Errors are loud, breaks are quiet: how to tell a lead integration has stopped
An error is easy: something turns red, someone gets an email. The failures that cost the most are the ones where nothing is red. The integration has simply stopped doing its job, and the first sign is a colleague asking why there have been no webinar leads this week.
People who run lead flows ask in public forums how to catch an automation that silently stopped, and the answers are hand-built: a "heartbeat" that alerts when nothing has arrived in a while, or a daily comparison of source and destination counts. This article puts those habits in one order.
Three ways a lead flow goes quiet
1. Nothing runs. The connection was switched off, a login expired, or a limit was reached and later calls were refused. One platform's documentation states that an integration that exceeds its limits receives an error response for all later calls until the limit resets, and one marketing tool's help centre notes that a CRM's web-to-lead form accepts at most 500 submissions a day. If nobody is watching the error log, this looks like a slow day.
2. Something blocks some records. A filter, a required field or a validation rule starts rejecting records. The form still says thank you; the lead never reaches the CRM. In a related case, a vendor employee answering a public forum question wrote that the marketing side cannot see which CRM validation rule fired.
3. It runs, but wrong. Records arrive with blank fields, the wrong owner, the wrong source. Counts look normal. This one is easy to miss, and a renamed field is a common cause (our reading, not a measured share; see the guide to renamed fields).
Three checks, one for each
Check 1: compare the count at both ends of every hand-off
For each hand-off (form to CRM, CRM to sales tool, sales tool back to marketing), keep a table of the last eight weeks with two columns: records that left, records that arrived. Divide arrived by left.
A healthy hand-off holds a steady ratio. A drop from, say, 97% to 62% in one week says something broke that week. This check catches the "blocks everything" failure, and it works even when every individual number looks plausible, because you are comparing two numbers that should agree.
Check 2: watch the gap since the last record
For each flow, write down how long it normally goes without a new record: every few minutes on a busy form, a day on a quiet one. Then check the time since the last record. A flow that has been silent three times longer than its normal gap deserves a look.
Tune it by weekday. A Saturday gap is not a Tuesday gap, and a webinar flow is quiet between webinars. This check catches "nothing runs" within hours, rather than at the end of the week.
Check 3: watch the blank rate of the fields that matter
Pick five important fields (source, owner, product interest, country, lifecycle stage). For each, record the share of new leads where it is empty. When one jumps, a mapping or a form or a rule changed. This is the check that catches "runs, but wrong".
Do it once by hand, then decide how often
Set up the three tables in a spreadsheet and fill them from exports. It takes an afternoon the first time. Look at them weekly for a month. You will learn what normal looks like, which is the part tools cannot guess.
Then decide honestly how often you will really do it. Weekly is our suggestion for the minimum. If you know you will not, the work belongs to something that does not forget.
Some limits to know about
- Consent and browser settings reduce what tracking can see. When a visitor declines non-essential cookies, one platform's developer documentation says it still counts the page view in aggregate but does not tie it to a person. A drop in tracked leads can be a policy change, not a broken form. Check form submissions, not only tracked visits.
- Time zones shift counts. A daily comparison needs both tools set to the same day boundary, or late-evening leads land on different days.
- A baseline is not a promise. Thresholds drawn from last quarter will be wrong during a campaign. Review them when you launch one.
Watching this all the time
Navceed is being built to watch hand-offs like this continuously: follow each record through every step, notice when one breaks, show the cause in plain words and prepare the fix for the data owner to approve. Today it works on an example company; connecting your own tools is coming next. Take the workflow break check on your own setup.