Data and performance

Turn data into readable indicators, measure meaningful performance and build dashboards that genuinely support decisions.

62 articles · Latest article updated

Search covers every article in this topic, including the archive and article text.

Understand the topic

A practical guide to this topic

Measurement is useful when the definition, source and decision behind each indicator are explicit. These analyses explore data quality, trend interpretation and monitoring of real digital journeys. A performance measure becomes actionable when its population, time window, unit and exclusions are known. Authentication signals, network latency and user failures describe different events. Combining them without definitions can hide the dependency needing correction.

What to examine
  • Define units, scope, period and owner before comparing figures.
  • Separate collection changes, behaviour changes and the effect of an intervention.
  • Connect technical metrics to tasks and outcomes experienced by users.
  • Segment field metrics by page and device, then locate the responsible element or interaction.
  • Show counts, period and limitations before interpreting a rate or recommending action.
  • Define entry, return and horizon, distinguish no return from an unobservable period and compare cohorts with completed observation windows.
  • Verify grain, keys and cardinality before totaling sales, costs or conversions.
  • Avoid shifted reporting days, duplicate counts and inconsistent comparisons between exports and dashboards.
  • Compare performance using consistent populations, units and time windows.
  • Prevent silent replacement of absent values from changing the meaning of metrics.
Useful questions
  • Which event creates this data?
  • Which decision changes when the measure moves?
  • Can the calculation and source be reproduced?
  • Which change comes from traffic or collection rather than the fix?
  • What is the denominator?
  • Does a blank cell mean zero?
  • Does DISTINCT fix an incorrect join?
  • Is converting everything to UTC enough?
  • Can p95 values be averaged?
  • Should missing values become zero?

Illustrative decision example

Situation
Fictional example: these inputs describe no customer or observed result. A story announces a 100% increase when the result moves from one success in twenty to two in twenty.
Check
Show both counts, the 5% and 10% rates and the five-percentage-point gap. Describe period comparability and the analysis method.
Act
Separate observed change, uncertainty and interpretation. Relative doubling does not establish that the studied intervention caused it.

Mastering Data Storytelling: Turning Analytics into Strategic Business Actions

Small samples: communicate measurement uncertainty

NIST/SEMATECH — Confidence intervals for proportions

Qualify the number before interpreting it

Check freshness, completeness and correctness separately. A decline can come from business activity or collection failure; partial periods and missing values should be visible before comparison.

Which reconciliation establishes comparable periods and denominators?

Explore the methods

Read data · verify · decide

Engagement: add counts before rates

Keep percentages connected to their definition and period. Add sessions and engaged sessions, then calculate the combined rate.

Combined rate
54%540 / 1,000 sessions
Simple average of rates
45%(60% + 30%) / 2
Combined bounce rate
46%1 − 54%, with matching definitions

Fictional teaching case — figures describe neither CIE nor a client.

Questions to support your decisions

Find answers and practical checks in the analyses in this category.

Explore the archive

Further analysis in this topic beyond the recent articles listed below.

References: data, uses and limitations

Documentation reviewed on .

References: data, uses and limitations
Source and purposeVerified data or benchmarkHow to use itLimitation to retain
Google · Core Web Vitals

Distinguish loading, responsiveness and visual stability.

 Good thresholds are LCP ≤ 2.5 s, INP ≤ 200 ms and CLS ≤ 0.1 at the 75th percentile, segmented by mobile and desktop. Separate field measurement from laboratory tests. Record period, URL, device, population and sample size before comparing changes.

Explore the method →

 These thresholds are not measured results for this site. A good lab score alone does not establish visitor experience.

Media: sanitization and evidence

Compare the supplier’s deletion promise with evidence

A data-deleted clause may cover an application operation, copies or media retirement. Request explicit scope and reconciliation between items delivered to a provider and those actually processed.

  • Identify method, technique, tool and medium behind each attestation.
  • Request verification results and the validation decision separately.
  • For cloud storage, examine encrypted boundaries, key management and copies under contractual responsibilities.

Read the methods, evidence points and annotated situations →

NIST · SP 800-88 Rev. 2 · Guidelines for Media Sanitization · Final · 2025-09-26 · consulted 2026-10-23

Core Web Vitals: thresholds and field data

Measure performance without losing context

Keep collection metadata and separate field assessment from lab diagnosis.

  • Compare the same device, scope and documented periods.
  • Do not turn missing data into zero.
  • Verify functions and accessibility after the fix.

Read the thresholds and try the p75 reader →

Google · About PageSpeed Insights · Living documentation · consulted 2026-10-23

A/B allocation: data and SRM

A/B tests: check allocation before deciding

An unexplained discrepancy may arise from assignment, the executed journey, logs or analysis filters. SRM supports investigation; it does not measure product improvement.

  • Check intended allocation and unit stability before comparing conversions.
  • Do not interpret absence of a signal as proof of overall validity.
  • Document the quality check, measured effect and deployment decision separately.

Explore the allocation check and its limitations →

Microsoft Research · Diagnosing Sample Ratio Mismatch · Published 2020-09-14 · consulted 2026-10-23

NIST/SEMATECH · Binomial distribution · e-Handbook §1.3.6.6.18 · consulted 2026-10-23

Join cardinality, NULL and aggregates

Reconcile rows before interpreting a dashboard

A metric may change with identical source events: a join creates pairs, a filter removes rows and null measurements reduce contributions. Keep a reference at the source grain and inspect retained identifiers.

  • Compare COUNT(*), distinct identifiers and non-null measurements.
  • Inspect multiple matches before summing.
  • Test the active predicate in ON and WHERE on an unmatched case.

Inspect pairs and aggregates in the workshop →

PostgreSQL 18 · Table expressions · PostgreSQL 18 · documentation reviewed 2026-10-23

PostgreSQL 18 · Aggregate functions · PostgreSQL 18 · documentation reviewed 2026-10-23

PostgreSQL 18 · Comparison functions · PostgreSQL 18 · documentation reviewed 2026-10-23

Reference material

Primary guidance published by the organizations responsible for these methods and standards.