01 / Section
Activity tells you location, not the work itself
A list of visited apps or websites is a weak starting point for process discovery. It can show that a person moved between a customer system, a spreadsheet, and email, but it does not show which business outcome those actions supported.
The more useful unit is a completed task: a submitted request, reconciled record, approved document, or delivered response. Starting with the outcome gives each captured batch a clear question: what sequence of work appears to have produced this result?
This does not make the sequence a complete account of the work. A task may include offline judgment, conversations, interruptions, or actions outside the captured evidence. The task is a practical boundary for organizing evidence, not proof that every step was recorded.
- Use the task outcome as the anchor.
- Treat app and site activity as clues within that task, not as the task itself.
- Record what is visible separately from what the sequence may suggest.
02 / Section
Reconstruct one attempt before generalizing
For each completed task, assemble the relevant captured batch in time order. Link documents, records, messages, and page visits to the outcome only when the evidence supports that connection. A citation establishes where an observation came from; it does not establish that the observation caused the result or that the sequence was complete.
Then describe the attempt in plain language. For example: a request was received, details were copied into a working file, a record was checked, an exception was resolved, and a response was prepared. Mark each statement as recorded evidence or interpretation. If the evidence does not explain why a step occurred, say so.
This discipline prevents a common mistake: turning a plausible story into a verified workflow. A reconstructed attempt should remain useful even when parts of it are uncertain.
03 / Section
Compare completed attempts to find recurring patterns
One task can reveal a possible path. Several distinct attempts can show whether that path recurs. Compare tasks with the same outcome and note repeated inputs, handoffs, checks, rework, and exception points.
Look for patterns such as the same information being entered in two places, the same approval being sought through different channels, or the same missing detail causing a later clarification. Also record meaningful differences. Variation may reflect customer needs, policy requirements, or judgment rather than an avoidable flaw.
The goal is not to rank people or infer how they spent their time. It is to understand the shape of the work: which steps appear common, which are conditional, and which parts remain unexplained by the captured evidence.
- Common across attempts: possible core workflow.
- Present in only some attempts: possible exception or condition.
- Unexplained or inconsistent: area for further inquiry, not a conclusion.
04 / Section
Keep app sequences as discovery signals
App and site sequences are still useful, but they should remain separate from the task reconstruction. A recurring sequence may point to a handoff, a missing integration, or a place where information is repeatedly looked up. It is a discovery signal that helps determine what to investigate next.
Do not treat a sequence as an automatic recommendation or a complete procedure. The same sequence can support different outcomes, and a different sequence can support the same outcome. Confirm its meaning against task evidence, interviews, documents, and policy context.
Once a recurring pattern is understood, teams can write a clearer process description and identify planning options. Those options might include simplifying a form, clarifying an exception rule, or reducing duplicate entry. They remain proposals until the business confirms the requirement and decides what to change.
05 / Section
A practical review format
For an operations review, create one short record per completed task and use the same fields each time: outcome, captured evidence, reconstructed attempt, uncertainties, and possible follow-up. Add a comparison view that separates repeated steps from task-specific variation.
This format keeps the conversation grounded. Reviewers can challenge an interpretation without disputing the underlying citation, and they can identify missing evidence without treating absence as proof that an action did not happen. It also makes planning assumptions visible before they become process decisions.
The result is a more reliable discovery unit: not a person’s activity, and not an isolated click path, but an evidence-linked attempt to produce a defined business outcome.
