Why lead generation changes the Meta Ads setup
Nobody on Facebook or Instagram asked to be shown a clinic, a solar installer or an accountant. Delivery predicts who is likely to complete the action the ad set optimises for, and when that action is a form submission it gets very good at finding people who submit. Some of them are buyers. Others tapped through because the form was already filled with their name and email and it took two touches to finish. The platform counts both the same. That is the gap this setup closes: between what Meta can observe, which is a submission, and what the business is paid for, which happens later in a sales conversation Meta never sees. Where capturing enquiries sits against creating demand in the first place is the subject of the lead generation versus demand generation guide; everything below assumes the demand exists and the job is to collect it without paying for noise.
Recommended setup for lead generation
| Setting | Recommended for lead generation | Why it differs from other objectives | Feature page |
|---|---|---|---|
| Objective | Leads | Only this objective offers instant forms, conversion leads and lead-specific events as goals | Campaign objectives |
| Conversion location | Instant form when the site converts poorly on mobile; website when the page does the qualifying | Sales and Traffic always send people away; Leads can finish the job inside the app | Campaign objectives |
| Form type | The higher-intent option with a review step and one or two qualifying questions | Purchases carry their own proof of intent; a form submission does not | Custom audiences for form-opener pools |
| Optimisation goal | Leads at first; conversion leads once CRM stages flow back | No other objective can learn from an outcome recorded in a sales system | Conversions API |
| Audiences | Broad delivery plus a lookalike seeded from qualified leads; existing clients excluded | A buyer seed exists for sales; for leads the seed must be filtered by sales first | Custom audiences |
| Follow-up | Leads routed to the team within minutes, with the ad and form named in the record | A shop's checkout closes itself; a lead decays while it waits | Attribution settings |
Conversion location: instant form or website lead event
The first choice is where the enquiry happens. An instant formAn ad format that collects contact details inside the platform, prefilled from the user's profile. opens inside Facebook or Instagram with contact fields prefilled from the profile, so the person never waits for a page to load. A website conversion location sends them to your own page and optimises for a lead event recorded there. The test is how well the site converts on a phone: if the page is slow, long or awkward in the in-app browser, the form removes that loss; if the page carries proof that helps people decide, the website route keeps it. Website leads should be sent by the pixel and the server together, as the Conversions API guide explains, so blocked browsers do not hide enquiries from delivery. The native lead form test shows how to settle the choice with a split test read on qualified leads from the CRM. When buyers want to ask questions before they give their details, the comparison of click-to-message ads and instant forms sets a conversation in a messaging app against the form.
Form friction: volume or intent
Meta's form setup lets an advertiser lean towards more submissions or towards higher intent, and the help centre documents a review screen in the higher-intent version where the person confirms their details before sending. Add one or two qualifying questions as multiple choice: budget band, postcode area, timescale. Each question costs some submissions and removes more of the wrong ones. Prefilled fields are the reason lead forms are easy; they are also the reason sales receives an old university email address from someone who does not remember enquiring. The campaign objectives guide places the form among the other Leads goals.
Teaching delivery with conversion leads
The Leads objective offers an optimisation goal that learns from stages a CRM reports back, which Meta calls conversion leads. Once the integration sends a stage such as booked or qualified, delivery can favour the people who reach it rather than everyone who submitted. Pick the stage by volume as well as truth: a stage that occurs rarely leaves the ad set stuck in learning, so a booked-call stage often works before a signed contract would. Agree the stage definitions with sales first, using the MQL and SQL stages guide, so the word "qualified" means the same thing on both sides.
Lookalike seeds built from qualified leads
A Lookalike AudienceAudience similar to a seed list. modelled on every submitter copies the casual responders along with the buyers. Seed it instead from a customer listAn ad audience built from the advertiser's own sources: customer lists, site visitors or engagers. of leads that sales qualified or that converted, and refresh it as that list grows. Under Advantage+ audience a lookalike acts as a starting suggestion, which the lookalike audiences guide covers. Exclude current clients and recent submitters from prospecting with lists built as the custom audiences guide describes, and keep form openers who did not submit as a separate retargeting pool.
