rate is driven by something other than your advertising deserves its own box, or it will contaminate every average it touches: existing customers, one enormous bestseller, a seasonal line. The question isn't whether they're different. It's whether their presence makes the rest of the account unreadable.

Where keywords enter — and why they're not the strategy

Keyword research belongs here, at the end of structure, and this ordering is deliberate. Done before the objective, it produces an inventory of everything anybody might type, which is data, not a plan. Done after, it answers a specific question: where does the demand for this objective actually live, and how much of it is there?

The output I want from a google ads marketing strategy at this point isn't a list. It's a shape: how much demand exists per product or service, at what intent, at roughly what cost per click, and therefore what the account's realistic ceiling looks like. That shape tells me whether the structure I sketched is fundable. If half my boxes have no demand behind them, the structure was fiction and better to know now.

Match types and negative keywords are the next layer down, and they're control mechanisms rather than strategy — the mechanism by which the structure you chose stays what you said it was. Broad match without a coherent structure isn't aggressive, it's just noise with a budget. The keyword strategy layer is where those decisions get made properly, but note the dependency: it inherits from the boxes, and the boxes inherit from the objective. Every ad group in the account is downstream of a decision made three layers up.

Fourth: priorities, or where the budget actually goes

Objective, measurement, structure — now the hard part of any google ads strategy, which is that you can't do all of it and the account will be shaped by what you fund first.

Demand capture before demand creation

My default, and I'll defend it: fund the existing demand before you fund creating new demand. Capture is measurable in weeks and it funds the experiments; creation takes quarters to prove and gets cut in the first bad month, which is exactly when it needed to survive. Reverse the order and you're spending your credibility before you've earned any.

The exception is real and worth naming: when capture is already saturated — you're at 90% impression share on everything commercial — more capture budget buys nothing and creation is the only growth left. That's a genuinely different account and it deserves a different plan. But it's rarer than the people proposing it believe, and the check takes ten minutes.

The account has a budget floor, not a budget

Small accounts fail on strategy more than on execution, and the mechanism is specific: automated bidding needs a conversion volume to learn from, and spreading a small budget across a sensible-looking structure starves every box below the learning threshold. Nothing works, and nothing works for a reason that never shows up in any report. The strategy on a small budget is therefore mostly a strategy of refusal — one objective, one channel, one campaign, and the discipline not to add a second thing until the first one is winning.

The arithmetic is worth doing on paper before you build. Take the budget, divide by a plausible click price, divide by a plausible conversion rate, and look at the monthly conversions per box your structure implies. If that number is in single digits, you don't have a structure problem — you have a structure you can't afford, and building it anyway is choosing to fail slowly.

What I deliberately don't do first

Not audiences. Not extensions. Not testing ad copy and landing pages against each other. All useful, all secondary, and all of them are optimisations on top of a decision — worth something once the decision is right, worth nothing when it isn't. I've watched teams spend a quarter A/B testing calls to action in a campaign that was pointed at the wrong conversion the entire time. The tests were rigorous. The results were noise.

There's an order to which optimisations pay, and it isn't the order of how interesting they are. Landing page relevance feeds quality scores, which feed what you pay, which changes the whole account's arithmetic — that's leverage. A high performing headline variant is a rounding error by comparison. Yet the headline test is the one that gets a slide, because it's legible and finite and the landing page belongs to another team.

The priority I keep coming back to

If I had one intervention per account, it would be the conversion definition. Not the copy, not the structure, not the bids. It's upstream of everything the model does, it's usually wrong somewhere, and fixing it changes results without anyone launching anything. It's also the least popular answer in the room, because it implies the last year of work was aimed slightly off-target — which it usually was, and saying so is the job.

The order, and how to use it backwards

Objective, measurement, structure, priorities. It's not arbitrary — each layer is an input to the next, and the reason most problems originate here is that these decisions are invisible once campaigns exist. Nobody audits an objective. People audit bids.

Reading a failing account in reverse

This is the most practical thing in the article. When an account is underperforming, walk the layers upward, not downward. Is the structure hiding the problem inside an average? Is the measurement pointing at something that isn't the business? Is the objective one that was chosen by someone who's since left? A google ads strategy fails upward, and so must the diagnosis. The lower you find the fault, the more of the work above it is rubble — which sounds bleak and is actually good news, because the fix is a decision rather than a year of grinding.

The tell that you're in the wrong layer: you're making changes, the numbers move, and nothing improves. That's the signature of optimising within a bad frame. Genuine strategic faults don't respond to tactics, they absorb them — which is why an account can look busy and diligent for years while going nowhere, and why "we've tried everything" is nearly always true and nearly always evidence for the diagnosis rather than against it.

What a successful google ad campaign inherits

When something works, the credit usually goes to the last thing that changed. It's almost never the last thing. A campaign that can reliably drive qualified traffic is standing on an objective that named one outcome, a conversion definition that pointed at money, a structure with a lever where the problem was, and a budget concentrated enough for the model to learn. The final tweak got the applause; the four decisions did the work.

This matters practically, not just philosophically, because it tells you what to copy. Teams try to replicate the tactic — same bid strategy, same structure — into an account where the layers underneath are different, and it fails, and they conclude the tactic was overrated. What's actually transferable are the decisions, not their output. Effective google ads work is boring at the bottom and only looks clever at the top.

When to revisit

A google ads strategy isn't a document you write once. Objectives, once a year or when the business changes. Measurement, whenever the site changes — which is more often than anyone tells you, and it's the reason accounts decay quietly rather than breaking loudly. Structure, only when the priority test starts giving a different answer than it did when you built it. And in an inherited account, all four in the first week, before you touch anything, because auditing an account you didn't build means finding the decisions nobody documented, not the settings anybody can see.

Write the four decisions on one page before you build anything. Name the objective and the cost you're accepting for it. Define what counts and what it's worth. Draw the fewest boxes that keep the levers you need. Fund capture before creation. That page is the google ads strategy — the campaigns are just its consequence, and if the page is wrong, no amount of execution downstream will disagree with it.