Behind the Dash
Data Product Thinking

The Architecture You Ignore Eventually Becomes Your Process

When the system cannot express how the business works, people invent manual workflows to compensate.

July 25, 20262 min read

Behind the Dash

  • The manual work is compensating for assumptions the system never captured.
  • Users are not bypassing the product. They are completing it.

Front of the Dash

Visible symptom
Exports and spreadsheets
Common response
More training and documentation
Missed issue
The model cannot express reality

The request sounds completely reasonable.

"Can we reduce the number of manual steps?"

Every leadership team wants less manual work. Fewer spreadsheets. Fewer recurring meetings. Fewer people reconciling reports by hand.

The usual response is to document the process better or automate another task.

Sometimes that helps.

Often it misses the real problem.

Manual work is frequently treated as operational inefficiency. In reality, much of it exists because the product cannot represent how the business actually operates.

When people export data into spreadsheets, they are usually adding information the system does not understand.

When they maintain lookup tables outside the application, they are preserving relationships the model never captured.

When they schedule recurring meetings to explain reports, they are resolving ambiguity that the product leaves unanswered.

Those activities are easy to dismiss as bad habits.

More often, they are evidence.

They point directly at the assumptions the architecture failed to encode.

The temptation is to eliminate the visible process because it looks inefficient.

Delete the spreadsheet.

Remove the reconciliation step.

Discourage local tracking.

Consolidate everything into one workflow.

If the underlying problem remains, users simply invent a different workaround.

The manual effort comes back because the business requirement never disappeared.

This is why process improvement initiatives sometimes produce remarkably little improvement.

The process was never the root cause.

It was the symptom.

Good data products treat recurring manual work as design feedback.

Every spreadsheet should raise a question.

What decision is being made here that the product cannot support?

Every exception list should raise another.

What rule exists in practice but not in the model?

Every recurring reconciliation meeting should prompt the same investigation.

Why does the system require human interpretation before it becomes trustworthy?

Not every manual step deserves automation.

Some exist because humans should remain responsible for judgment.

Many exist because the architecture stopped one layer too early.

The healthiest teams stop viewing workarounds as user behavior.

They view them as requirements that arrived late.

Instead of asking how to eliminate another spreadsheet, ask why that spreadsheet became necessary.

The answer usually says more about the product than it does about the people using it.

Related essays