Search
Enter at least two characters without personal information.

    LibraryArticles & FAQ Your projectRequest a quote
    LanguageEN
    Data and measurement

    Understand data and measurement

    Before the chart, define what a row represents and how a metric is calculated. Keep period, unit and rules for unknown values. Comparison quality depends on these definitions as much as on the tool used.

    43 concepts to explore with definitions and practical advice.

    Search the full glossary
    Tablet, laptop and notebook on a desk
    AI-generated illustration

    A reading path

    1. Describe data

    Identify each entry and its fields. Exports should retain relationships and units needed for correct reuse.

    • Data table

      Values organised in rows and columns with explicit headers. Preserve associations between cells and headings, and identify missing values and units to prevent misleading readings.

    • Unit of measurement

      The reference giving a value meaning, such as seconds, euros or kilograms. Publish it with the quantity and context; distinguish counts, percentages and denominators before comparing values.

    • Data export

      Extraction of content or data into a reusable format; it does not necessarily include code and settings needed for a full restoration.

    2. Define an observation

    Specify the observed action or population. A triggered event proves receipt or payment only when the corresponding journey is actually verified.

    • Event tracking

      Recording defined journey events, such as clicks, submissions or confirmations. Name each stage precisely: sending analytics or counting a click does not establish that the business operation succeeded.

    • Conversion rate

      Share of visits, sessions or people completing a defined action, depending on the chosen denominator. For example, 20 enquiries from 500 visits is 4%. State the population, period and action definition with the rate.

    • KPI

      Key Performance Indicator: metric chosen to track progress toward a clearly defined objective.

    3. Compare methodically

    Separate field measurement, laboratory measurement and hypotheses. Groups, periods and devices should be comparable for a difference to be meaningful.

    • Field data

      Measurements collected during real visits under observed users’ conditions. They describe a specific population and period; public CrUX data does not automatically cover all browsers or pages.

    • Synthetic monitoring

      Monitoring through automated scenarios run at defined times and locations. It can detect repeatable regressions and complements user observation without reproducing every usage condition.

    • Real User Monitoring

      Real User Monitoring (RUM): measurement of performance and errors encountered during real visits. Results depend on the devices, networks, pages and visits actually represented in collection.

    Useful distinctions

    Null value / Unit of measurement

    A missing value is not a measurement equal to zero. A unit explains how to read a present value. Define both rules before calculations and data imports.

    Attribution / Conversion rate

    Attribution seeks to distribute credit for an outcome among touchpoints. Conversion rate relates action count to a defined population. These metrics answer different questions.

    Educational example

    Use the vocabulary to make a decision

    Hands comparing two printed tables beside a calculator
    AI-generated illustration · Check period, population and calculation rules before comparing rates.

    Situation

    Two dashboards give different rates for the same enquiries.

    Before deciding

    Compare period, filters, action definition and denominator. Calculation differences can explain the gap without incorrect data.

    A concrete check

    Manually recalculate a small subset and record the chosen rule.

    From a word to an observation

    Give every rate a denominator

    Two percentages become comparable when their numerator, denominator, period and filters are known. Show volumes too when they change substantially.

    On a narrow screen, scroll the table horizontally to read all three columns.

    Illustrative observations to adapt to your project
    Observed signalWhat to checkWhat it means
    The rate rises but enquiries fallRead enquiries and visits together for each period.A smaller denominator can increase the rate.
    Two tools report different visitsCompare sessions, visitors, exclusions and time zones.Units and collection rules may differ.
    An average hides device differencesSeparate relevant groups and their volumes.An overall result may hide a mobile difficulty.

    Next step : Write the formula and exclusions beneath the chart, then compare the same unit and duration.

    Technical reference : W3C — Data on the Web Best Practices. The scenarios and numerical examples above are educational, not measurements of this website.

    Examine the interpretation

    Mean and median tell different stories

    Fictional teaching case. Open each interpretation to examine its scope; there is no score or professional assessment.

    1. Values ordered
    2. Summary named
    3. Limit retained

    Fictional set of five durations in minutes: 1, 1, 1, 1 and 16. Their sum is 20. These values illustrate two calculations only.

    The five fictional durations, in minutes
    1. Case 11 min
    2. Case 21 min
    3. Case 31 min
    4. Case 41 min
    5. Case 516 min

    Mean: 4 min · Median: 1 min · Largest observed value: 16 min.

    Which summary are you trying to read?

    The mean is 4 minutes

    The calculation is 20 ÷ 5. The 16-minute duration contributes strongly to the mean, even though four of the five observations are one minute.

    The median is 1 minute

    The values are ordered, and the third of five, in the middle, is 1. This median does not make the 16-minute duration unimportant.

    A useful next step : Name the metric and retain useful values or distribution details. Neither summary alone establishes a maximum duration or overall quality.

    Frequently asked questions

    Can unknown data become zero?

    Only when a justified rule actually gives it that meaning. Otherwise retain explicit absence to avoid distorting totals, averages or filters.

    Are more metrics always needed?

    Add a metric if it informs a decision. Keep its definition and owner; a wider dashboard cannot compensate for poorly understood data.

    Primary reference documents

    These documents explain technical concepts. Reading paths and examples are editorial methods to adapt to your project.

    Related topics