Research & insights

Source-backed reporting, original comparisons and practical analysis across our eight subject areas.

October 19, 2026 · Business & Industry

One order can have several dates. ORIGINAL ORDER: When was the transaction created? LATER EVENT: When was the return or reversal processed? REPORT VIEW: Which date groups the analytical result? Original explanatory diagram.

Returns and Reversals: Which Reporting Date Does the Number Describe?

Separate original order dates from later returns and reversals so sales trends and customer cohorts use compatible reporting conventions.

Read the article →

October 19, 2026 · Countries & Cities

Response is not coverage. TARGET: The population the survey aims to describe. FRAME: Who can enter the selection process? RESPONSE: Who supplies information among eligible units? Original explanatory diagram.

Why a High Survey Response Rate Does Not Prove Full Coverage

Separate survey response from population coverage and learn what a strong participation rate can—and cannot—tell you about a dataset.

Read the article →

October 19, 2026 · Economy & Finance

One scale factor preserves ratios. ORIGINAL: 120 to 126 is a five-percent increase. REBASE: Multiply both values by 100 / 120. RESULT: 100 to 105 is still a five-percent increase. Original explanatory diagram.

Rebasing an Index: New Levels, Same Percentage Change

Work through an index rebasing example and see why changing the reference level preserves percentage growth when the series is otherwise unchanged.

Read the article →

October 19, 2026 · Technology & AI

Read the disclosure, then verify. IDENTITY: Which model, version and document? EVIDENCE: Which task, population and evaluation? LIMITS: Which uses remain untested or excluded? Original explanatory diagram.

How to Read a Model Card Without Treating It as Certification

Read a model card for its intended uses, evaluation evidence and omissions, with a practical worksheet for deciding what still needs checking.

Read the article →

October 19, 2026 · Healthcare & Public Data

New cases and existing cases differ. INCIDENCE: New cases over a defined time framework. PREVALENCE: Existing cases at a point or during a period. CHECK: Case definition, risk population and time unit. Original explanatory diagram.

Incidence and Prevalence: New Cases and Existing Cases Are Different

Separate incidence from prevalence by checking which cases, population and time period are counted, with a clearly fictional example.

Read the article →

October 19, 2026 · Sports & Statistics

Aggregate before dividing. SCORING: Total eligible runs / eligible overs faced. CONCEDING: Total eligible runs / eligible overs bowled. NRR: Scoring rate minus conceding rate. Original explanatory diagram.

Cricket Net Run Rate: Aggregate Runs and Overs Before Comparing

Calculate cricket net run rate from aggregate runs and eligible overs, and avoid decimal-over and unweighted-match-average errors.

Read the article →

October 19, 2026 · Real Estate

Vacancy depends on the inventory. TOTAL STOCK: All units in the defined housing account. RENTAL INVENTORY: Relevant occupied and for-rent categories. HOMEOWNER INVENTORY: Relevant occupied and for-sale categories. Original explanatory diagram.

Rental and Homeowner Vacancy Rates Use Different Inventories

Compare rental and homeowner vacancy rates using their proper inventories, and see why all vacant units do not equal homes available to rent or buy.

Read the article →

October 19, 2026 · Products

Three different display stages. FRAME RATE: Images generated by the rendering system. REFRESH RATE: Display updates per second. PIXEL RESPONSE: A transition under a named test method. Original explanatory diagram.

Refresh Rate, Frame Rate and Pixel Response Time Are Different

Separate monitor refresh rate, generated frame rate and pixel response time, with simple timing calculations and clear limits on specification claims.

Read the article →

October 19, 2026 · Business & Industry

Read the order average with its total. NUMERATOR: Which revenue and adjustments are included? DENOMINATOR: Which orders belong to the period? MIX: Which order sizes became more or less common? Original explanatory diagram.

A Higher Average Order Value Can Hide Fewer Orders

See how average order value can rise while order revenue falls, and compare the ratio with order counts, mix and reporting definitions.

Read the article →

October 19, 2026 · Countries & Cities

Do not turn unknown into zero. ZERO: A numeric observation under the definition. UNAVAILABLE: No usable value supplied. SUPPRESSED: A value withheld under release rules. Original explanatory diagram.

Missing, Suppressed and Zero: Three Different Public-Data States

Keep missing and suppressed values separate from zero when reading public datasets, with a practical example for tables and charts.

Read the article →

October 19, 2026 · Economy & Finance

A statistic has two time fields. OBSERVATION: Which quarter or year is measured? RELEASE: When was this estimate published? VINTAGE: Which release supplied each comparison value? Original explanatory diagram.

Why GDP Estimates Change: Read Releases as Data Vintages

Separate a GDP observation period from its release date and compare revised estimates using a transparent data-vintage worksheet.

Read the article →

October 19, 2026 · Technology & AI

Original explanatory diagram: training fits transformations and parameters, validation selects settings, and final testing evaluates held-out cases.

Data Leakage: Why a Clean Test Split Matters More Than a Strong Score

Trace data leakage through preprocessing, tuning and repeated records, then build a test-split audit that keeps evaluation data separate.

Read the article →

« Previous 1 … 3 4 5 6 7 … 12 Next »