The analyst is literal in the way good analysts are. When question planning is on, it declares how it read your question before it answers — a line like "I read 'sold most' as Quantity Sold" — and every planned answer carries a Read as: line under it naming the measures it used (by their plain names), the period it resolved — as dates, so "last quarter" shows exactly which days it ran over — and any filter it applied. If the reading is not the one you meant, rephrase and ask again; the line is there so a misreading is visible, never silent.
Some habits still get the best from it:
- Say what you mean, in your words. "Which products sold most this quarter?" is a complete question — you do not need the measure names of 2.9, though using them never hurts. The example chips above the box show the shape of questions it answers well.
- Name the period and the slice. "Supplier spend by supplier, this quarter" or "stock value by warehouse". It understands today, yesterday, the last 7 days, this week, this month, last month, this quarter, last quarter, year to date, this financial year, the last 30 days and the last 12 months — resolved against your company's calendar and timezone — and the dimensions of 2.8 are the slices it can cut. The day words mean exactly the days they say, on your company's local calendar: today is your today (the same "today" the dashboard's anomaly banner means), the last 7 days is the seven days ending today, and this week runs from Monday to today. A question that names a day is answered for that day — never quietly widened to the whole month — and the Read as line shows the exact dates. One honest difference to know: the banner's "7-day average" is the average of the seven days before today, because today is the day being judged against them; an Ask question over the last 7 days includes today. A breakdown answers with the top 10 entries and gathers the rest into an Other row. Balances and positions — "What do we owe suppliers?" — are answered as of the period's end date, and the Read as line says so.
- Ask "against what?" Comparisons are built in (2.10): "against last period" or "against the same period last year" turns a number into an answer. And ask the follow-up — "which customers drove that?" is a complete question, and usually the right second one.
- Rank in your own words. "Who is our best customer?", "top supplier", "worst-performing product" — the analyst reads best/top/worst as a ranking over a declared default (for customers, the default is invoiced revenue — what was billed, not what has been settled), and the reading line says so before the ranked list. Breakdowns are labelled with names, never bare record numbers: a by-customer answer names the customers.
- Two slices and trends work too. "Revenue by customer and branch" cuts by both at once, and "monthly revenue trend" (or weekly, quarterly) answers as a series over time. One field can also be compared against another: "Which products are below reorder point?" compares each product's stock value against its own reorder point. Stock transfers answer by the warehouse pair: "What moved from the main warehouse to the depot this month?" reads the quantity transferred between each pair of warehouses — each transfer counted once, where it arrived, in each product's stocking unit — and a reversed transfer reads as a transfer back along the pair.
- The indicator family is askable. "How many days do customers take to pay us?" (DSO), "What is our cash conversion cycle?", "What is our inventory turnover?", "What is our current ratio?", "What is our budget vs actual this year?", "Who is the cheapest supplier for SKU-007?" — each is a governed measure with its basis declared in the reading line (the day measures run over the trailing 90 days, as the answer states). When a question carries two honest readings, the analyst asks which you mean and lists the readings, rather than guessing.
- Delivery lead time is askable too. "What is our average delivery lead time?" reads the mean calendar days from order date to completed delivery, over delivered orders only: an order counts when its posted waybills cover everything that was ordered, and its delivery day is the completing waybill's date (a split delivery counts at the waybill that finished it, not the first one that left). Orders still partly delivered do not count yet. Orders imported from another system carry no delivery evidence, so they stay out of the average rather than reading as instant deliveries. The same measure is available in Explore for charts and breakdowns (by customer, by month, and more).