Skip to content
Guides · Advertising

How to optimize for qualified leads, not form fills

Teach Meta and Google to find leads who buy. Which sales stage to report, what each platform requires, and how to switch without losing performance.

Daniel Kalcher13 min read

This guide is for anyone running Meta or Google Ads for lead generation whose sales team keeps complaining about the leads. You will see why campaigns are so good at finding the wrong people, which sales stage to report back to the platforms instead, and what Meta and Google need technically to learn from it.

By the end you will have a plan: which data to store with each lead, which goal to pick, how to assign values, and how to tell whether the switch is working.

Why campaigns bring cheap, bad leads

The bidding algorithms at Meta and Google do exactly what you ask. If the optimization goal is "lead", they look for people who will submit a form with as little friction as possible. Whether that person answers the phone, has a budget or buys is invisible to them.

That produces a familiar pattern:

  • Prefilled forms: Instant Forms prefill name, email and phone from the profile. That lowers the barrier so far that curious and accidental submissions get through too.
  • Cheap audiences: People who click quickly and like filling in forms are cheap for the platform to reach. The algorithm moves budget there because your CPL drops.
  • A feedback loop: Every form filler who never buys still counts as a success. The algorithm goes looking for more people like them.

A low CPL is therefore not good news on its own. Look at cost per booked call or cost per sale instead. The cost per qualified lead calculator shows how two campaigns with different CPLs compare once you follow the leads through to sales.

The fix is not smarter manual targeting. It is a better signal: the platform has to find out which leads actually move forward.

Which sales stage to report

Between the form and the revenue sit several stages. Each one is a possible optimization goal with its own trade-offs:

StageWhat it tells youVolumeAs a goal
Lead (form)Someone showed interesthighStarting point, says nothing about quality
ContactedThe details are real and the person talks to youhighFilters spam and fake entries
MQLFits your target profilemediumGood if your criteria are strict
SQL or booked callSales confirms the needmediumOften the best compromise
Closed dealRevenue is inlowIdeal, but rare and late
Revenue with valueHow much the customer is worthlowFor value-based bidding

What counts as an MQL or SQL is for marketing and sales to agree on. What matters is that the line is always drawn in the same place. Otherwise the algorithm learns from moods rather than criteria.

Finding the right stage

The ideal goal meets three conditions at once:

  1. Close to revenue: People who reach the stage buy far more often than average.
  2. Frequent enough: The algorithm needs a certain number of events per week. Meta marks an ad set as "Learning limited" if it is unlikely to get around 50 optimization events in the week after its last significant edit.
  3. Fast enough: The stage has to happen within the platform's deadlines, covered below.

Say you spend €6,000 a month and get 400 leads (€15 CPL). Your team reaches 120 of them, 60 book a call and 12 buy. A sale closes on average five weeks after the lead. Twelve sales a month is too few, and five weeks is too late for Meta. Booked calls, on the other hand, happen 60 times a month, mostly in the first week, and are closely tied to revenue. Here, the booked call is the better goal.

Prerequisites: identifiers, deadlines, volume

Before you change anything in your campaigns, the feedback loop has to work. Three things decide whether it does.

Store identifiers with the lead

The platform must be able to match a sales stage to a click or a user. For that it needs at least one identifier from the original contact:

IdentifierPlatformWhere it comes from
Lead IDMeta, LinkedInThe platform's own lead form
gclid, gbraid, wbraidGoogle AdsURL parameter on the click
fbclid / fbcMetaURL parameter on the click
Email, phoneallThe form, sent hashed

Store these values with the lead in your CRM or in a tracking tool that ties them to the person. If the stage happens weeks later, they are gone otherwise. How to send the data from your server to each platform is covered step by step in how to set up a Conversion API.

Platform deadlines

Every platform accepts late conversions only within certain limits:

PlatformLimitSource
Meta, Conversions APIEvent no more than 7 days old when sentServer Event Parameters
Meta, conversion leadsOptimized stage within 28 days of the leadConversions API for CRM integration
Google Ads, with gclidUp to 90 days after the clickOffline conversion imports FAQs
Google Ads, contact data onlyUp to 63 dayssame

Meta's seven-day limit applies to the time of the event, not the lead. A call booked three weeks after the lead is fine as long as you send it within seven days of booking. On Google, the action's conversion window also limits what gets counted: 90 days at most.

That leads to the most important rule: report continuously, at least once a day. Meta requires daily uploads for the CRM integration, and Google recommends them for Smart Bidding.

