Run It On a Schedule

Outcome: the import-repair-write loop running on a schedule, with per-field cadences, a review gate on updates, and a digest that makes the work visible.

Surface: Sync GTM app
Level: Intermediate
Uses: Scheduled runs · CRM integration
Credits: ~10–15% of the initial repair per month
Prerequisite: Lessons 02–07

Cadence per field, not per table

The whole cost control is here.

Field group	Cadence	Trigger
New records	Immediate	On creation — never let a record be born incomplete
Job-change check on contacts	Monthly	Scheduled
Contact repair	On change detected	Event
Company firmographics	Quarterly	Scheduled
Verified email	On bounce, or on job change	Event
Score and quadrant	Monthly	Free, so recompute often
Dedupe check	Monthly	Free
Archive rule	Quarterly	Scheduled

Most of it is event-driven. Scheduled full passes are the fallback for fields with no available change signal.

A scheduled job that re-enriches everything monthly costs as much as the initial repair, every month, forever — and it does not look like a mistake, it looks like keeping data current. Set cadences before you schedule anything, and check consumption specifically on the second and fourth runs.

Enrich at creation

The highest-value automation in this course, and the cheapest.

A record created empty stays empty until a batch job finds it, and in the meantime it routes wrongly, scores wrongly and reports wrongly. Enriching on creation means the database never accumulates the debt this course exists to pay off.

Trigger on new account or contact creation: normalize the identifier, enrich company, verify the email, score, route. Under two credits per record and it removes most of the future work.

What stays manual

Merges of records with open opportunities

Revenue-affecting. A human decides.

Update-type writes on the first months

Fills can automate immediately; updates deserve review until the diff has proven itself over several runs.

Archive decisions

Review the list before it applies, at least quarterly.

Anything flagged low-confidence

An explicit queue, not a silent skip.

The monthly digest


CRM hygiene — <month>

  New records enriched at creation      412
  Contacts checked for job change     4,880
    moved                                74
    departed, flagged and suppressed      41
    routed to champion play               33
  Fields filled                         1,206
  Updates staged for review               88
  Duplicates found                        27  (12 merged, 15 queued)
  Archived                               310

  Completeness: verified email 78% → 84% · industry 91% → 94%

The completeness line is what makes the case for continuing to fund this. It is also the number that quietly degrades if the schedule breaks.

Guardrails

A scheduled job with write access needs limits that do not depend on anyone watching.

Write cap per run. A run proposing to change 5,000 fields when it normally changes 300 should stop, not proceed.
Never-overwrite list enforced in the job, not in a review step someone might skip.
Alert on zero. A change check returning no movers across 5,000 contacts means something broke.
Alert on spike. Ten times the normal change count means a data error, not a hiring boom.
Credit balance check covering several runs.

Do this now

Set cadence per field group

Written down, in the column description.

Build the enrich-at-creation trigger

The highest-value piece.

Schedule the monthly change check

Automate `fill` writes, keep `update` in review

Schedule the quarterly archive review

Review, not automatic application.

Build the digest

Including the completeness line.

Add the guardrails

Write cap, zero alert, spike alert, balance check.

Check consumption on runs two and four

Check your work

Every field group has a cadence and a trigger
New records are enriched at creation
update writes still pass through review
Guardrails are in the job, not in a human step
A digest is delivered with completeness tracked over time

Where this breaks

A scheduled job with write access and no cap can do a lot of damage quietly. A mapping change, a provider returning a different shape, or a filter that stopped matching can produce thousands of wrong writes overnight — and the first person to notice is usually a rep whose account data looks wrong. The write cap is the guardrail that turns that from an incident into a stopped job.

Course complete

You now have a CRM repair loop: a justified slice imported, records matched and deduped, gaps filled without touching human input, records scored and quadranted, changes staged and written back safely, hygiene driven by change detection rather than a timer, and the whole thing scheduled with guardrails.

Where to go next:

You want	Course
The movers as a full play	Job Changes & Champions
Better contact coverage on the repaired records	Waterfall Enrichment
New records enriched at creation	Automated Inbound
The loop run from chat	AI for RevOps

Reference for this lesson: CRM integrations, Credits, Workspace integrations, Webhooks.