Scoping and Success Metrics

Scope a thin, valuable slice

The first deployment should be the smallest thing that delivers real value to real users, end to end: one team, one workflow, real data. Avoid "phase 1 builds the platform, value arrives in phase 3".

Success criteria

Agree on criteria before building, written down and measurable:

• Outcome metrics: hours saved per week, error rate, revenue recovered, time to decision.
• Adoption metrics: weekly active users, share of cases handled in the new workflow.
• Quality metrics: accuracy on a test set, data freshness, uptime.

hours_saved_per_week = 4 * 3            # 3 analysts save 4 hours each
hourly_cost = 60
recovered_payments_per_month = 18_000   # failed payments chased within 24 h instead of 5 days
recovery_rate_gain = 0.15

yearly_value = hours_saved_per_week * 52 * hourly_cost + recovered_payments_per_month * 12 * recovery_rate_gain
print(f"time saved:     ${hours_saved_per_week * 52 * hourly_cost:>9,.0f} / year")
print(f"cash recovered: ${recovered_payments_per_month * 12 * recovery_rate_gain:>9,.0f} / year")
print(f"total value:    ${yearly_value:>9,.0f} / year -> success criterion: chase 90% of failures within 24 h")

time saved:     $   37,440 / year
cash recovered: $   32,400 / year
total value:    $   69,840 / year -> success criterion: chase 90% of failures within 24 h

Write a one-page scope

Problem, users, success criteria, in scope, out of scope, data sources and owners, dependencies (security review, network access), milestones, risks. Get the customer sponsor to sign off.