Start with the work
Select one task and a small enough sample for a person to inspect. Define a baseline before configuring software. A pilot should reveal the operational conditions under which the workflow is useful.
Build the working checklist
- Week 1: record baseline time, errors and ownership.
- Week 2: run a limited reviewed sample, including exceptions.
- Week 3: compare task outcomes, handoffs and recovery effort.
- Week 4: decide whether to retain, revise or stop.
Check it against a real task
Track a useful task outcome as well as time. Faster responses that require more rework may be a poor trade. Keep test activity separate from live customer data and audience evidence.
What to take away
Do not expand while ownership, contact basis or critical recovery remains unresolved. Keep an export and a documented way to return to the earlier process.
A four-week decision log
Scroll horizontally to read all columns.
| Period | Work | Evidence to keep |
|---|---|---|
| Days 1–7 | Define the job and observe a manual baseline | Time, errors, ownership and the outcome for a reviewed sample |
| Days 8–14 | Run a limited trial with normal and difficult cases | Inputs, observed outputs, failures and recovery work |
| Days 15–21 | Repeat with the same definitions and review differences | Correct outcomes, rework and operator effort |
| Days 22–30 | Choose keep, change or stop | Decision, unresolved limitations, cost and the next review date |
Choose the measure that matches the category
For messaging, a useful response and correct human handoff are better measures than messages sent. For account research, count verified matches and review time rather than records returned. For helpdesk work, distinguish an acknowledgement from a verified resolution. Keep the definitions unchanged through the comparison.
Set an expansion limit
Choose a sample that a person can inspect and document the actual cap. “Ten reviewed accounts” is an example of a bounded research sample, not a scientifically sufficient dataset. Stop if an excluded record re-enters a campaign, a critical handoff is lost or the owner cannot recover a failure. Fix and retest the boundary before wider use.
Working files
Blank templates for your own planning. These contain no customer records and are not submitted to Sales Fit Lab.
Keep baseline and trial observations separate, with outcome, failure and operator-effort columns.