A working example
Compare evening calls against a real baseline
A business wants better evening coverage. Before routing calls to an assistant, it samples its own recent evening calls: how many arrived, how many were answered, what callers needed, and which requests became completed work. The same definitions are used during the pilot.
Define success before the first pilot call, then report captured, delivered, accepted, and completed work separately.
Record the current process first
Choose a period that reflects normal operations and note its unusual days. Count eligible calls, answered calls, messages, staff callbacks, completed requests, and any customer outcomes your business can verify. If you do not have a reliable measure, label it unknown rather than filling in a market statistic.
Record the staff time and provider charges actually involved. A missing call may have been a low-value question or a high-value inquiry; assigning the same revenue to each would distort the decision.
Keep the pilot small enough to interpret
Choose one phone path or request type and document when the agent should answer, collect information, hand off, or stop. Keep the comparison period, business hours, and definitions consistent where possible. Note changes in staffing, advertising, seasonality, or call routing that could affect the result.
For every pilot request, follow the work after the conversation. A transcript, form record, or successful notification can prove an intermediate step. The downstream system or staff record is needed to show that the requested action was accepted or completed.
Build a scorecard from observable events
Useful rates are ratios with explicit denominators: answered eligible calls divided by eligible calls; complete requests divided by requests started; correct escalations divided by cases that should have escalated; accepted bookings divided by eligible booking inquiries. Decide how to classify duplicates, test calls, and callers who hang up.
For a financial view, compare measured benefit with the actual pilot costs: setup, provider use, staff review, corrections, and support. Report uncertain benefits as scenarios using your own assumptions. A positive estimate is not a verified return until the underlying outcome is observed.
- Call coverage: did the intended calls receive the intended response?
- Workflow quality: were details complete and exceptions handed to the right person?
- Business outcome: did staff or the system confirm the requested result?
- Operating cost: what did the business actually spend and still have to do?
Before setup
Questions to bring to the conversation
- 01Write the pilot's eligible call and success definitions before launch.
- 02Collect a comparable baseline from your own records.
- 03Mark each event as captured, delivered, accepted, or completed.
- 04Track exceptions, corrections, and staff time alongside apparent wins.
- 05Agree the stop, change, or expand decision and who makes it.
Common questions
Clear answers before a pilot
What if we do not know how many calls we miss now?
Start by measuring that baseline from your own phone and staff records. Do not claim a recovered-call or revenue result without it.
Is an answered call a completed lead?
No. Answering is coverage. A useful lead needs relevant details and a verified next step; a completed sale or booking needs its own evidence.
Should we assign a fixed value to every call?
Use observed outcomes when available. If you model possible value, label the assumptions and keep the model separate from verified results.
Can a browser demo establish pilot ROI?
No. It can help review a proposed interaction, but ROI needs the business's own costs and accepted outcomes over an agreed period.
See the next step
Discuss a pilot scope
Bring your current call path and the outcome you want to measure.
Continue