Start with an audit of what you already have
Before we change anything, we look at the account in the same order we would audit it for ourselves. Tracking comes first, because if Meta is not receiving accurate lead data, every decision further down is built on guesswork. We check whether the Pixel and the Conversions API are firing the lead event once and at the right moment. We also check whether the form, the follow-up and the sales figures match what Ads Manager is reporting.
We also want to know what has already worked, what you are paying per lead and where those leads go once they land. If you are starting from scratch, we look at your offer, the area you can serve and your sales process instead.
Pick a conversion event Meta can learn from
We set campaigns to optimise for an action Meta can see and count reliably, usually a lead form submission. That gives the algorithm a steady stream of results to learn from. Meta generally wants around 50 conversions a week in an ad set before it leaves the learning phase, so we avoid spreading budget so thinly that no ad set ever gets there.
Where it helps, we then layer on value so Meta can learn which leads are worth more. That only works if the lead event itself is clean. If you have been optimising for a sale that happens weeks later, we will usually move to a lead event first and keep the purchase data for reporting rather than for optimisation.
Connect leads to sales
A cost per lead figure is only useful if you know what those leads are worth. We work with you to confirm your lead-to-sale rate and your average order value. With those two numbers, we can estimate the revenue your Meta leads produce, even when the sale happens on the phone or in person. This is how we showed the real value of National Numbers' campaigns when most of their customers took longer than Meta's 7-day attribution window to buy.
We also look at how the lead event reaches Meta in the first place. The Conversions API sends the same lead events from your server that the Pixel sends from the browser, so fewer enquiries are lost to ad blockers and privacy settings. If both are sending the same lead, they need a matching event ID so Meta counts one enquiry rather than two. Events Manager's match quality score then shows how well Meta can tie those leads back to the people who saw the ads. Higher match quality usually means better optimisation, not just tidier reporting.
Clean, simple targeting
We tidy mixed audiences so each ad set has a clear job, while leaving Advantage+ enough room to find buyers. Some overlap is expected once Advantage+ is running. Audience segments show us how much of the spend reaches new customers and how much reaches existing ones, so we can see what the algorithm is actually doing.
Lookalikes, where we use them, are built from purchasers or high-value customers rather than everyone who visited your site. Uploaded customer lists help Meta understand who your best customers are. Existing customers are excluded from prospecting campaigns, so you are not paying to reach people who bought last week.
Location targeting is limited to areas you can actually serve. If you are a national business and can take enquiries from across the UK, targeting the whole UK is usually best because it gives Meta more room to find buyers. Tight location targeting only makes sense where there is a genuine physical limit, such as a service area or a grant that only applies in certain regions. For Elite Energy, tightening location targeting to eligible areas meant less spend on people who were never able to buy.
Build forms that qualify
For Meta instant forms we choose between a volume form and a higher intent form. A volume form is quicker to complete and usually brings more leads at a lower cost per lead. A higher intent form adds a review step before someone submits, which puts a little friction in the way and tends to filter out people who were tapping through without reading.
We add one or two qualifying questions where they make sense, such as postcode, budget or timescale. Every extra question helps weed out poor leads and also puts some genuine customers off, so we aim for just enough. We then look at the leads with you in the first weeks, not only the cost per lead, so we can tell whether the form is being too soft or too strict.
We often test a Meta instant form against a website form as well. There is more on that below.
Test creative with a purpose
Old creative is not something we rip out by default. If an ad is still bringing in good leads at a steady cost, we leave it running and make sure new ideas are ready for when performance does drop. If nothing new has gone live in months and cost per lead is rising as frequency climbs, that is when creative is part of the problem.
Every plan includes new creative each month from our in-house design team: 1 to 2 new creatives on Base Camp, up to 4 on Ascent and up to 6 on Summit. We test with a purpose. If a format is working, we build on it. If it is not, we try something else rather than adding variants for the sake of it. Static images, carousels and short video each do a different job. For Elite Energy, the first reel-style creatives we tested came in at a cost per lead 24% below the account average.
Move budget carefully
Large jumps in budget, or lots of significant edits in a short space of time, can send an ad set back into Meta's learning phase and make results less stable. We shift spend gradually and give each change a few days to settle, so a new ad set can build data without knocking what is already working. Read more in our guide to the Meta ads learning phase.
Report on what matters
You get at least a monthly report covering leads, cost per lead and lead quality, not just clicks and reach. On Ascent you also get a bi-weekly update. On Summit you get a weekly update plus a call every two weeks.
We look at Meta in two ways. First on its own, using the leads and cost per lead it reports on the platform. Then as part of your wider marketing, alongside your own sales figures and the feedback from your team on which leads were worth a conversation. Meta's numbers will not line up exactly with anything else, because each platform measures in its own way. Looking at both together gives a fairer picture of what the campaigns are contributing.