Establish the Current Workload
Observe the manual process over a representative period before changing it. Record task volumes, handling time and meaningful variations by channel, team or case type. Use timestamps, work records and staff observations where possible, rather than relying only on memory. Note unusually busy or quiet periods so they do not distort the baseline.
Include Review and Exception Effort
Automation rarely removes every human step. Measure time spent reviewing outputs, correcting incomplete records, resolving failed runs and handling cases that require judgment. Track both the frequency and duration of exceptions. This creates a more useful post-pilot estimate than counting only the time taken by the automated step itself.
Separate Capacity From Cash Savings
Time released by automation is capacity, not automatically a cash saving. Document whether that capacity is absorbed by higher volumes, reassigned to other work or linked to an actual reduction in expenditure. Report these outcomes separately so decision-makers can see operational benefits without assuming that every saved hour changes the organisation’s costs.
Record Full Pilot Costs
Capture implementation work such as process design, configuration, testing, training and data preparation. Also record recurring software, support, monitoring and maintenance costs. Include internal staff time where it is material. Replace estimates with invoices, time records or approved budgets as evidence becomes available, and document any remaining assumptions clearly.
Count Verified Errors and Test Controls
Measure actual errors before and during the pilot using the same definition and sampling method. Record correction effort without assigning speculative lost revenue. For integrations, verify supported fields, permissions, sync direction, duplicate matching, failure and retry handling, and owner acceptance tests before enabling them. Do not automate marketing without appropriate permissions.
Compare Like-for-Like Results
Compare equivalent task types, volumes, teams and observation periods. Account for seasonality, staffing changes and process redesign that could affect results independently of automation. Review the evidence with process owners, document limitations and update the calculator’s user-entered assumptions. Expand only when results are repeatable and operational responsibilities are accepted.