Stop capturing fake leads with the new IPQS email verification

Experiment: Closing the door on “toxic leads” in real time

We had spent months looking at a chart that made no sense: our database was growing 15% quarterly, but revenue per subscriber was plummeting. The marketing team was celebrating, but the sales dashboard was in the red.

We came across the case of a retailer (a Data Innovation client) that lost €450,000 on a Black Friday because an aggressive acquisition campaign filled its list with garbage, causing Gmail to send all its emails (including those for VIP customers) to Spam.

That made us think: What if we are paying for Salesforce/Adobe licenses for ghosts? We decided to stop theorizing and audit the entry point.

The Hypothesis

If we implement real-time forensic validation on the form (not just checking for an @), the cost per lead will increase in the short term, but deliverability and conversion will skyrocket immediately. Learn more about Overcoming Resistance to Digital Change.

We suspected that reactive “Data Cleaning” (cleaning the list once a month) no longer works. We had to filter at the door.

What We Tested (The Setup)

Instead of launching a campaign and praying, we set up a 14-day experiment on a cold-traffic landing page (Display and Social Ads).

1. The change: We integrated a real-time validation API (such as IPQS or Data Innovation’s proprietary ones) directly into the form.
2. The filter: Before the lead entered the CRM, the system had <500ms to analyze:
   • Does the mailbox actually exist?
   • Is it a known spam trap?
   • Is the IP residential or does it come from a Data Center (AWS/Azure)?
3. The goal: Block anything with a high risk score before we were charged for hosting that data.

What Happened (The Raw Data)

This is where we got scared at first.

In the first 48 hours, our apparent conversion rate dropped 22%. The acquisition team panicked: “Did we break the form?”. Learn more about CRM with AI: Key to Autonomy in Retail.

No. The form was working perfectly. What was happening is that we were blocking the garbage.

When analyzing the rejected data (the “leads” that never reached the CRM), we found this: * The 3 AM pattern: We saw spikes in registration attempts at illogical hours for the local market. * Data Center IPs: 15% of attempts came from hosting servers, not residential or mobile connections. No real human registers from a cloud server IP to buy shoes. * “Burner” domains: A flood of @mailinator.com and variations of juan.perez01, 02, 03.

Intermediate result: Out of 10,000 registration attempts, we only allowed 7,800 in. Final Inbox result: The open rate for welcome emails in that clean segment jumped from 12% to 28% in the first week. Gmail detected that all emails were reaching real people who interacted, and our domain reputation was shielded. Learn more about Corporate Data Process Optimization.

What We Learned

1. The vanity trap: We were addicted to the “number of leads” metric. By cleaning the entry, the database grows slower, but each name is worth real money.
2. Bots have evolved: They are no longer dumb scripts. We saw bots that scrolled and filled fields imitating humans. Only IP and behavioral validation (what Data Innovation calls “forensic analysis in milliseconds”) could stop them.
3. The hidden cost is real: If we had let those 2,200 fake leads in, we would have paid for them in the ESP license and they would have dragged our legitimate emails to the spam folder. It is a double financial penalty.

What We Still Don’t Know: We still don’t have a clear picture of the exact “false positive” rate. Did we block a real human using a very strict corporate VPN? It’s possible. We are adjusting the algorithm’s sensitivity to find the perfect balance.

What We Would Test Next (Iteration)

For phase 2, we are going to test less technical but equally clever methods before calling the API: * Honeypots: Add invisible fields (CSS display: none) like “fax_phone”. If the field comes filled, we know it’s a bot (because humans don’t see it) and we discard it for free. * Source Audit: Cross-reference quality data with traffic source. I suspect that certain affiliate channels are bringing us 80% of this garbage.

Question for you: If tomorrow you deleted 20% of your database (those who never open), do you think your sales would drop… or would you discover you were paying for ghosts?

Have you ever done a radical cleanup and lived to tell the tale? I’m reading.

Let’s talk today https://datainnovation.io/contacto/!