I like automated bidding. I’ll say that first, because what follows can sound like a complaint about the tools. It isn’t. It is a complaint about how they get used.
The pattern goes like this. An account is switched to fully automated campaigns. The platform’s recommendations are accepted, mostly without reading them. The budget is set, the dashboard shows a healthy return, and everyone moves on to something else. Six months later someone asks why sales haven’t moved, and nobody can explain what the money has been doing.
The machine is doing exactly what it was asked
Automated campaigns are very good at one thing: finding more of whatever you have told them is a success. That is the whole trick, and it is also the whole risk.
If a success is defined as a real sale with a real value, the system goes looking for buyers. If a success is defined as a page view, a click on a phone number or a form that is mostly spam, it goes looking for more of those, and it finds them cheaply. The report looks wonderful. The business gets nothing.
So the first question I ask about any automated account has nothing to do with the campaigns. It is: what counts as a conversion here, and is that what the business actually wants more of?
Where the lax part comes in
Automation didn’t make account management unnecessary. It moved the work. The jobs that matter now are the ones that are easiest to skip:
- Conversion quality. Is the thing being counted worth paying for? Are better customers given a higher value than worse ones?
- What you’re showing up for. Automated campaigns will happily spend on your own brand name and on people who were about to buy anyway. That inflates the return and adds very little.
- What you’re feeding it. Product data, images, headlines and audience signals. Thin inputs get generic ads.
- What you’re excluding. Existing customers, job seekers, irrelevant searches and placements that will never convert.
- Whether it is working at all. Tested against a period, a region or a product range where it was switched off.
None of that is done by ticking “apply all recommendations”. Those recommendations come from the company that is paid when you spend more. Some are useful. They all deserve to be read by someone whose interests are yours.
A fair test
If you want to know whether an account is being managed or just left running, ask three things:
- What changed in the account last month, and why?
- How much of the reported return comes from searches for our own name?
- What would happen to sales if we turned this campaign off for a fortnight?
A good manager answers the first two in a minute and has an honest view on the third. If the answers are vague, the account is on autopilot, and you are paying a management fee for a machine to supervise itself.
What I’d do
Keep the automation. Fix what it learns from. Get the conversion data right, give real values to real outcomes, separate brand from non-brand so you can see each plainly, and put a person back in charge of asking whether any of it is adding sales you wouldn’t otherwise have had.
The tool is good. It still needs someone who is a bit suspicious of it.