Speed to lead
An instant-form lead was given in a feed, often between other things, and the memory of it fades quickly. Connect the form to the CRM or a notification route rather than downloading files, record which ad and form produced each lead, and call or message while the person still recalls the ad. Follow-up speed is part of the setup, not an afterthought, because a slow call turns a good lead into a poor one in the CRM data that conversion leads learn from.
A worked lead-generation setup on Meta
Limberwend Physio, an invented group of physiotherapy clinics
Limberwend runs four clinics and wants booked first assessments, each worth about £65 with an average course of five sessions. Its first Leads campaign uses the more-volume instant form with name, email and phone prefilled, a £45 daily budget and broad delivery. Leads were downloaded twice a week.
| Setting | First eight weeks | After the rebuild |
|---|---|---|
| Conversion location | Instant form | Instant form (site is slow on mobile) |
| Form type | More volume, no questions | Higher intent with review screen, two questions: clinic and pain duration |
| Optimisation goal | Leads | Leads for four weeks, then conversion leads on "assessment booked" |
| Audiences | Broad only | Broad, plus a lookalike of booked patients; current patients excluded |
| Follow-up | CSV download twice weekly | CRM integration, call within the hour during opening times |
Before the rebuild the campaign spent £2,520 for 336 leads, £7.50 each, but only 29 booked an assessment, a cost per booking of £87. After the rebuild leads fell to 141 over the next eight weeks on the same spend, at £17.87 each, yet 58 booked, a cost per booking of £43. Of those bookings, 21 came from form submitters who answered "more than three months" to the pain-duration question, which the clinics now prioritise on the call list.
The cost per lead more than doubled, and that was the right result. The questions and review screen removed people who would never have booked, the integration meant calls happened the same day, and the booked stage gave conversion leads a definition of success that matched the clinic diary.
Measuring lead generation in Meta Ads
Ads Manager reports Leads and Cost per lead as the result columns for a Leads campaign; read them per ad set and per form, since forms with different questions are different products. Cost per result in Meta is always tied to the event an ad set optimises for, which the glossary entry on CPA in Meta Ads sets out with its attribution and learning-phase caveats. Where CRM stages are sent, conversion leads appear as their own results and are the closer figure to revenue. For instant forms, compare form opens with submissions: many opens and few submissions points at a question or a privacy notice that stops people, while almost no drop-off can mean the form is too easy. For website leads, check that lead events arrive from both the pixel and the Conversions APIServer-side event sharing with Meta. and are deduplicated. Watch Frequency on smaller local audiences, where the same people see the ad quickly.
Beside those, the business measure is cost per qualified or booked lead, and the CPA calculator turns a booking value and a close rate into the most a lead may cost (CPACost of each conversion.). The rate at which leads book, effectively a downstream conversion rate (CVRShare of visits or clicks that convert.), is what the conversion rate calculator tracks. Review routing and lead counts daily in the first weeks, cost per booked lead weekly, and form questions and seeds monthly. The measurement plan framework is the place to write those definitions down before the campaign spends. For how much of a lead plan paid social should carry across networks, see the channel-level guide to lead generation through paid social.
Where the lead setup goes wrong
Changing the form and the optimisation goal in the same week makes it impossible to say which one moved the cost. Conversion leads starved of stage updates behave like an ordinary lead goal, so an integration that silently stops sending stages undoes the main advantage. Attribution windows that are shorter than the gap between form and booking leave the booking outside the window, which is why the attribution settings guide matters even for forms. And local audiences fill up: if frequency climbs while cost per lead rises, the problem is reach, not the form.
When a running campaign already produces leads that sales rejects, the diagnosis page for low-quality Meta lead ads splits lead outcomes by form, placement and audience to find which part of this setup has slipped.
Common mistakes
- Judging on cost per lead alone. The cheapest forms are usually the least likely to book.
- Downloading leads by hand. Delay turns an interested person into one who does not answer.
- Seeding lookalikes from all submitters. The model learns to find casual responders.
- Asking qualifying questions as free text. Multiple choice is quicker and easier for sales to sort.
- Switching to conversion leads before stages flow. The ad set has nothing to learn from.
