The Comfort (and Seduction) of Clean Numbers

I’m at Atlassian Team 26 right now and just saw a slide that said:

“Current Data Quality in Jira: 90%”

“↓ 40% Decrease Manual Reporting”

Clean. Confident. Applause-ready.

And completely incomplete.

The Seduction of Clean Numbers

If you’ve spent any time in enterprise software, you’ve seen this movie before.

A system gets implemented. Governance gets layered in. Automation rules are added. Validation fields get tightened. And eventually, someone stands up and declares:

“Our data quality is 90%.”

Everyone nods. Leadership feels good. The system is “working.”

But here’s the uncomfortable question:

90% of what, exactly?

Data Quality Isn’t a Native Metric

Let’s be clear—Jira does not have a built-in “data quality” score.

That number didn’t come from the system.

It was constructed.

And when something is constructed, the definition matters more than the number.

Because in most cases, “data quality” really means:

- Required fields are filled in

- Values conform to expected formats

- Workflows weren’t bypassed

- Tickets aren’t sitting in obviously broken states

In other words:

“People are following the rules we defined.”

That’s not nothing. But it’s also not what most executives think they’re hearing.

Compliance ≠ Truth

A ticket can be perfectly “compliant” and still be completely useless.

- A priority field can be filled in—and wrong

- A description can be present—and meaningless

- A due date can exist—and have no connection to reality

So what does 90% actually represent?

90% of evaluated records passed internal governance checks.

Not:

- 90% accurate

- 90% predictive

- 90% aligned to execution

Just… 90% compliant.

And compliance is not the same thing as truth.

The 40% Reduction That Needs a Footnote

Then there’s the second claim:

“↓ 40% Decrease Manual Reporting”

Again—sounds great. But measured how?

There are a few ways this number could have been derived:

- Time spent building reports (before vs after)

- Number of manual spreadsheets replaced

- Volume of automated dashboards introduced

- Or, less rigorously, user perception (“this feels easier now”)

Without a defined baseline and method, the number floats.

And floating numbers don’t hold up under pressure.

The Real Problem: We Measure the System, Not the Outcome

This is the deeper issue.

Enterprise tools often optimize for internal integrity, not external effectiveness.

We measure:

- Field completion

- Workflow adherence

- Automation coverage

But we don’t measure:

- Whether decisions got better

- Whether execution got faster

- Whether reality is actually reflected in the system

So we end up with dashboards that look strong…

and operations that still feel heavy.

What an Operator Actually Cares About

If you strip away the theater, there are only a handful of signals that matter.

1. Completeness

Are the critical fields filled in at the moment of creation?

Late data isn’t useful data.

2. Stability (Drift)

How much does work change after it starts?

If ownership, scope, or priority are constantly shifting, your system isn’t stable—no matter how “complete” the fields are.

3. Latency

How long does it take for work to meaningfully update?

If tickets sit untouched for days, your system is lagging reality.

4. Consistency

Do similar types of work look the same?

If two identical requests produce wildly different data structures, your system isn’t coherent.

Reporting Should Disappear, Not Shrink

The real goal isn’t a 40% reduction in manual reporting.

It’s this:

You shouldn’t need reporting to understand your system.

If your data model is aligned to reality, insight becomes ambient.

Dashboards don’t answer questions.

They confirm what you already know.

The Slide I Wish I Saw

Instead of:

- “90% Data Quality”

- “40% Reduction in Manual Reporting”

I’d rather see:

- “Time to answer a basic operational question: 3 minutes (down from 2 hours)”

- “% of work items changing priority after start: 8% (down from 27%)”

- “Average time from intake to first meaningful update: 4 hours”

Now we’re talking about system behavior, not system compliance.

Final Thought

There’s nothing inherently wrong with the slide I saw.

It’s directionally right. It shows progress. It reflects effort.

But it’s also a reminder of something most organizations miss:

A well-governed system can still be a poorly aligned system.

And if you’re serious about operational clarity, that distinction isn’t academic.

It’s everything.