Why Most Business Problems Are Really System Design Problems

Business problems often look like software problems because the missing information is visible on a screen. The underlying fault may be elsewhere: nobody owns the next decision, two departments use different definitions of completion, or an exception has no route back into the process. Buying another application before identifying that fault can add another handoff.

Key takeaways

• Identify whether the blockage is information, authority, capacity, or software.
• Agree on ownership and a measurable completion milestone before building.
• Test a small process change and measure rework across the whole handoff.

Start with one delayed decision

Choose a real transaction that disappointed a customer or consumed unnecessary time. Follow it from the first request to its final outcome. Ask each person what they received, what they needed, what they changed, and who they expected to act next. Record timestamps where available. Do not begin by drawing the process everyone wishes they followed.

For example, a quotation may wait because sales is missing a delivery estimate. Operations may believe sales is still confirming quantities. Both teams can be working diligently while the customer hears nothing. A new quotation tool will not resolve that disagreement unless the required inputs and response owner become explicit.

Distinguish four kinds of bottleneck

An information bottleneck means the necessary facts are missing or scattered. An authority bottleneck means nobody knows who may approve the next step. A capacity bottleneck means the right person has more work than they can complete. A system bottleneck means a reliable decision exists but the software cannot execute or communicate it efficiently.

These require different interventions. Better forms help incomplete requests. Delegated approval limits help routine decisions. Capacity planning helps overloaded teams. Integration helps repeated copying. Treating all four as automation opportunities leads to fast movement of work that is still fundamentally unresolved.

Define the smallest useful operating contract

For the quotation example, write down the minimum specification: confirmed items, requested delivery date, responsible estimator, promised response window, and reasons an estimate may be blocked. Define what sales should tell the customer while an exception is open. Attach those rules to the record rather than relying on someone's memory.

The contract should include an ending. Is the work complete when the quotation is sent, when the customer acknowledges it, or when an order is accepted? Choose one measurable milestone for this workflow. Other milestones can belong to subsequent processes without making the initial step ambiguous.

Decide what deserves software

Test the operating contract manually on a small sample first. If the team still cannot agree on ownership, pause the build. If the rules are stable and the remaining friction is repeated data movement, automate that movement. If several teams need to see shared status, add a lightweight queue or dashboard before commissioning a large platform.

A useful decision record compares the proposed change with doing nothing. List implementation effort, continuing maintenance, permissions needed, and the expected reduction in delay or rework. Estimates should be labelled as estimates, then checked against actual results after a pilot.

Know when the intervention worked

Measure elapsed time to the chosen milestone, the proportion of requests returned for missing information, and the number of unresolved exceptions. Also speak to the people doing the work: fewer steps in one department may have created extra steps elsewhere. Review customer outcomes alongside internal efficiency.

The aim is not a perfectly drawn process. It is a business where the next action is understandable and recoverable. Share a delayed workflow through the contact form, or explore business automation services to discuss whether the answer is process design, integration, or custom software.