Service
Google Ads, managed properly.
Search, Shopping, Performance Max, Demand Gen and App Campaigns. What changes between a good account and a bad one is rarely the bidding – it is what the bidding was told to value.
- From
- $2,500 / mo
- Minimum spend
- None
- Minimum term
- None
- Free audit
- ~5 working days
The first month
Week one: what is the account optimising towards?
Every automated bid strategy is a learning system pointed at a target you chose. Point it at a conversion that fires on page load and it will faithfully find you the cheapest page loads in your category. The campaign will look excellent and sell nothing.
This is the most common thing we find, and it is rarely carelessness. Tracking decays: a checkout gets rebuilt, a consent banner starts blocking a tag, a conversion gets duplicated when someone adds a second tag manager container. Nobody notices because the dashboard still shows numbers.
Weeks two to four: structure that matches your decisions
Campaign structure should mirror the choices you actually make about money. If you would move budget between the US and Germany separately, they are separate campaigns. If you would never treat two product categories differently, they do not need to be split – and splitting them just divides your conversion signal.
| Split by | When it earns its place | When it just fragments signal |
|---|---|---|
| Brand vs non-brand | Always – brand flatters every blended number | Never; this one is not optional |
| Market / country | Different budgets, margins or seasonality | Same language, same economics, thin volume |
| Margin band | Product margins vary widely | Uniform margin across the catalogue |
| Funnel stage | Separate targets for capture and discovery | One objective, one budget |
| Catalogue category | Categories genuinely behave differently | Usually – this is the default split and usually wrong |
What we do differently
Brand is separated and reported separately, permanently
Brand search converts at several times the rate of everything else and costs a fraction as much. Any account reporting a blended figure is mostly reporting how many people already knew about it. Splitting it out makes the first month look worse and every month after that legible.
Signal quality before bid tuning
Enhanced Conversions, server-side events, consent handling that does not silently drop a third of your conversions, and offline conversion imports where the sale happens later. An account with good signal outperforms an identically managed account without it, and this work is unglamorous enough that most agencies skip it.
Reporting that opens with money
Cost per acquisition, contribution, payback period, then what we changed and why. Impressions and click-through rate are inputs, and inputs belong further down the page.
Questions people actually ask
- What do you change first?
- The conversion actions. In most accounts we inherit, at least one is firing wrong – duplicated, triggering on page load, or counting a step nobody values. Smart bidding optimises towards whatever you report, so until that is fixed every other change is tuning a system learning from bad data.
- How do you structure campaigns?
- Around decisions you need to control separately – market, margin band, funnel stage – and merged everywhere else so each unit gets enough conversion signal to exit the learning phase. Splitting a modest budget across nineteen campaigns is the most common way accounts stall.
- Do you use broad match?
- Where the conversion signal is strong enough to guide it, yes – broad match plus good data now finds intent that exact match misses. Where the signal is weak it is an expensive way to discover search terms. The answer follows from your conversion volume, not from a principle.
- How do you handle Performance Max?
- As one campaign among several, with brand traffic fenced into its own Search campaign so PMax cannot absorb it and report it as incremental. We also extract the search term, placement and product-level reporting the default interface does not show.
- What reporting do we get?
- A written monthly review that opens with cost per acquisition and payback rather than with impressions, plus a live dashboard. If your funnel supports cohorts, reporting runs to the month a cohort turns profitable.
- How long before it works?
- Tracking fixes land in week one. Structural rebuilds take two to four weeks, then another two to exit learning. Month three is when the numbers mean something. Anyone promising results in week two is reading noise.
Free
Get the Google Ads audit
Related.
Performance Max in detail, ecommerce specifics, and what agencies charge across the market.