Enough volume

For conversion leads, Meta asks for at least 200 leads a month. For Target ROAS on Search campaigns, Google requires at least 15 conversions in the past 30 days. These are minimums, not a promise: more events mean steadier learning. If you are below them, pick an earlier stage or consolidate campaigns.

Meta: setting up conversion leads

For lead ads with Instant Forms, Meta has a dedicated performance goal: conversion leads. Instead of as many leads as possible, the campaign looks for leads that reach a sales stage you choose. It is fed by CRM events you send through the Conversions API.

  1. Get leads into your CRM: Pull Instant Form leads into your CRM automatically and keep Meta's lead ID (15 to 17 digits) with each one.
  2. Define your stages: Decide which stages a lead goes through, such as "lead", "contacted", "qualified", "won". Meta wants at least two stages, ideally three. Sending only "sale" is not enough.
  3. Send every stage: Each stage change sends an event to the Conversions API with the lead ID, action_source set to system_generated, event_source set to crm, and your CRM's name as lead_event_source, as the payload specification requires. Report the raw lead as the first stage, too.
  4. Pass validation: Meta needs at least seven days of data, checks it in the Diagnostics tab of Events Manager and then analyzes your funnel.
  5. Pick the stage to optimize for: It should happen within 28 days of the lead and convert at between 1 and 40 percent.
  6. Switch the campaign: In Ads Manager, set the ad set's performance goal to maximize the number of conversion leads.

Be patient. Meta says that after validation there is a training period of one to two months before the integration fully feeds the model. Switching to a different dataset later restarts training.

Website leads on Meta

Conversion leads is built for Meta's own lead forms. If your leads come in through your website, send the later stages through the Conversions API anyway, with hashed email and phone and, where available, the fbc value from the click. According to Meta, all action sources can be used for ad optimization. How well a given event works as a campaign's conversion event depends mostly on how many of those events Meta can match. Check that in Events Manager before you optimize for it.

Google offers two ways to report sales stages, and today they work together:

  • Offline conversion import with gclid: You store the click ID with the lead and send it back with the stage.
  • Enhanced conversions for leads: You also send hashed contact data. Google calls this an upgraded version of offline conversion import and recommends it for new setups. Include the gclid whenever you have it.

According to Google, both kinds of upload have gone through the Data Manager API since 15 June 2026 and are blocked in the Google Ads API.

  1. Create conversion actions: Create one action per sales stage of the type "Import from clicks", for example "Call booked" and "Deal won".
  2. Accept the customer data terms: Without them, Google does not use hashed contact data.
  3. Capture identifiers: On your website, store gclid, gbraid or wbraid plus the contact details from the form.
  4. Report stages: Through the Data Manager API, with the click ID, normalized and SHA-256-hashed contact data, timestamp, value and currency, and a unique transaction ID.
  5. Start as secondary: Let the new actions observe first. Google recommends uploading daily for one to two conversion cycles before adding an action to the "Conversions" column, and with it to bidding.
  6. Make it primary and pick a strategy: Only then turn the stage into a primary action and bid with Target CPA, Maximize conversion value or Target ROAS.

Primary actions drive bidding; secondary actions only show up under "All conversions". If leads, calls and sales are all primary, one customer counts three times and the strategy keeps optimizing for the most frequent stage: the form. As the goal, Google recommends either the qualified lead or the converted lead. For a worked setup, see the use case Send offline conversions to Google Ads.

Assign values and bid on value

Not every customer is worth the same. With values, the platform learns which audiences bring the big deals, not just the many.

Which value to send

While actual revenue is still unknown, give each stage an expected value:

StageValueExample
LeadDeal size × close rate from lead€3,000 × 3% = €90
Booked callDeal size × close rate from call€3,000 × 20% = €600
Closed dealActual deal value€4,200

These numbers are hypothetical. Take your own rates from the CRM, over a period long enough that a single month doesn't dominate. The lead value calculator does the arithmetic.

Value-based bidding on Google

Maximize conversion value needs transaction-specific values. Add a target ROAS and it behaves like Target ROAS bidding. Google recommends reporting values for four weeks or one to two conversion cycles before setting a target, and giving a new target one to two cycles to settle.

Three rules keep values helpful rather than harmful:

  • Net, not gross: Subtract cancellations and refunds. Otherwise the platform learns from inflated values.
  • Only one valued stage in bidding: If calls and sales both count as primary, the campaign looks twice as good as it is.
  • Realistic targets: A target ROAS that is too high throttles delivery. Start close to the actual ROAS of recent weeks.

