Workflow Pilot Evaluation: Choose Metrics and Review Evidence

起動プロセスを理解します。

選択したワークフローがパイロットの前にどのように動作するかを、ワークフローに影響を与える条件や例外も含めて記録します。対処している問題に関連する観察を選択してください。慎重にサポートされたパイロットと文書化されていないベースラインを比較し、すべての違いを新しいツールのせいにすることは避けてください。

証拠と限界に同意します。

チームが何をレビューするか、誰がそれを収集するか、ワークフローを理解するためにどの程度の参加が必要かを定義します。合意された対策と併せて、ユーザーのフィードバックと運用上の問題も含めます。特にプロセスのごく一部のみが含まれる場合、パイロットで確立できないものを文書化します。

拡張は個別に決定してください。

ワークフロー所有者と一緒に調査結果を確認し、未解決の要件を特定します。パイロットを調整するか、繰り返すか、延長するかを決定します。前向きな実証があったとしても、変更に応じて他のチームが開始する前に、より広範なサポート、統合、導入を計画する必要性がなくなるわけではありません。

Define the comparison before collecting results

Write the question the pilot is meant to answer and define the workflow boundary. Agree when an item starts and finishes, which request types are included and which conditions make records unsuitable for comparison. Collect a baseline using the same definitions where a comparison is intended.

Choose a few measures connected to the question, such as elapsed time, correction work or unresolved requests. Record workload differences, exceptions and changes elsewhere in the process. Review qualitative feedback alongside the figures. Explain missing data and limits of the observation period. Use the results to decide whether to refine, extend or stop the pilot, rather than treating every measured difference as proof of a lasting improvement.

A workflow-pilot measurement plan

Prepare the plan before reviewing results. This makes it easier to explain which conclusions the pilot can and cannot support.

• Decision question: what the team needs to learn before choosing the next step.
• Scope: process boundary, audience, request types and observation period.
• Measure definition: start, finish, unit, data source and recording responsibility.
• Comparison: baseline or alternative process and relevant differences in conditions.
• Exceptions: missing data, unusual cases, correction work and unresolved items.
• Review outcome: findings, limitations, decision owner and the next experiment or action.

Do not report a single average without context

An average can conceal a group of difficult requests, and a small pilot may not represent the wider workload. Inspect the distribution or meaningful categories where appropriate, and keep unresolved cases visible. Explain changes in request mix before comparing periods.

Avoid converting an observed association into a guaranteed business outcome. Training, staffing or policy may have changed at the same time. State what was measured and what remains uncertain. A useful pilot report supports the next decision without overstating what its evidence establishes.

What if the pilot has too little data?

Say so and use the findings to refine the measurement method or plan a further observation period. Qualitative issues can still guide a correction, but limited evidence should not be presented as a reliable estimate for the entire organization. Keep the next decision proportionate to what is known.

ILLUSTRATIVE PLANNING EXAMPLE

Example: evaluating an approval pilot

Define when an approval request starts and finishes before measuring elapsed time. Compare similar request types and record incomplete or urgent requests separately. Review user questions and correction work alongside timing. Check whether a difference could reflect a changed workload or missing records rather than the new workflow.

Record the pilot scope, observation period and limits of the comparison. Use the evidence to choose the next step without presenting a small pilot as a guaranteed organization-wide result.

FURTHER READING

Official resources to explore

Explore the related official guidance below. The templates and examples are practical starting points to adapt to your own project.

• GOV.UK Service Manual: how discovery works