Evidence-first implementation case study
From unmeasured follow-up to a measurement-ready contractor system
This case study documents the implementation specification and evidence standard for an anonymized home-service workflow. The source archive did not contain a defensible client outcome export, so lead and revenue lifts are deliberately marked pending, not invented.
Scope
The operating problem
Phone-based contractors commonly handle leads across calls, inboxes, calendars and handwritten reminders. The reference rollout standardizes five points where a contact can otherwise be lost: missed calls, new web leads, estimates awaiting a decision, appointments and completed jobs.
5Workflow specifications selected
20Documented action steps
26Minimum QA scenarios
Implementation artifact 1 · Lead-to-completion architecture
Inbound call or form
Contact dedupe and ownership
Estimate decision workflow
Appointment status
Completion and honest review request
What changed at the system level
| Control | Before specification | After specification | Evidence |
|---|---|---|---|
| Defined workflow triggers | Undocumented | 5 written triggers | Published workflow specifications |
| Documented action steps | Undocumented | 20 steps | Four per selected workflow |
| Explicit exit conditions | Undocumented | 5 sets | One set per workflow |
| Primary measurement definitions | Undocumented | 5 KPI groups | Workflow measurement sections |
| Minimum functional QA | Not defined | 26 scenarios | Benchmark dataset |
Implementation artifact 2 · Outcome evidence register
| Outcome | Required denominator | Status |
|---|---|---|
| Median human response time | Eligible inbound leads by week | Pending live export |
| Missed-call reply rate | Delivered acknowledgments | Pending live export |
| Estimate decision rate | Estimates entering cohort | Pending live export |
| Appointment kept rate | Confirmed appointments | Pending live export |
| Published review rate | Eligible completed jobs | Pending live export |
| Specification completeness | Five selected workflows | Measured: 5/5 |
Measurement plan
- Freeze a 30-day baseline export before enabling automation.
- Record eligibility rules and suppress tests, duplicates and opt-outs.
- Launch one workflow at a time and retain a change log.
- Compare equal-duration cohorts using the same denominator.
- Report raw counts with rates; do not attribute every downstream sale to automation.
Build the same evidence standard
Start with the workflow specification
Copy the reference workflows, estimate implementation effort and use the checklist to create a testable handoff.