By Amy Jones, Senior Paid Media Consultant, Sonar Group
The campaign is generating more leads. Cost per lead is down. On the platform report, things look good.
Then the client calls. The enquiries are coming from areas they don’t service, or the sales team is spending too much time following up people who aren’t suitable.
Would you still increase the budget?
As a Senior Paid Media Consultant at Sonar, I manage campaigns across Google, Meta and TikTok. In more than 12 years working in paid media, I’ve learned that strong platform results don’t always tell the full story.
That’s why I’d start with this question when discussing AI. Automation can improve the numbers we ask it to improve. Our job is to make sure those numbers reflect what the business actually needs.
Before you scale, check what’s coming through
One of the more useful conversations in campaign management is the one between the person looking at the advertising account and the person handling the enquiries.
They can be seeing quite different things.
The account might show a campaign generating leads efficiently. The sales team might be asking for more enquiries from particular locations, fewer enquiries for an unsuitable service or customers who better understand the offer before submitting.
Both perspectives matter.
Consider a campaign targeting a broad geographic area. Keeping that audience broad may support lead volume, but it can also produce enquiries from places the business can’t service. Narrowing the targeting may reduce volume. That doesn’t automatically make it the wrong decision.
I’d want to understand where the suitable enquiries are coming from, what the sales team can handle and which locations the business wants to prioritise. Then we can make a recommendation with the trade-off clearly explained.
A lower cost per lead is useful. A lower cost per lead that creates more unsuitable enquiries needs a closer look.
A form submission is a starting point
When we ask a platform to generate leads, we need to be clear about what we’re counting.
Someone downloading a guide, requesting a call and becoming a qualified sales opportunity are different actions. If the account mainly measures the first step, we shouldn’t assume it tells us what happened at the last.
Google’s Performance Max uses the conversion goals and other inputs supplied by the advertiser to guide optimisation. Choosing those inputs is part of the strategy, rather than an account setup detail we can forget about. Google’s Performance Max guide
Before recommending more investment, I’d ask:
- Which actions are guiding optimisation?
- Are those actions still relevant to the campaign objective?
- What proportion of enquiries progress?
- Can we connect later outcomes with the advertising activity?
Where practical and permitted, qualified-lead or sales data can help close that gap. Google supports importing later lead outcomes through its lead measurement tools. Google’s lead measurement guidance
If that information isn’t available yet, we should be clear about the limitation. We can report that enquiries increased without claiming that sales increased.
More creative variations mean more things to check
Creative automation is another area where the practical detail matters.
Meta offers tools that can adapt creative, including expanding an image to suit a vertical Reels placement. That can save production time, but the adapted version still needs to work as an ad. Meta’s Reels advertising guidance
For a property campaign, I’d check whether the home is represented accurately, whether important text remains visible and whether the image still matches the offer.
For a retail campaign, I’d check the product, price and conditions. A variation that attracts more clicks is only useful if it communicates the right thing.
The review needs to cover the whole combination: image, wording, destination and placement. Approving the original asset doesn’t tell us whether every adapted version is appropriate.
Before enabling a creative feature, agree which elements can vary and which need to remain fixed. Then review the available previews and check the advertising once it’s live.
Give each test a job
“Let’s try the AI feature” is too broad to be a useful test.
A better starting point would be: we want to understand whether this setting can generate more qualified enquiries within our service area at an acceptable cost.
That gives us something to assess.
I’d confirm the budget, the comparison and the review period before making the change. Where possible, I’d use a controlled experiment and avoid changing the audience, creative, budget and conversion goal together.
Otherwise, if performance improves, we may not know why. If it declines, we may not know what to adjust.
The review period also needs to reflect the buying journey. A quick increase in form submissions won’t tell us much about sales quality if those enquiries take several weeks to progress.
Some problems sit outside the account
An automated campaign won’t resolve a landing page that leaves customers confused about eligibility. It won’t fix an enquiry form that asks for too much information too early, or a follow-up process that leaves leads waiting.
If a campaign is generating interest but people aren’t progressing, I’d look beyond bids and targeting. What does the customer see after clicking? Is the offer clear? Do they understand who it’s for and what happens next? How quickly does someone respond to their enquiry? Where are potential customers dropping off?
Make AI earn its place
Getting value from AI takes clear business objectives, human oversight and meaningful measurement. That means regularly reviewing what goes in, what’s enabled and what comes back.
Start with three questions:
- What have we allowed the platform to change?
- Are we attracting the right customers?
- What will our next test help us learn?




