Running analytics as an ongoing practice

0:00 The analytics operating rhythm

Here's a hard truth about analytics. You can build a perfect implementation in January, and by June it will be quietly broken. Sites change, campaigns launch, developers refactor, new tags appear, and people leave. Analytics decays without ownership. In this lecture you'll learn how to run analytics like a product: roles and responsibilities, a daily, weekly, monthly and quarterly rhythm, the change log, how to audit an inherited property, how to automate the boring parts, and how to build a culture where data is trusted.

0:37 Why it matters

Why does this matter? Because every number your organization trusts depends on dozens of small things staying correct: tags, triggers, consent settings, UTM habits, definitions. Without routines, those things drift, and nobody notices until a big decision is made on a broken number. An operating rhythm turns analytics from a project into a product, with owners, routines and documentation. It's not glamorous. It's the reason some teams trust their data and others don't.

1:09 Roles

Start with roles. Think of a simple RACI: who's responsible, accountable, consulted and informed. The measurement plan: the analyst is responsible, the marketing lead accountable. The tracking plan and tags: the analyst or implementer responsible, the analytics owner accountable, developers consulted. The data layer: developers responsible, the tech lead accountable. Consent and privacy: analyst and legal responsible, the data protection lead accountable. Dashboards: the analyst responsible, the marketing lead accountable. In a small business, one person holds several roles. The point is that every activity has a named owner.

1:48 The rhythm

Now the rhythm itself. Daily, five minutes: a sanity check during active campaigns, and alert review. Weekly, thirty minutes: K P Is versus target, channel and landing page movers, anomaly notes. Monthly, two hours: the QA script on top journeys, the PII audit, the Unassigned review, dashboard commentary, and an access review for leavers. Quarterly, one day: measurement plan review, tracking plan clean-up, consent re-audit, the attribution and experiment roadmap, and stakeholder feedback. And on every release: developer notice, QA script, change log, annotation.

2:25 The change log

The cheapest, highest-value analytics artifact is the change log. Date, change, who, reason, and data impact. For example: the second of March, generate lead moved to a data layer push, by Hassan, because the thank-you page created false positives, and leads dropped as expected. When a number moves unexpectedly, the change log is the first place to look. Think of it like a flight recorder. You hope you never need it, but when something goes wrong, it tells you exactly what happened.

3:01 Auditing an inherited property

Agencies and new hires often inherit a messy property. Audit it in six areas. Access and ownership: who owns the account, who has admin, remove stale users. Configuration: retention, filters, referrals, cross-domain, key events, links. Implementation: duplicate tags, the container's state, data layer quality. Data quality: not set rates, Unassigned share, personal data leaks, revenue reconciliation. Consent: pre-consent firing, the C M P setup, non-Google tags. And reporting: which dashboards exist and who uses them. Score each red, amber or green, then prioritize by impact and effort.

3:39 Automate the rhythm

Now automation, which makes the rhythm sustainable. Daily, a BigQuery scheduled health query posts anomalies to a team chat. Weekly, a Data A P I script pulls the K P I table into a sheet, and an AI assistant drafts the first version of the commentary from the aggregated numbers, which the analyst edits and approves. Monthly, the PII scan and UTM leak query run on their own. And quarterly, a simple query lists every event name seen in the last thirty days, which you compare with the tracking plan.

4:18 Example 1: a solo consultant

First example, a simple one. A solo consultant runs analytics for three small clients. She can't do everything, so she shrinks the rhythm to fit. A weekly twenty-minute review per client, reading from an automated sheet. A monthly thirty-minute check with the QA script on the top journey and the PII scan. And a quarterly one-hour review with each client. Her rule: if a routine keeps getting canceled, shrink it rather than abandon it. A small rhythm that runs beats a big one on paper.

4:55 Example 2: a 90-day turnaround

Second example, a business case. A new analyst at a Dubai e-commerce company inherits duplicate tags, no change log, and a dashboard nobody trusts. Days one to thirty: audit, remove the hard-coded duplicate, fix payment gateway referrals, add the PII scan. Days thirty-one to sixty: rebuild the tracking plan, move purchase to a data layer push, launch the UTM builder, run a consent audit. Days sixty-one to ninety: a leadership dashboard on BigQuery summaries with weekly commentary, and the first geo holdout designed with the paid media team.

5:33 Measuring the rhythm

How did they know it worked? At day ninety, leadership meetings referenced the dashboard by name, and the change log explained the two numbers that moved unexpectedly that quarter. That's the real measure of an operating rhythm: people use the data in decisions, and surprises have explanations. Other signs: fewer escaped errors found by stakeholders, faster answers to how was this measured, and no single person who holds all the knowledge.

6:04 Watch me do it

Watch me run the quarterly event inventory. In BigQuery, I select event name and count events over the last thirty days, sorted by volume. Then I paste that list next to the tracking plan. Two lists come out of the comparison. Events that are live but undocumented: either document them or remove the tag. And events that are documented but show zero volume: investigate whether they broke, or retire them. This week, I find a leftover test event from a vendor and a signup event that stopped firing after a redesign.

6:44 Culture

Building culture matters as much as routines. Teach questions, not clicks: train stakeholders to ask good questions, and they'll find the reports. Celebrate decisions made with data, including decisions to stop something. Be honest about uncertainty: say directionally and illustratively when you mean it, because credibility comes from calibrated claims. Keep documentation close to the work, linked from the dashboard. And whenever an analyst, agency or freelancer changes, hand over a pack: plans, container export, UTM taxonomy, consent setup, dashboards, the change log and open issues.

7:21 Mistakes and recap

Common mistakes. No change log, so historical anomalies become mysteries. One person holding all knowledge and access. Audits that list eighty issues without prioritizing. Stakeholders who only see numbers, never the reasoning. And routines that exist on paper only. Recap: run analytics like a product, with owners, a daily-to-quarterly rhythm, a change log, prioritized audits, automation with human approval, and honest communication.

7:48 Try this now

Try this now. Create a change log and a monthly analytics routine for your organization or a client, and schedule the first review in your calendar. If you have the BigQuery export, run the event inventory query and compare it with your tracking plan. And if you're ready to go deeper, the Privacy-First Measurement course covers consent-aware collection and server-side tagging, and AI for Data Analysis and Decision Making will help you turn these numbers into defensible decisions.