Monitor and Tune

Outcome: a funnel view of the inbound pipeline, alerts on the failures that are otherwise invisible, and a tuning loop driven by outcomes rather than opinion.

Surface: Sync GTM app
Level: Advanced
Uses: Formula columns · Slack · webhooks
Credits: 0
Prerequisite: Lessons 02–08

The funnel

Six counts, per day and per source.


Submissions received          412
  ↓ validation
Passed validation             398   (14 quarantined)
  ↓ enrichment
Fully enriched                317   (81 partial or failed)
  ↓ routing
Routed to a named owner       398   ← must equal "passed validation"
  ↓ response
First touch within SLA        341
  ↓ outcome
Meetings booked                23

The fourth line is the integrity check: routed must equal passed validation. Any gap is leads that went nowhere, and it is the single most important number on the page.

Alerts that matter

Zero submissions from a source

That normally produces some. A broken form is silent otherwise.

Routed count below validated count

Leads with no owner. Should be impossible; alert if it happens.

Unactioned lead past SLA

Fires on the lead regardless of owner, so a rep on leave does not hide it.

Enrichment failure rate above normal

A provider issue or a shift in traffic quality.

Latency above budget

The whole pipeline exists for speed; alert when it stops delivering it.

Quarantine volume spike

Either a bot campaign or a validation rule that broke.

Alert on zero, not only on errors. Every silent failure in this pipeline — a broken webhook, an expired credential, a form that stopped submitting — produces an absence rather than an error, and absence looks exactly like a quiet day. A rule that fires when a normally-active source goes silent for a few hours is the single most valuable alert here.

Tune the score against outcomes

The score you wrote in lesson 06 was a guess. After a quarter you have data.

Compare	What it tells you
Score of leads that booked meetings vs those that did not	Whether the score predicts anything
Conversion by intent signal	Which form and which page produce real intent
Conversion by source	Which channels are worth more budget
Conversion by score confidence	Whether unenriched leads really are worse — often they are not

The last row is the one that surprises people. If low-confidence leads convert as well as high-confidence ones, your enrichment coverage is a data-quality artefact and should not influence priority.

Tune the SLA by band

An SLA that is the same for everyone is either too slow for high-intent leads or unachievable for the rest.

Band	Target first touch
High fit, high intent	5 minutes
High intent, any fit	30 minutes
Everything else	Same day
Low confidence	1 hour — a human resolves what enrichment could not

Measure attainment per band, not overall. An aggregate SLA of 85% can hide complete failure on the band that matters.

The weekly review

Fifteen minutes, five questions:

Did every source deliver? Any zero counts?
Did routed equal validated? Any leads without an owner?
What was SLA attainment by band?
What did the quarantine catch, and was any of it legitimate?
Which source produced the most meetings per lead?

The fourth question is the one that finds problems nobody reported. Legitimate leads in quarantine are people who tried to contact you and did not get through.

Do this now

Build the six funnel counts

Per day and per source.

Add the routed-equals-validated check

As an alert, not a report line.

Configure all six alerts

Especially zero-submission.

Set SLA targets per band

And measure attainment per band.

Review quarantine weekly

After a quarter, compare score against outcomes

And adjust the weights.

Schedule the fifteen-minute weekly review

With a named owner.

Check your work

The funnel reports per source, not just in aggregate
Routed count is checked against validated count automatically
Zero-submission alerts exist for every source
SLA is measured per band
Quarantine is reviewed by a human weekly

Where this breaks

Monitoring aggregates hides single-source failures completely. Total submissions look normal because your highest-volume form is fine, while the demo request form — the one that produces most of your pipeline — has been broken for a week. Every count in this lesson must be per source, and the alerts must fire per source too.

Course complete

You now have every inbound source feeding one table, webhook intake with idempotency and validation, enrichment on arrival with defined fallbacks, a tolerant score, a routing cascade that always assigns an owner, a fast and honest first response, and monitoring that catches the silent failures.

Where to go next:

You want	Course
To repair the records this creates	CRM Enrichment
Better contact coverage on inbound leads	Waterfall Enrichment
Outbound against the accounts that never convert	Automated Outbound
Triggers on accounts that went quiet	Signals & ABM

Reference for this lesson: Webhooks, Slack, CRM integrations, Tables.