Optimizing campaigns for sales: feeding closed-customer conversions back to Meta and Google

Optimizing campaigns for sales means sending Meta and Google back the information about which leads became opportunities and customers, so the platforms' algorithms learn to find people who buy. Without that feedback, the platform only sees the form that was filled out or the conversation that was started, and it goes looking for more people who fill out forms. For B2B marketing, where many leads never turn into a sale, that difference matters.

This article explains how it works, what needs to be in place first and how to implement it. It's part of the guide to AI in marketing for B2B companies.

Why does the platform optimize for the wrong thing?

Meta and Google campaigns use machine learning to decide who sees each ad. The algorithm looks at who completed the event chosen as the objective and searches for similar people.

The problem is the event. In most B2B accounts, the objective is the lead: a form submitted or a conversation started on WhatsApp. So the algorithm learns to profile people who submit forms. That includes the curious, students, competitors, people with no budget. The more the campaign "works," the more of that audience it brings in.

When the company sends the platform the events that happen later, such as qualified lead, meeting, proposal and sale, the algorithm gets a better signal. It starts to tell apart the lead who buys from the lead who just fills things out.

It's the same logic as AI anywhere: it learns from the data it receives. In marketing, giving the platforms' AI the right data usually pays off more than producing more creatives.

What needs to be in place first

Before any integration, check four things:

1. Source recorded on the lead. Every lead needs to reach the CRM with the campaign information and, when possible, the click identifier the platform generates. Without it, the platform can't tie the sale to the ad.
2. Clear stages in the CRM. Lead, qualified lead, opportunity, sale, lost. Each stage with written criteria and kept up to date by the sales team.
3. Reliable contact data. Complete and correct email addresses and phone numbers help the platforms recognize the person when the click identifier isn't available.
4. Legal basis and transparency. Using personal data to measure and optimize ads needs to be provided for and disclosed in the privacy policy.

The first point is where most companies stumble, especially when the sale happens on WhatsApp. The conversation starts outside the form and the source gets lost. The way to fix it is in how to integrate forms, CRM and WhatsApp without losing leads and on the page about media and CRM integration.

Which events should you send back?

It doesn't have to be only the sale. In B2B, the sale can take a while and happen in low volume, which makes learning harder. A good choice combines stages.

A common strategy is to optimize the campaign for the deepest event that has enough volume and send the others for measurement. As volume grows, you can move the objective up a stage.

Include the sale value when there is one. That lets the platform tell a small customer from a large one, and lets you compare campaigns by revenue.

How it works on Meta

On Meta, events can be sent through the Conversions API, a server-to-server connection that doesn't depend on the user's browser. In addition to website events, it accepts events coming from the CRM, such as qualified lead and purchase, linked to the original lead.

To match the event to the person, Meta uses identifiers such as the click ID and contact data, which must be sent hashed, as the platform's documentation instructs. The platform also offers ways to optimize lead generation campaigns for quality using these CRM events.

For ads that lead to WhatsApp, the principle is the same: the conversation needs to be recorded with its source so the sale event can be tied to the ad.

How it works on Google

In Google Ads, the equivalent feature is offline conversion import. The company sends which leads moved forward or bought, linked by the click identifier or by hashed contact data, through the enhanced conversions for leads feature.

With these conversions imported, automated bidding strategies start optimizing for what happens after the form. You can also assign a value to each conversion.

The technical details of each platform change often. Before implementing, check the current Meta and Google documentation for accepted formats, required data and time windows, because events sent too late may not be used.

Want your campaigns to learn from people who buy, not people who fill out forms?
Discover Nexo's AI for marketing

Implementation in six steps

1. Map the lead's path. Where it comes from, where it enters, where the source gets lost.
2. Capture the source at entry. Campaign parameters and click identifier recorded on the lead, including in WhatsApp conversations.
3. Define the events. Which CRM stages become events for the platforms, with the same name and criteria on both sides.
4. Connect the CRM to the platforms. Through a native integration, a middleware tool or scheduled uploads. What matters is that it's automatic and frequent.
5. Validate. Check in the Meta and Google dashboards whether the events are arriving and being matched to clicks. Adjust before changing the campaign objective.
6. Change the objective carefully. Switch optimization to the new event once there's enough volume, and watch the learning period before passing judgment.

What changes in how you read campaigns

With the sale recorded, the conversation between marketing and sales changes. Instead of arguing about whether a lead is good, both look at the same number: how much opportunity, sales and revenue each campaign brought in, and at what cost.

Some campaigns that looked great on cost per lead turn out to be expensive per customer. Others, with pricier leads, bring in better sales. Without the sale fed back, those decisions are made in the dark. How to compare channels and campaigns is covered in how to measure marketing with AI, and the concept of attribution helps you understand the limits of that reading.

What if an agency runs your media?

The agency can keep running the campaigns. What changes is the source of the signal. The sale lives in the company's CRM, not in the ad account, and the agency usually doesn't have access to it.

There are two possible setups. In the first, the company (or whoever manages the CRM) builds the integration that sends events to the platforms, and the agency starts optimizing with them. In the second, the agency receives a periodic report of which campaigns generated sales, which helps with analysis but doesn't teach the algorithm.

The first setup is the one that actually changes how campaigns behave. Agree with the agency on which events will be sent, under what names and when the campaign objective changes, so nobody is caught off guard by a temporary drop in lead volume during the learning period.

Common mistakes

• Optimizing for leads when the goal is sales.
• Sending events without the click source or without contact data.
• Using different names and criteria in the CRM and on the platform.
• Changing the campaign objective before validating the events.
• Sending the sale too late, outside the accepted window.
• Leaving the sales team without a deadline to update CRM stages.

Conclusion

Feeding sales back to the platforms is one of the most direct ways to use AI in marketing, because it uses the AI that already exists in the campaigns and gives it the right signal. The work isn't in the ad but in the path: source captured, stages in the CRM, events sent and validated.

With that in place, your media starts seeking out people who buy, and marketing and sales start discussing the same number.

Want to know whether your campaigns are optimizing for sales? Take Nexo's free assessment.
Take the free assessment