Troubleshooting

4 min read

Oracle is built to tell you when it cannot answer rather than to answer anyway. Most of what looks like a failure is it saying so.

”Not something this data can answer”

You asked for something your project does not record.

Which tester is fastest? How many defects will we have next month?

Nothing in Hawzu stores how long a person takes to work through a case, and nothing predicts future counts. Oracle refuses rather than ranking your team on a number it would have to invent, or extrapolating one.

Questions about duration, effort, cost, speed, who is fastest, or what will happen next all land here. That is a boundary, not a gap — a plausible answer to an unanswerable question is the worst thing a tool like this can produce, because it comes back with rows and therefore looks like it worked.

”Which did you mean?”

Your question was answerable but underspecified, so Oracle asked one short question back with two to four answers as buttons.

Show me the important ones Which did you mean? — High priority test cases · Open critical defects · Releases at risk

Select one and it runs. Note that answering is a full re-ask — Oracle plans the original question again, now settled by your reply — which is different from removing a condition, and slower.

”There is no member called Sarah”

A name in your question did not match anything in the project.

Names are looked up against your project rather than guessed at, so this is a refusal rather than an empty result. That is deliberate: a zero-row table for a misspelled name reads as good news, and it is not.

Check the spelling, or that the person is on this project rather than only in the workspace.

The Count Looks Wrong

Read the conditions before you doubt the data. Every answer prints Understood as — the conditions Oracle applied, in resolved terms — and that is almost always where a surprising number comes from.

Two that catch people out:

“The last 10 days” is shown as the date it computed. If that date is not what you meant, the count will not be either.
A person’s name is shown as the member it matched. Two people with similar names are a real possibility.

Select the × on a condition to remove it and the answer re-runs on the spot, calling no model. Comparing the count with and without one condition is the fastest way to find which one is doing the damage.

It Understood Something Differently

Remove the condition rather than rewriting the question.

Retyping asks Oracle to plan the whole thing again from scratch, and it may resolve something else differently on the way — so a rewrite to fix one condition can quietly change another. Editing the conditions changes only what you touched.

The Answer Is Empty, and You Do Not Believe It

Work backwards by removing conditions one at a time. The condition that changes the count from zero is the one that was wrong.

If every condition looks right and the answer is still empty, it probably is empty — but check what Oracle can see first. Your role’s access shapes what is askable, and a scope you cannot read is not a scope that returns nothing.

It Will Not Do the Sum

Oracle returns records or counts them. It does not sum, average or group — asked for those, it says so rather than approximating.

That work belongs to Observatory, which is built for it. Oracle can draw a chart from a fixed vocabulary, but a chart is not the query language growing arithmetic.

The Shape You Want Is Not Available

Some question shapes are outside what Oracle expresses at all:

Not available	Instead
Conditions joined with or	Ask the two questions separately
Nested or bracketed logic	Narrow in steps, checking each one
Sorting by more than one column	Sort by the one that decides the order
Very large result sets	Narrow the question — results are capped

Every condition combines with and. If you need either, ask twice — and you will usually find the two answers are more useful apart.

A Chart Redrew Differently Than Expected

Charts come from a fixed vocabulary of combinations plus a small set of trends at daily granularity.

A window your data does not support, or a grouping outside the vocabulary, is refused rather than approximated — so an unexpected redraw is Oracle declining to guess, not Oracle getting it wrong. See Charts.

Nothing in the Documents Covers It

For document questions, Oracle says so plainly rather than filling the gap:

I couldn’t find that information in the Hawzu documentation.

If you expected an answer from your own documents, check that the topic is actually in Canon and that you are asking inside the right project — Canon is project-scoped. See Answers From Documents.