Establish your measurement baseline

Make measurement serve a decision

The first analytics episode answers a practical question: what evidence can help us choose the next useful piece of work? A dashboard is useful when its numbers have clear meanings and lead to decisions. Start with the journey chosen during the foundations, rather than collecting every metric offered by every platform.

Basic measurement should have been enabled before launch. This week makes that setup visible, verifies it and establishes a baseline. If measurement started late, state the actual collection date. Visits that were never recorded remain unavailable. They are not zero, and they cannot be reconstructed by guessing from social reach.

1. Name the outcome and its source

For an audience and affiliate business, a confirmed email subscription can indicate continuing interest; provider-reported paid referrals and approved commissions describe commercial outcomes. For a service business, distinguish an enquiry, a booking, attendance and a new customer. For a digital product, distinguish checkout, payment, delivery and first use.

Choose one or two headline outcomes for the current stage. Record who owns each source. An email provider confirms subscriptions; a scheduler confirms appointments; a payment system confirms transactions. A website click is evidence of an action on the page, with a narrower meaning than a completed event in another service.

Write a simple definition before building a chart. For example: “Unique email subscriptions confirmed by the provider during the reporting week, excluding controlled tests.” This describes a count of new confirmations. It does not mean net list growth or the number of paying customers.

2. Verify the measurement journey

Create an event dictionary with the event name, meaning, unit, source, timestamp, relevant identifier and known limitations. Use names such as resource_link_clicked and booking_confirmed when they accurately describe the event. Record a download click as a click; it does not prove that the resource was opened or used.

Run a controlled test through the active journey. Check the expected event, its timestamp and any duplicate records. Mark test activity and exclude it from business results. Where a provider sends repeated notifications, use its stable event or record identifier to avoid counting one completion several times.

Keep personal information out of campaign tags. A content identifier can describe the post or guide that led to a visit. It should not contain the visitor's name or email address. When a shared profile link prevents attribution to an individual post, report the known channel and leave the post unknown.

3. Establish a comparable baseline

State the start and end dates, reporting time zone and any outages. Retain the original export or a dated source snapshot so a later review can reproduce the count. Separate observed zero, unavailable data and activity excluded as tests. These states lead to different decisions.

Show counts before rates. A conversion rate also needs a denominator drawn from the same eligible population and period. Do not divide confirmed bookings by unrelated Instagram reach and describe the result as a tracked person-level funnel. Platform reach, website sessions and provider records are different units unless a justified connection exists.

Small samples still help test clarity, reliability and collection quality. They usually give limited support for ranking close alternatives. Add reader questions and usability observations to explain what remains uncertain.

4. Turn the baseline into one investigation

Choose the largest useful question the evidence leaves open. It might concern a confusing destination, missing subscription confirmations or the absence of any reliable referral report. Assign an owner, a next action and a review date. Repairing measurement can be the correct first improvement.

For the public episode, explain what was measured, how it was checked, what the result means and what it does not establish. Publish aggregated evidence. Keep identifiable customer records in the private operating system.

Illustrative case

An educator sees resource-link clicks in website analytics but has no evidence that readers used the worksheet. They retain the click metric, add a short optional question about completion and choose a small usability check. Their next action addresses uncertainty without upgrading clicks into invented customer success.

If Clearweft has no live results when this lesson is published, teach the measurement map and show the missing-data status. Replace the review placeholders only after the source records are checked.

You are finished when: The review includes a decision, owner and next review date without inventing missing results.