How to Query Business Systems for Accurate Data
Learn to define the reporting period, units, scope, and document status in every data query, so you can retrieve consistent figures, reconcile discrepancies, and choose the right source for each business task.
Looking up figures in management software seems simple, but a question missing conditions can produce a result that cannot be used. “How much revenue this month?” for instance, does not say whether revenue is counted by order, by invoice, or by cash received. When two people run the same query but understand it differently, reconciliation easily turns into an argument about numbers instead of a check of the business process. Asking clearly from the start lets users control the data scope and know where to verify the result.
A question needs to fix the period, unit, scope, and status
Before querying, write the question as a data request rather than phrasing it as a spoken exchange. The four elements below should all be present; even if the system has default values, restate them to avoid misunderstandings.
- Data period: state the start and end dates, for example “from 01/07 through 31/07”. For document data, also state whether the period is defined by invoice date, posting date, or payment date.
- Unit of measure: specify Vietnamese dong or foreign currency, quantity or value, before tax or after tax. For inventory, state the unit such as piece, carton, or kilogram when the same item has several conversion methods.
- Scope: the whole company or one branch, warehouse, department, or customer group. If only one unit is needed, do not ask for “the company” and then exclude rows by guesswork.
- Document status: whether to include only approved, posted, or completed documents; whether drafts, cancelled documents, and adjustments are included. Status names must match the configuration actually in use at the business.
A question with enough conditions might read: “Give the total pre-tax value, in Vietnamese dong, of sales invoices posted at the Gia Lai branch, by invoice date from 01/07 through 31/07; exclude cancelled invoices.” It is longer, but each phrase removes one possible alternative reading.
Why a vague question produces two different numbers
The difference usually lies not in the addition but in the dataset selected before calculating. For the same “July revenue”, one person may take posted invoices while another includes draft invoices. One report uses pre-tax values and the other adds tax; or one looks at the whole company while the other sees only the branch they manage.
- The transaction date and the posting date of a document can fall in two different periods.
- Sales orders, invoices, and receipts reflect different stages of a transaction.
- Draft, cancelled, or adjusted documents change the total if they are not explicitly excluded.
- Access rights by branch or department can mean each person sees a different data scope.
So when a number arrives, do not rush to conclude that the system is wrong. First compare the question itself: are both sides using the same date basis, the same document type, the same unit, and the same status? If any of these conditions is unclear, two different numbers can both be correct under two different readings.
Frame the question around the business check needed
Beyond the four required conditions, state how the figure will be used. The number needed to reconcile receivables is not the same as the number needed to track sales progress; the inventory figure used for a stock count is not the same as the value issued from the warehouse in a period. When the purpose is clear, the person querying is more likely to choose the right source documents and avoid pulling a familiar report that does not fit.
“List outstanding receivables as of the end of 31/07 by customer, in Vietnamese dong, within the Gia Lai branch; count only posted invoices, after deducting payments and adjustments recorded up to that point.”
For a request like this, the result should come with a detailed list, not just a total. The list helps reveal customers that were missed, documents recorded in the wrong period, or payments not yet deducted. If an AI assistant is used to query data, still ask it to show the filter conditions and the link to the original documents instead of accepting only a summarized answer.
Verify by asking again in a different way
A figure is more usable when it can be traced back to detailed data. After the first query, change the wording or the order of checks to see whether the result stays consistent. There is no need to repeat the identical sentence, because that only confirms the same calculation.
- Ask for the total first, then the detail: starting from total revenue, request the list of invoices that make up that total, with date, status, and value.
- Split along one data dimension: break the figure down by branch, by date, or by owner and add the parts back to compare with the original total.
- Rephrase the conditions: replace “July revenue” with “total pre-tax value of posted invoices with an invoice date in July”.
- Pick a few documents to compare: open the original documents, check the amount, tax, status, and scope, and see whether they appear correctly in the list.
If two ways of asking still give two different totals, save both questions, the filter conditions, and the detailed lists. Comparing the difference document by document is more concrete than asking “why is this wrong”. Once the condition causing the difference is identified, the business should standardize how frequently used figures are named and queried.
What to do now
Start with the figures queried most often, before expanding to other reports.
- Pick one familiar question and add the period, unit, scope, and document status.
- Save it as a shared template, with a clear example of documents that are counted and excluded.
- Query again by requesting the detailed list, then compare a few original documents.
A good question does not need to be long, but it must carry enough conditions for someone else to repeat it and get the same data scope. VASCOM recommends keeping these templates inside the internal query process, so that reconciling figures starts from documents rather than from impressions.