Lead Quality Scoring: How to Stop Your Ads From Optimizing for Bad Leads

The problem with counting form fills

In lead generation accounts, the default conversion action is a form submission. The bidding algorithm then optimizes to produce as many form submissions as possible at your target cost.

It does this extremely well. Within weeks you have more leads, a lower cost per lead, and a sales team that is quietly furious because none of them close.

Lead quality scoring is the practice of assigning a value to each lead based on how likely it is to become revenue, and sending that value back to the ad platforms so they optimize for outcomes rather than form fills.

Define your lead stages first

Before any technical work, agree with sales on the stages a lead passes through. A workable default:

1. Raw lead - form submitted.
2. Contactable - valid phone or email, reachable.
3. Qualified - fits the ideal customer profile and has stated need.
4. Opportunity - meeting held, proposal in play.
5. Closed won - money received.

Each stage gets a value. If your average closed deal is €4,000 and 10% of qualified leads close, a qualified lead is worth €400 and you can work backwards from there.

Assign values, not just labels

The values you send should be monetary and proportional. A common starting structure:

• Raw lead: €0 or a token €1.
• Contactable: €10.
• Qualified: €100.
• Opportunity: €400.
• Closed won: actual deal value.

The absolute numbers matter less than the ratios. What you are telling the algorithm is that one qualified lead is worth ten contactable ones.

Getting the data back to the platforms

Three mechanisms cover almost every setup:

Offline conversion imports (Google Ads). Capture the GCLID on the form, store it in your CRM, and upload conversions with values as leads progress. Google supports this natively and it is the single most effective lead-gen improvement available.

Conversions API (Meta). Send the same stage events server side with the lead ID and value, matched back to the original lead event.

Enhanced conversions for leads. Where GCLID capture is unreliable, hashed email matching lets Google connect a CRM outcome to the original click.

Whichever route you take, the CRM has to be the source of truth. If sales does not update stages consistently, the whole system reports garbage.

Filter the junk at the front door

Optimization fixes the average. Filtering fixes the floor. Worth adding:

• A honeypot field to catch bots without a visible captcha.
• Email validation that rejects disposable domains.
• One qualifying question in the form, phrased as a helpful routing question rather than a gate.
• Phone number format validation for your target countries.

We regularly see 10% to 20% of raw lead volume disappear when these go in, with zero drop in qualified volume. That is not a loss. That is a cleanup.

Give it time to learn

Offline conversion imports have a lag. If your sales cycle to qualification is two weeks, the algorithm is learning from data that is two weeks old. That is fine, but it means changes take longer to show results than in ecommerce.

Set a 60 to 90 day evaluation window. Upload daily. Do not change targets weekly.

What good looks like

After a full cycle you should see cost per raw lead rise, cost per qualified lead fall, and total pipeline value per euro spent improve. If sales stops complaining about lead quality without you reducing volume targets, the system is working.

Optimize for the outcome you actually sell, and the ad platforms will find it for you.