How do I build analytics around decisions instead of every possible event?

Build a decision-led analytics plan by defining owners, questions, outcomes, events, properties, data quality, privacy, reporting and action thresholds.

Digital Marketing
16 November 2026Updated 16 Nov 20266 min readBukhosi Moyo

Quick Answer

List the recurring decisions teams must make, who owns each one, what options are available and what evidence would change the action. Build a measurement tree from business outcome to leading indicators, then instrument only the events and properties required for those questions, diagnostics and quality checks. Use recommended events where they fit, maintain a tracking specification and test every release. Give each report a decision cadence and threshold, and retire data that has no owner, purpose or lawful retention need.

Key Takeaways

  • Start with named decisions and accountable owners.
  • Map outcomes to leading and diagnostic measures.
  • Instrument the smallest sufficient event model.
  • Test data quality, consent and business definitions.
  • Give every report an action rule and lifecycle.

Want the full breakdown? Scroll below.

People reviewing work together at a desk with laptops
On this pageJump to a section
  1. 1Inventory recurring decisions
  2. 2Write the decision question
  3. 3Build a measurement tree
  4. 4Define the entity model
  5. 5Choose the smallest event set
  6. 6Define properties deliberately
  7. 7Write a tracking specification
  8. 8Implement consent and minimisation
  9. 9Test the data layer
  10. 10Reconcile with systems of record
  11. 11Add data-quality measures
  12. 12Design reports from actions
  13. 13Set thresholds and review rules
  14. 14Train users on definitions
  15. 15Retire unused data
  16. 16Document transformations
  17. 17Plan analysis before the result
  18. 18Connect measurement to experiments
  19. 19My take: analytics is a decision service
  20. 20Run a decision-led pilot
  21. 21Frequently asked questions
  22. 22Measure what changes a decision
  23. 23Related guides
  24. 24Sources

Share this article

Bukhosi Moyo

Growth Partner

Need help growing your company?

We build SEO-first websites and growth systems for South African businesses.

Get Started

Start with the recurring decisions the business must make, not the interactions a tool can collect. Name the owner, options, evidence and action threshold for each decision. Build a measurement tree, then instrument only the outcomes, leading indicators and diagnostics needed to support it.

This is a marketing analytics and digital strategy problem. An enormous event list can create more disagreement while making the important data harder to trust.

Google provides recommended events with defined names and parameters for common business actions, but implementation still needs to reflect the actual business process Source: Google for Developers.

Inventory recurring decisions

Ask marketing, sales, product, service and finance what they decide weekly, monthly and quarterly. Capture options, owner, deadline and consequence.

Exclude vague requests to understand users better until they become a decision. Rank by business value and reversibility.

Write the decision question

Turn each decision into a question with population, outcome, time window and comparison. For example, which acquisition cohort produces qualified value after the full sales cycle?

Avoid dashboard questions such as what was traffic last month unless an action follows. State what evidence would change the choice.

Build a measurement tree

Start with the business outcome, then map leading indicators, operating drivers and diagnostic measures. Identify assumptions between levels.

Do not promote every diagnostic event to a key performance indicator. A field error can explain form loss without becoming an executive target.

Mark financial and CRM sources outside analytics.

Define the entity model

List users where permitted, organisations, sessions, leads, opportunities, orders, subscriptions and products. Define stable identifiers and relationships.

Avoid combining person and company metrics in a B2B journey. Document anonymous and consent gaps.

Use transaction and lead IDs without exposing sensitive data.

Choose the smallest event set

For each event, state the decision it supports, trigger, properties, owner and retention need. Prefer existing recommended definitions when they match.

Google's event reference includes common actions for lead generation, ecommerce and subscriptions Source: Google for Developers.

Do not track decorative clicks without a diagnostic purpose.

Define properties deliberately

Capture dimensions needed for segmentation, quality and reconciliation, such as service, form type, product, value and customer type. Use controlled vocabularies.

Do not send names, email addresses or free-text form content into analytics. Review each property for privacy and cardinality risk.

Version schema changes.

Write a tracking specification

Document event name, business definition, trigger, parameters, consent condition, examples, exclusions, data destination and test cases.

Keep the specification in source control or another governed system. Link implementation changes to releases.

Implement consent and minimisation

Collect only data needed for stated purposes and configure consent behaviour before tags fire. Define retention, access and deletion processes.

Avoid using analytics as a shadow customer database. Review advertising activation separately from measurement.

Obtain South African and other applicable legal advice.

