What a decision-grade workflow baseline actually contains

A workflow baseline is not last year’s cost divided by twelve.

It is a defensible account of what enters the workflow, what happens, where it varies, what leaves and what the current result costs or constrains.

Without that structure, a business case can look precise while comparing unlike periods, ignoring exceptions or assigning value to activity that never changes the economics.

The Australian Government Department of Finance describes a business case as a way to define the problem or opportunity, compare options and analyse costs, benefits and risks. That discipline applies just as well to one operating workflow.

Start with the boundary

Write one sentence that names:

• the trigger;
• the start point;
• the finish point;
• the output;
• the accountable owner.

For example:

From receipt of a supplier invoice to approved posting, including matching, exception resolution and credit handling.

If the boundary expands whenever a new problem appears, the baseline will never settle. Adjacent problems go into a register for later triage.

Count the population, not only the average

Record the number of cases by period and the mix that matters. A monthly average can hide end-of-quarter spikes, acquired entities, customer segments, transaction values or channels with very different behaviour.

At minimum, test:

• volume;
• value;
• frequency;
• seasonality;
• mix;
• recent structural changes.

A workflow processing 10,000 simple cases and 200 difficult cases is not represented by one average handling time.

Separate the standard path from exceptions

Map the normal path first. Then identify where a case leaves it.

For each material exception, record:

• the condition that creates it;
• how often it occurs;
• who detects it;
• who resolves it;
• elapsed time;
• touch time;
• external cost or commercial consequence;
• whether the same case can fail more than once.

This matters because the expensive part of a workflow is often not the standard transaction. It is the repeated investigation, waiting, approval and correction around a minority of cases.

Use timestamps carefully

Elapsed time and working time answer different questions.

A case may take four days from receipt to completion but require only twenty minutes of human work. The remaining time may be queueing, waiting for evidence, approval latency or system batching.

Keep these measures separate:

• touch time;
• waiting time;
• total elapsed time;
• time outside an agreed service level.

A faster screen does not solve an approval queue. More staff do not solve missing evidence at the start.

Trace the evidence

Every material baseline figure should have:

• source system or record;
• extraction date;
• period covered;
• filters and exclusions;
• owner;
• confidence rating;
• known limitation.

Useful sources include workflow logs, ledgers, invoices, queue histories, timestamps, tickets, contracts and structured samples.

Interviews remain valuable. They explain workarounds, unofficial decisions and why fields cannot be trusted. They are context, not a substitute for observation.

Establish the economic bridge

Keep the operating measure separate from the economic claim.

For example:

• 1,200 exceptions per quarter is an operating fact;
• 35 minutes per exception is an observed or sampled effort measure;
• 700 hours is calculated capacity;
• $42,000 is a cost allocation;
• cash savings exist only if an external cost, overtime, headcount plan or other approved economic result changes.

This separation prevents an activity calculation from becoming a cash claim by accident.

Read why hours saved are not automatically cash savings before using labour capacity in a financial case.

Record confidence, not false certainty

Use a simple evidence scale:

• High: directly observed from reliable records across a representative period.
• Medium: reconstructed from multiple sources with known gaps.
• Low: interview estimate, small sample or weak source quality.

Low-confidence evidence does not always stop the work. It changes the decision. The next step may be instrumentation, sampling or a short observation period rather than implementation.

Compare options against the same baseline

Once the baseline is accepted, compare options on consistent terms:

• no change;
• workflow or policy change;
• control improvement;
• system integration;
• conventional automation;
• narrow AI;
• combined intervention.

Include implementation, licence, support, control, change and residual exception costs. The cheapest build is not necessarily the lowest operating cost.

Baseline checklist

Before recommending investment, confirm:

• the workflow has a fixed start and finish;
• one person owns the result;
• the population and period are representative;
• normal cases and exceptions are separated;
• elapsed time and touch time are not confused;
• material figures trace to evidence;
• confidence and limitations are visible;
• operating measures are separated from financial value;
• options use the same comparison basis;
• Finance has reviewed any financial classification.

A baseline is decision infrastructure. If it cannot withstand a sensible challenge from Operations, Technology and Finance, it is not ready to carry an implementation claim.

Pressure-test one workflow.

Further reading

• Developing a Business Case, Australian Government Department of Finance
• How to choose one expensive workflow worth fixing