Blog · AI and self-service analytics
Interactive business analytics follows successive questions from a company total to the rows that explain it. This worked revenue example shows how to compare regions, customers and products while preserving a checked total.
Interactive business analytics means following a business question through successive, checked comparisons until the change is understandable. A revenue decline may lead from the company total to a region, then to a customer and product. Each answer should preserve the original scope and show how its rows add back to the total.
The interaction is the sequence of questions, not a particular screen design. Start with what changed. Ask where the change occurred, which transactions explain it and what additional evidence would help decide an action.
That approach resembles exploratory analysis: NIST describes exploratory data analysis as a way to gain insight, uncover structure and test assumptions. For a business reader, the practical discipline is to investigate the file before settling on an explanation.
A recurring report may tell a sales leader that revenue fell. The next useful answer identifies the accounts behind the fall and the accounts moving in the opposite direction. That is a different job from arranging recurring measures into dashboard layouts.
Begin with an uploaded sales workbook containing transaction date, customer identifier, region, product, invoice identifier and net revenue. Include quantity, credits and discounts when available. Keep the original export intact for checking.
Write down whether revenue means orders, invoices or recognized revenue. Use the same definition throughout the investigation. The SEC's guide to financial statements distinguishes revenue earned during a period from associated costs and explains that returns and allowances reduce gross revenue to net revenue. A question about revenue should not silently become a question about profit; see revenue versus profit for that boundary.
Choose complete comparison periods and a consistent region assignment. If a customer moved between territories, decide whether both periods use the original assignment or a restated one. Otherwise a territory transfer can appear to be growth in one region and a decline in another.
An illustrative US distributor compares August and September 2026. These summarized workbook rows contain the full revenue scope for the example.
Revenue change = Current-period revenue - Prior-period revenue
Percentage change = Revenue change / Prior-period revenue
| Region | Customer | Product | August revenue (USD) | September revenue (USD) | Change (USD) |
|---|---|---|---|---|---|
| East | Harbor | Standard | 120,000 | 80,000 | -40,000 |
| East | Harbor | Premium | 60,000 | 70,000 | 10,000 |
| East | Lake | Standard | 90,000 | 70,000 | -20,000 |
| West | Ridge | Standard | 130,000 | 140,000 | 10,000 |
| Total | 400,000 | 360,000 | -40,000 |
Revenue fell by 360,000 - 400,000 = $40,000, or -40,000 / 400,000 = -10.0%. The first answer is a measured decline, not yet a reason for it.
Regroup exactly the same rows by region.
| Region | August revenue (USD) | September revenue (USD) | Change (USD) | Percentage change |
|---|---|---|---|---|
| East | 270,000 | 220,000 | -50,000 | -18.5% |
| West | 130,000 | 140,000 | 10,000 | 7.7% |
| Total | 400,000 | 360,000 | -40,000 | -10.0% |
East declined by more than the company because West offset part of the loss. Percentages here are rounded to one decimal place; dollar calculations use unrounded values. The check is -50,000 + 10,000 = -40,000, matching the original total.
Do not discard West once East becomes the focus. Keep its offset visible in the company explanation, even while the next question uses only East's rows.
Harbor totals 180,000 in August and 150,000 in September, a $30,000 decline. Lake contributes the remaining $20,000 decline. Together, -30,000 - 20,000 = -50,000, which reconciles to East.
Harbor's Standard product lost $40,000 while Premium gained $10,000. The useful statement is that Standard's decline was partly offset by Premium growth. Calling Harbor a uniform decline would hide an important product difference.
The same question pattern works for location and category comparisons in automotive business analytics. Keep each page's business context separate, but retain the common check: the child rows must add to their parent.
This workbook establishes where the decline sits. It does not show whether Harbor ordered less, received shipments later or paid a lower price. Those explanations require quantities, prices, delivery records or documented account information.
A useful next question names the missing field: compare units and net revenue for Harbor's Standard product, or inspect the invoice dates for orders near month-end. If the export lacks those fields, say what cannot be established. An explanation should not convert a plausible story into a recorded fact.
Revenue movement can guide the sales review without proving its cause. The account owner can check customer circumstances, operations can examine recorded fulfillment timing, and finance can review credits and cutoffs. Each owner should return evidence tied to the relevant transactions.
For each answer, retain the measure definition, date range, included rows and reconciliation. A short log can record the question, answer, remaining uncertainty and action owner. This makes the investigation usable in the next management review.
When using AI analytics, keep the calculation evidence available alongside the written explanation so reviewers can check both.
Finance owns the starting revenue total. Sales operations owns customer and territory mappings. The business manager owns the decision to investigate or act. A business analytics center of excellence can formalize those responsibilities when many teams use the same measure.
Upload a sales workbook: Covirage maps the columns, and its tools compute and check the comparisons behind your successive business questions. The external AI model explains the results with every figure traced to its rows. Create a free Covirage account and bring a sample of your own data.
See self-service analytics for the supporting business workflow.
A dashboard organizes selected measures into a recurring view. Interactive analysis follows the questions raised by those measures, including comparisons that were not anticipated when the view was designed. Both are useful, but a changing question does not always require a new dashboard.
Specify the measure, comparison period and business dimension. Ask which customers explain a regional revenue decline, for example, rather than asking why sales are bad. A useful answer names the contributing rows, preserves offsets and identifies what evidence is still missing.
It can identify the recorded customers, products and transactions behind the decline. It may also show changes in quantity or realized price if those fields exist. It cannot establish a customer's motives or an operational cause that was never recorded in the workbook.
Covirage's tools compute and check the figures from the supplied file. The external AI model explains those results with every figure traced to its rows. Keep the distinction clear when reviewing an explanation: the calculation and its business interpretation serve different purposes.