Write the measurement contract
A measurement system fails when teams use the same word for different events. Agree on definitions for inquiry, qualified lead, booked appointment, estimate, won job, revenue, and cancellation. Specify who can change a stage and when the change is recorded.
The contract should also describe exclusions: spam, existing customers, duplicates, vendors, wrong geography, and calls unrelated to the campaign. Publish the definitions where marketing, sales, dispatch, and finance can review them. A shared imperfect definition beats several incompatible dashboards.
- Name the source of truth for each field.
- Record timestamp, owner, and status history where possible.
- Version definitions when business processes change.
Event definition
Owner
Source of truth
Decision supported
Map the full customer journey
Start with how customers actually contact the business: phone, form, chat, booking tool, email, walk-in, referral, or repeat service. Map each path from first interaction to revenue and identify where identity or source information is lost. Include manual steps; hiding them does not make them disappear.
For calls, distinguish answered and missed calls, new and existing customers, service fit, and eventual outcome. For forms, preserve the original landing page and campaign context while avoiding unnecessary personal data. The goal is useful continuity, not surveillance.
- Draw one journey for urgent demand and another for considered purchases.
- Mark handoffs between marketing, intake, sales, and delivery.
- Document where an operator must correct or enrich data.
- 1
Discovery
- 2
Contact
- 3
Qualification
- 4
Booking
- 5
Outcome
Instrument only events you can use
Implement a small event vocabulary before adding every available platform event. Track meaningful actions such as qualified call, form success, appointment request, estimate accepted, and purchase or closed job. Distinguish an event that indicates intent from one that proves revenue.
Test events in the real customer path across mobile and desktop, including validation errors and thank-you states. Keep analytics and advertising conversions purposeful; duplicated page views or button clicks can inflate performance and create bad optimization signals.
- Maintain an event dictionary with name, trigger, parameters, and owner.
- Use stable identifiers only where consent and privacy practices permit.
- Audit tracking after site, form, phone, or CRM changes.
Handle source and attribution honestly
Source fields help answer where a customer first encountered the business, what prompted the current contact, and which touchpoint was present when they converted. Those are different questions. Choose a primary reporting view and label other views clearly rather than presenting competing totals as fact.
Use first-party records and platform data together, but do not assume a click caused a sale merely because it was recorded. Direct traffic, referrals, branded searches, call transfers, and offline recommendations can obscure the journey. Report unknown and untracked portions so instrumentation gaps remain visible.
- Preserve first-touch and latest-known source when the data supports both.
- Use UTMs consistently and document naming conventions.
- Never overwrite original source during later follow-up.
Discovery source
Conversion source
Assisted touch
Unknown or offline
Connect marketing records to revenue
The CRM or job system should receive enough context to evaluate lead quality: service, geography, status, owner, estimated value, outcome, and source. Keep marketing identifiers attached as a lead progresses, and define what happens when a household or company has multiple opportunities.
Reconcile records on a fixed schedule. Compare platform conversions with intake logs, CRM opportunities, completed work, and accounting where appropriate. Investigate discrepancies by category instead of “fixing” totals until they look familiar.
- Use a stable opportunity ID for pipeline joins.
- Separate booked value, invoiced revenue, and collected revenue.
- Record lost reasons and no-contact outcomes, not just wins.
Tracked inquiry
Qualified opportunity
Booked work
Won and delivered revenue
Build reports for decisions
A weekly operating report should show volume, quality, response, spend, and problems requiring action. A monthly report can include pipeline and revenue cohorts. Executives usually need fewer metrics: whether growth is profitable, whether capacity is available, and which decision is recommended.
Every metric needs a denominator, date range, source, and caveat. Annotate outages, promotions, staffing changes, seasonality, and long sales cycles. A clean chart without context invites confident but incorrect conclusions.
- Use qualified rate, booked rate, win rate, and cost by stage.
- Separate leading indicators from lagging financial outcomes.
- Put an explicit recommendation beside each recurring report.
- 1
Daily alerts
- 2
Weekly operations
- 3
Monthly cohorts
- 4
Quarterly allocation
Governance, privacy, and data quality
Assign ownership for tracking, CRM fields, dashboard logic, and access. Limit sensitive data to what the team needs, protect credentials, and document retention and consent practices with the appropriate advisor. Measurement should improve decisions without creating unnecessary privacy risk.
Run a data-quality checklist after releases and on a regular schedule: duplicate events, missing source, broken forms, disconnected phone paths, stale campaign names, and records stuck in one stage. Resolve systemic causes rather than repeatedly cleaning exports by hand.
- Use role-based access and remove access when responsibilities change.
- Keep a change log for tags, dashboards, definitions, and integrations.
- Create an escalation path for data incidents and broken conversion paths.
Close the loop with experiments
Measurement earns its cost when it changes behavior. For each experiment, state the hypothesis, audience, change, primary outcome, guardrail, required observation period, and decision rule. Preserve a control or baseline when practical and avoid calling noise a result.
Feed findings into campaigns, service pages, sales scripts, capacity planning, and the next budget review. If a result cannot change a decision, it probably does not belong in the core dashboard. The system should become simpler and more trusted over time.
- Keep a decision log linking action to evidence.
- Review lagging cohorts before closing long-cycle tests.
- Retire metrics that no longer support an operational decision.