Executive Questions That Should Never Be Ad-Hoc

If leadership asks it every Monday, it is not a fire drill. It is a standing question, and standing questions belong in the model. Heroic extracts are for surprises.

The pattern is familiar. The same cash, backlog, margin, or on-time question lands in the inbox before the ELT meets. Someone dumps a file or rebuilds a slice, and that person is a hero until the day they are out. Then the question is “ad-hoc” again.

Standing questions versus fire drills

A fire drill is a one-time cut: a customer rumor, a plant incident, a scenario that will not repeat. It can live as a labeled working paper with an expiry date. It should not get a workspace.

A standing question comes back on a clock. Cash by entity, bookings versus plan, inventory that is really cash, scrap that never made the pack. If a question has a weekday, it is product work, not a ticket.

Commissioning a dashboard per question is an intake problem: how the next page gets born. This post is about the executive cadence, and which questions may no longer live as extracts. Recurring leadership questions are the spec for the semantic product. Pages are optional. The measure and the grain are not.

Nor is it a pitch for talking to your data. Conversational analytics without a trusted model is just chat. Plain English inherits whatever you already certified. If a person with a SQL window still answers Monday’s questions, you are not ready to converse. You are ready to stop treating the clock as an emergency.

A question that survives four Mondays is a requirement. Calling it ad-hoc is how you keep underfunding the model.

How a Monday question stays a hero job

Someone pings. The certified app is close but not quite right: wrong entity, no as-of stamp, freight still mixed in. Extracting is faster than changing the product, so the extract wins and the product never learns.

Next Monday brings the same gap and another extract. The hero now has a private ritual, and the ELT thinks analytics is a person, not a loop.

The model could have held that measure all along, with the grain named, refresh owned, and access matched to who may see each plant. The standing question would have been a filter. You picked a fire drill because fire drills do not require a steward.

Screenshots of last week’s answer make it worse. A picture cannot take the follow-up, so the follow-up becomes another extract. Ask the model, don’t screenshot.

What heroic extracts actually cost

1. The clock never gets a product. Every Monday is day one and nothing compounds. The model stays a museum of almosts, and the meeting still waits on a person.

2. Definitions drift with the hero. Last week freight was in. This week it is out, and nobody wrote the sentence down. The ELT thinks it asked the same question twice. It did not. That is how data quality shows up as arguments.

3. Risk sits in the inbox. Row-level rules do not travel with a CSV. A standing cash question answered as an attachment is a standing leak. The live model can restrict access. The extract cannot.

4. Capacity and people burn on the wrong work. Refresh still runs for pages nobody uses. The analyst who should be fixing a measure is rebuilding last week’s dump. You pay for a program and run a print shop.

5. Real incidents lose their place. When everything is “urgent ad-hoc,” true emergencies wait behind the weekly ritual. You have taught the organization that the only way to get a number is to shout.

6. Trust never attaches to the app. If the CEO’s number arrived in email, the certified page is decoration. Adoption dies even when analyst usage looks fine, because the front door failed: nobody opens the dashboard.

You do not need a survey. Open last month’s executive inbox and count the repeats. That list is the backlog the PMO should already own.

How to put Monday in the model

1. Write the standing list. Ten questions, not forty. Pick the ones that come up at every ELT: cash, margin, backlog, inventory, throughput, on-time, or whatever your cadence actually fights about. If you cannot name them, you do not have a strategy. You have a suggestion box.

2. Turn each question into a measure contract. Include, exclude, grain, as-of, and owner, in one sentence leadership can say out loud. If two questions share a word but not a definition, give them different names. The product is the semantic model. The brochure can wait to be pretty.

3. Point the meeting at the product, not the person. The pack reads the certified app, or connected Excel on the same model. The old extract becomes a labeled working paper with an expiry. The real change work is removing the competing file: change management without the theater.

4. Allow conversation only on that product. Plain-language follow-ups help when they hit named measures and trusted dimensions. Sandbox models stay out of executive channels. Label every answer with the measure and the refresh stamp, and if the interface cannot show lineage, do not point it at the CEO.

5. Route true fire drills down a labeled path. A one-off gets a steward, a date it dies, and one test: will this come back? If it does, it graduates to the standing list and the model. If it does not, it does not get a report. Intake still says no to a page for every curiosity.

6. Review missed Mondays like incidents. An extract that still happened is a product bug or an ownership gap. Fix the measure and skip the praise for the hero. A Quickstart can land the first standing loop, and Managed Data & AI Advisory keeps the list honest. If the list is still “everything,” you need a roadmap before a microphone.

If the standing question is the close, see finance reporting consulting and cash, not charts. This post is the rule for any function that repeats itself weekly.

What good looks like

The ELT can ask this week’s version of last week’s question without a ticket. The answer matches the pack, and the follow-up is a filter, not a rebuild.

True incidents still get a human, and they are rare enough to notice. The hero has a different job now.

Frequently asked questions

What if the CEO’s question changes every week? Then you do not have standing questions. You have a strategy gap. Name the decisions that actually repeat. Curiosity still does not get a workspace.

Can people ask the model in plain language? Yes, once the standing measures exist, and only against that product. Chat on top of extracts is just a faster argument.

Do standing questions all need a dashboard? No. They need a certified measure, a grain, and a place the meeting already opens. A report page is one front end and connected Excel is another. The extract is not a third official source.

Get started

Stop calling the weekly ask ad-hoc. Put it in the model and save the heroics for surprises.

Want to know which executive questions still live in extracts? Book a session with Alluvium and we will name the standing list, the measures, and the fire drills allowed to stay one-off. Or start with a free Model Health check.