Landing-page view to reported Lead, main affected ad set
1.86% to 0.48%
- Before
- 1.86% (1–9 May)
- After
- 0.48% (10–13 May)
Online educationClient work
Meta Leads Collapsed Overnight: A Read-Only Forensic Reconstruction
I reconstructed a US education provider’s Meta lead collapse from campaigns, events and page checks, delivering a working hypothesis and remediation plan.
The brief
Context
A US online education provider’s Meta lead campaigns deteriorated at the transition from 9 to 10 May. Paid traffic continued to reach the landing pages, but the affected funnel stopped producing leads at its previous rate.
I worked within a read-only scope. I reconstructed the failure window from campaign exports, account activity history, event diagnostics and landing-page checks. I did not change the campaigns or tracking to test a theory.
Sounds familiar?
Symptoms
The investigation
Diagnosis
I aligned the campaign history with the funnel counts and event evidence. The measurable break came after the landing-page view. For the main affected ad set, landing-page view to reported Lead fell from 1.86% to 0.48% between the baseline and failure windows.
-
I checked the landing pages’ availability and the presence of their forms and tracking. The pages loaded. That did not prove that a completed form reached the CRM or that every Lead event was attributed correctly, but it ruled out a simple page-down explanation.
-
I then separated the dataset-wide Lead signal from the campaign being investigated. Lead events were still arriving from mixed sources. That evidence ruled out a global Lead pixel outage; it did not prove that the affected paid funnel was healthy.
-
Purchase feedback needed its own investigation. Much of the observed activity appeared to come from manual or offline uploads, while browser Purchase activity was sparse. The upload history suggested a possible delay across the failure window. At the same time, the account history confirmed substantial structural changes that could disrupt delivery and learning.
Handover
What I delivered
Delivery noteDelivered May 2026
I built a dated reconstruction and a report that keeps confirmed observations separate from open explanations.
The strongest working hypothesis was a combined failure: degraded Purchase feedback feeding the Sales campaigns, amplified by structural edits and learning disruption. That was strong enough to guide a remediation plan, but not enough to declare the root cause proved.
The plan was to reconcile leads against the CRM, inspect upload processing and matching, validate a live form submission, confirm the dataset and conversion-event settings, and then test a controlled campaign with fresh, non-duplicated Purchase feedback. Implementation was a separate step.
Before → after
Results
Client work Done for real clients. Client details are anonymised.
The full results table
| Measure | Baseline | Failure window | Report date |
|---|---|---|---|
| Landing-page view to reported Lead, main affected ad set | 1.86%, 1–9 May | 0.48%, 10–13 May | 14 May 2026 |
Notes on the numbers
The outcome was a localised break, an evidence-backed working hypothesis and an ordered remediation plan. Upload processing logs, CRM reconciliation, historical tracking settings and live form-submit validation were still missing. I did not claim a repaired account, recovered leads or a proven upload failure.
A note from Daniilbefore you decide
What this means for a similar business
When leads collapse, an active pixel does not settle the question. I need to identify which funnel failed, which conversion event the campaign learns from, and whether the platform received useful feedback during the failure window. A read-only reconstruction can put those checks in order before anyone changes the account.
— Daniil
Have a similar task?
Describe your taskThe first answer is free, within one working day. Or write directly: next@taskfordaniel.com