Switching over in five steps

Changing the optimization goal overnight resets learning and leaves you with nothing to compare against. A safer path:

  1. Measure: Make sure every sales stage comes back with identifiers, and watch the match rate in Events Manager and Google's diagnostics for a few weeks.
  2. Analyze: Compare campaigns by cost per call and cost per sale, not CPL. The ranking often changes. For a fair comparison you need your own attribution rather than each platform's view of itself; see attribution models for lead gen.
  3. Pick the goal: Choose the earliest stage that meets the conditions above.
  4. Test: Switch one campaign or ad set and keep the others running. Give the test at least the training period plus one sales cycle.
  5. Roll out: If cost per call or cost per sale goes down, switch the rest. A rising CPL is normal during this and no reason to stop.

Sending contact data from your CRM to ad platforms is a data transfer that needs a legal basis. What that means in practice is covered in GDPR-compliant ad tracking.

Common problems and solutions

ProblemCauseSolution
Meta matches few CRM eventsLead ID missing or in the wrong formatStore the lead ID on import and send it with every event
Events are rejectedEvent older than seven daysReport stages continuously instead of in a monthly batch
Campaign stays in learningToo few events at the chosen stagePick an earlier stage or consolidate ad sets
Google shows no imported conversionsClick older than 90 days or outside the conversion windowAlso report an earlier stage, check the window
Enhanced conversions for leads don't matchCustomer data terms not accepted or data not normalizedAccept the terms, hash lowercase email and E.164 phone numbers
Bidding still chases form fillsThe lead action is still primarySet leads to secondary, keep only the target stage primary
Platform ROAS higher than in your booksCancellations and refunds not deductedSend net values and correct canceled deals

How to do it with TSO

TSO handles the technical part between your sales team and the platforms, through each platform's Conversion API:

  • Collect stages: Your CRM or your team reports calls and deals by webhook, through the API or by CSV import, also via Zapier or Make. Payments and refunds come from Stripe, Digistore24 or CopeCart. Alongside system events such as "deal won" you can add your own, such as "sales call". There is no direct connector for Salesforce or Pipedrive.
  • Remember identifiers: TSO stores click IDs on the visit and merges email, phone and external IDs into one person. A deal that only carries an email still lands on the original click.
  • Meta CRM stages: TSO pulls leads from Meta Instant Forms automatically. Deal created, won and lost go to Meta as CRM stages by default, and you can mark further events as stages per destination. The lead ID goes along with every later conversion of the person.
  • Google Ads: Through the Data Manager API with gclid, gbraid or wbraid, hashed contact data, value and currency, to a conversion action of the type "Import from clicks". In test mode Google validates each event without recording it.
  • Send rules: Per destination, you decide which events go out and on what condition, for example only after a click from that platform. A delivery log shows what arrived and why anything was skipped.

Two things worth knowing: TSO stores contact data in plain text and hashes it only when sending. IP address and user agent go to the Conversion APIs by default even without consent; you can switch that off per destination. The use case Optimize for qualified leads walks through the setup, and Match Meta and CRM numbers shows how to reconcile the results with Meta afterwards.

Sources

How TSO helps

Frequently asked questions

Why do my Meta lead ads bring so many bad leads?

Because the campaign optimizes for the form. It finds people who submit forms with little effort, and Meta never learns which of them move forward in sales. Only when you report sales stages back can the algorithm learn what quality looks like.

Which stage should I optimize for?

The earliest stage that reliably signals revenue and happens often enough. On Meta it should happen within 28 days of the lead and convert at between 1 and 40 percent. Google recommends the qualified lead or the converted lead.

How many leads do I need for Meta's conversion leads goal?

Meta asks for at least 200 leads a month and daily uploads from your CRM. Below that, the algorithm has too little to learn from.

How long after the click can I send a sale to Google Ads?

Up to 90 days with a gclid, and up to 63 days with hashed contact data only (enhanced conversions for leads). Google only counts it within the conversion action's conversion window.

Should I bid on leads, calls and sales at the same time?

No. Create one conversion action per stage, but make only one of them the primary bidding goal. Keep the others as secondary actions for reporting.

Terms in this guide

Written by
Daniel KalcherCTO

Daniel is CTO and leads the technology behind tracking, attribution and conversion APIs at TSO.

More guides

All guides →

From the guide to clean measurement.

Set up TSO with us in a demo or request beta access.