Test the data layer

Validate triggers, values, currency, identifiers, duplication, sequence and failure states. Test mobile, browsers, consent choices and cross-domain journeys.

Compare client events with server or system records where appropriate. Do not rely only on a real-time debug view.

Keep automated regression tests for critical events.

Reconcile with systems of record

Compare analytics leads, orders and values with CRM, commerce, payment and finance records. Explain expected gaps and timing.

Define whether reports use event time, processing time or outcome time. Deduplicate retries and refunds.

No executive metric should depend on an unreconciled event.

Add data-quality measures

Monitor missing events, unexpected volume, invalid parameters, duplicate IDs, consent state and delayed feeds. Set alert thresholds.

Show freshness and known incidents on reports. A precise chart built from incomplete data should not drive a high-consequence decision silently.

Assign incident ownership.

Design reports from actions

Give each report an audience, decision, cadence, comparison and action rule. Put outcomes and guardrails before diagnostic detail.

Use annotations for launches, campaigns and tracking incidents. Avoid dozens of tabs with no narrative or owner.

Provide drill-down only where it changes investigation.

Set thresholds and review rules

Define when evidence triggers monitor, investigate, test, scale or stop. Include minimum sample and maturity windows.

Avoid reacting to daily noise in a long sales cycle. Use ranges and uncertainty where appropriate.

Record decisions made from the report.

Train users on definitions

Create a concise metric dictionary and examples. Teach teams the difference between observed events, modelled attribution and system outcomes.

Resolve definition disputes through governance, not by creating another dashboard. Keep training aligned with the live specification.

Review new decision owners.

Retire unused data

Audit events and reports quarterly. Remove fields with no owner, decision or diagnostic use after checking dependencies and retention obligations.

Retirement reduces implementation cost, privacy exposure and interpretation noise. Preserve historical definitions where comparisons need them.

Announce breaking changes.

Document transformations

Record how raw events become sessions, users, cohorts, channels, leads and financial measures. Version filters, joins, attribution choices and currency logic. A metric can remain visually identical while a transformation changes its meaning. Require review for material changes and annotate reports. Keep enough lineage that an analyst can trace an executive total back to source systems without exposing customer data.

Plan analysis before the result

For each decision, define segments, comparisons, exclusions, maturity window and minimum evidence before viewing performance. This reduces the temptation to search dozens of dimensions until one supports a preferred action. Preserve exploratory analysis, but label it and validate new hypotheses on later data. Use qualitative customer or sales evidence when event data cannot explain motivation.

Connect measurement to experiments

Analytics can identify patterns and prioritise questions, while controlled changes test whether an action improves outcomes. Maintain an experiment register with hypothesis, exposure, primary metric, guardrails and decision. Avoid launching overlapping changes that share the same audience and outcome without a contamination plan. Feed results back into the measurement tree and remove indicators that repeatedly fail to predict value.

My take: analytics is a decision service

My take is that analytics succeeds when a team can make a better decision at the right time and explain the evidence used. Collection volume is an infrastructure cost, not a measure of insight.

Every event should earn its place.

Run a decision-led pilot

Choose one valuable recurring decision and implement its measurement tree, events, reconciliation, report and review cadence end to end.

Observe whether the owner changes an action. Fix definition and data gaps before adding more decisions.

Use the pilot to define governance.

Frequently asked questions

How many events should a website track?

There is no universal number. Track the smallest sufficient set for decisions, diagnostics, quality and required reporting.

Should every button click be an event?

No. Track an interaction when it represents an outcome or helps diagnose a material journey.

Can recommended events replace a measurement plan?

No. They support consistent implementation, while the business must still define decisions, outcomes and properties.

Who should own analytics definitions?

Business and data owners should agree definitions, with implementation specialists responsible for faithful collection and quality.

Measure what changes a decision

If your business needs a governed decision-led measurement plan, get in touch for marketing analytics.

Related guides

For further background, read our marketing dashboard metrics and attribution for lead generation. The analytics glossary explains the terminology used in those guides.

Sources

Share this article

Bukhosi Moyo

Written by

Bukhosi Moyo

CEO & Founder

Bukhosi is the founder and lead SEO strategist at Symaxx. He architects search-first digital systems for South African businesses, combining technical engineering with commercial strategy to build long-term organic assets.

Feedback

Was this helpful?

Tell us how this article felt in one click.

Back to Insights

Need help executing this strategy?

Connect your campaigns, landing pages and enquiry handling in one practical plan.