Most expensive Meta lead campaigns I have audited were broken before the first impression. They optimised for raw form fills, let anyone submit with one tap, sent no booked-job data back to Meta and split a small budget too many ways. No mid-flight targeting tweak fixes that stack. For a clinic, a trade or any local service business that lives on booked jobs, I run these 20 checks in launch order. The point is not to win a prettier CPL report. It is to give Meta a way to find people who book, give the business a way to reach them, and put enough budget behind both. Run the list before you approve spend, or before you pay someone else to spend it for you.

Tracking: can Meta recognise a job? (lines 1–5)

  1. Pixel and Conversions API both fire for the same lead. Check the browser and server events together, not in separate dashboards. The browser eventID must match the server event_id, and the event names must match; Meta can then keep one event when both arrive within 48 hours. That is deduplication: not counting one enquiry twice. If the IDs differ, your reported lead count can make an ad look like a winner when the CRM says otherwise. Verify the pair using the Pixel and CAPI setup.
  2. Lead events show as deduplicated in Events Manager. A plausible total is not proof. Open the event and check the deduplication status against what the CRM actually received. If the CRM has 40 leads and Meta reports 81, do not use Meta’s number to decide what to scale. Fix the count before you ask the account to learn from it. This check catches the quiet version of line 1: both events are arriving, so the setup looks alive, but each enquiry still appears twice.
  3. CRM stages go back to Meta: booked, showed and sold. An Instant Form submission tells Meta who was willing to tap a button. These stages tell it what happened after that tap. Connect the CRM so outcomes return through Conversions API for CRM; that feedback makes later Conversion Leads optimisation possible. If nothing after the form goes back, the account cannot learn the difference between a booked job and someone who never answers the phone. Put the loop in place before the first lead arrives.
  4. The optimisation event reflects lead quality, not just form volume. Choose the deepest useful event the account can support. Conversion Leads aims to find people likely to become customers rather than merely submitters, but it needs enough data: roughly 200 leads a month, daily uploads and a target stage that converts at 1–40% within 28 days. The Conversion Leads setup cites Meta tests at around 21% lower cost per quality lead. If your volume is not there yet, do not pretend the deeper signal is ready. Optimise as deep as you can and check quality manually.
  5. A test lead reaches the CRM before spend starts. Submit one through the Test Events view and follow it all the way through. Confirm that browser and server events arrive and deduplicate, then check Event Match Quality after a few days. A green-looking event in Meta is not enough if the lead never reaches whoever makes the call. Direct CRM integration also beats relying on a manual CSV download: those leads expire for download after 90 days. The Pixel and CAPI setup is useful here. Find the broken hand-off now, not on day seven.

Offer and form: filter before the callback (lines 6–10)

  1. One offer runs in each campaign. Do not bundle a free consultation, a price list and a seasonal discount, then wonder which promise attracted the enquiry. Pick the offer the sales team can actually follow up on and make the ad and form say the same thing. Broad, stacked offers can produce cheap-looking leads without telling you who wants what; the fake-lead breakdown describes that trap. One clear offer gives you a clearer read on booked rate. Three offers in one campaign give you a CPL and a guessing game.
  2. The form slows down accidental submissions. Use a Higher intent Instant Form or add a qualifying question when the offer needs a call. Higher intent adds a review screen and follow-up notice; More volume favours fast mobile submissions. That extra pause matters when pre-filled details make it possible to enquire without much thought. There is a placement trade-off: higher-intent forms deliver to Facebook Feed and Instagram feed on mobile, as the Instant Forms comparison notes. Choose deliberately, rather than accepting the fastest form because its CPL looks good.
  3. At least one qualifier appears; two or three multiple-choice questions are better. City, timeline, budget and property type can each screen for a different kind of mismatch. Choose the questions that decide whether the business can take the job, and make the answers taps rather than typing exercises. The fake-lead breakdown recommends higher intent plus two or three multiple-choice qualifiers such as city, timeline and budget. Do not make sales discover on the phone that a lead rents outside your service area. The form can establish that first.
  4. Phone number and postcode can be checked. Require a valid mobile format and a usable postcode rather than accepting any characters in a free-text box. This is a small check with an irritatingly concrete failure mode: 07123 or xxx@yyy enters the lead count, costs money and gives the caller nowhere to go. A qualifier cannot rescue an enquiry the team cannot contact, and a fast callback cannot reach a number that was never complete. Test the form with a bad entry as well as a good one.
  5. The service area appears in both the ad and the form. Name the towns served in the ad, then repeat location as a qualifier. The ad sets the expectation; the form catches the person who missed it. Do not use a broad location setting as a substitute for telling people where you work. Wide reach and slow follow-up are a poor combination: by evening, some people barely remember which business they contacted, as the fake-lead breakdown observes. Avoid paying to quote jobs you would not travel to do.

Budget and structure: give one ad set room to learn (lines 11–14)

  1. The daily budget can buy enough weekly optimisation events. Meta needs about 50 events in the week after the last significant edit to exit learning; an ad set that cannot get there can remain learning-limited, according to the account-structure discussion. Work backwards from the event you chose. At $15 per lead, $30 a day buys about 14 leads a week, not 50. Raise the budget, consolidate the ad sets or optimise higher up the funnel while qualifying harder in the form. Do the arithmetic before calling a volatile CPL a creative problem.
  2. Ad sets are consolidated instead of split for appearances. Say the budget is $3,000 a month. One broad ad set at $100 a day has a chance to gather useful data; five at $20 a day each may leave every set short of events. Three overlapping interests or four lookalikes do not become a strategy just because they fill a campaign diagram. They divide the budget and can bid against one another. The account-structure discussion makes the case for pooling budget. Start with enough concentration to learn; split only when there is a reason.
  3. The location radius matches where the team will travel. Set the pin where the vans sleep, not where the city looks biggest on a map. Check the edge of the radius against jobs the business would actually accept, not just the centre against its address. A lead 45 minutes away can look perfectly qualified in a form and still become a quote nobody closes. Keep this check separate from line 10: telling people the service area and setting the campaign’s location are two different jobs. Both need to agree.
  4. Existing customers, recent leads and staff are excluded where appropriate. Check the customer lists and the people who have already booked before launch, then check again when those lists change. Paying for a regular to submit the same form does not create a new job, even when the campaign reports a tidy lead. This is not a reason to build an elaborate maze of exclusions that starves delivery. It is a reason to stop buying enquiries the business already owns. Make the exclusions match the campaign’s actual offer.

Creative: test ideas, not colour palettes (lines 15–17)

  1. Three genuinely different concepts are ready. Use one proof ad, such as a before-and-after or review screenshot; one offer ad that says what happens on the visit, how long it takes and what it costs to start; and one face-to-camera ad in which the owner says who should enquire and who should not. Three versions of the same headline in different colours test a colour preference, not a reason to book. If all three struggle, you should be able to tell whether the proof, offer or audience fit missed the mark.
  2. Real job photos beat stock for this test. Show the van, tools, actual bathroom or actual engineer when those are what the customer is hiring. A stock smile can make a local service ad look like a national call centre and attract enquiries from people who have not understood where the business works. Watch booked rate alongside click-through rate. A popular image is not doing its job if the comments ask whether you cover their city and the diary stays empty. Use the creative to show the work, not merely to collect attention.
  3. A price or starting point appears when it filters poor-fit enquiries. From £89 callout, deducted from the job or Boiler services from £95 gives a reader a useful decision before they open the form. Show a starting point when price objections regularly kill the close. If every job is custom, let a qualifier such as timeline or property type do more of that filtering instead. Neither choice is automatically better; the check is whether the ad prepares someone for the conversation the business will actually have. A full diary of quotes that never convert is not a win.

Follow-up: turn a submission into a booked job (lines 18–20)

  1. The thank-you screen promises a callback window, and a person calls inside five minutes. Put the exact promise in the form: We call back within 15 minutes, 8am to 6pm. Then staff the promise. The speed-to-lead data describes sharply better contact and qualification odds for a five-minute response than a later one; that is about reaching and qualifying someone, not a guaranteed close rate. Slow follow-up can also make an ordinary lead look fake when the person has forgotten the enquiry. If nobody can call promptly, fix coverage before buying more submissions.
  2. The no-answer sequence is written before launch. Set two call attempts, one SMS and one email inside day one, and give the SMS the business name, the reason for calling and a one-tap callback number. Do not let the whole process end with one silent missed call. The form has already done the work of getting permission to talk; the follow-up needs a clear next step when the first call does not connect. Check who owns each attempt, including the hours named on the thank-you screen. Otherwise the lead may book a competitor while your dashboard still calls it new.
  3. Lead quality is reviewed by ad every week, not just CPL. Log each lead as Not reachable, Not relevant, Qualified, Booked, Won or Lost; compare ads by qualified rate and cost per booked job, then send booked, showed and sold stages back to Meta. The outcome-tracking breakdown covers the per-ad check. Even below the roughly 200-leads-a-month threshold for Conversion Leads, use the stages to move budget manually. If you hire a manager, ask to see the deduplicated event, CRM feedback, chosen optimisation event, weekly event count and cost per booked job in the account. A PDF of CPL averages is not the same proof. The 12 changes that lower cost per qualified lead and eight weekly jobs to ask your manager to prove are useful next checks. groas runs Facebook and Instagram ads with round-the-clock work on ads, audiences and budgets, a named strategist, logged actions and a $1,499 flat monthly fee, whatever you spend. This is the line I see skipped most: scale an ad at $9 CPL that books at 4%, pause one at $22 CPL that books at 31%, and the cheaper-looking leads cost you the jobs.