Purchase disappears or starts duplicating
The campaign report still runs, but optimisation and revenue use a faulty signal.
Data quality monitoring · GA4, GTM and BI
Monitor the data before an error reaches the report.
I design controls, alerts and a response rhythm for critical events, conversions, campaigns and dashboards, so silent measurement failures are found before the monthly review.
Monitoring covers more than detection. We also agree priority, ownership, verification and the condition for closing each incident.
Control and responsean alert starts the process; it does not finish it
Risk-based coveragedifferent rhythms for revenue, leads, consent and dashboards
Recurring or continuousa standalone review or an ongoing analytics support module
Silent failures
Data breaks between implementations, not only during them.
A website, CMP, tag, campaign or integration change can disrupt measurement without producing a visible error.
The CMP changes tag behaviour
Consent Mode, advertising and analytics behave differently after a banner update.
A dashboard stops refreshing
The source, permissions or schema changes while recipients continue to see an old result.
A release changes the dataLayer
A field disappears, changes format or is sent at a different moment.
A campaign uses incorrect tagging
UTMs, ad identifiers and traffic sources diverge across platforms.
A person finds the error first
The anomaly appears only during a manual comparison of reports, invoices or backend orders.
Primary outcome
A control system for critical data.
Scope, thresholds and frequency follow the impact each signal has on revenue, campaigns, reporting and compliance.
Scopecritical KPIs, events, sources and reports
Rhythmcontrols matched to risk and data availability
Responsealert, triage, owner, repair and retest
- 01
Control definition
We define the expected range, tolerance, seasonality and comparison source.
- 02
Signal and alert
A rule identifies a variance, missing data, duplication or refresh issue.
- 03
Triage and ownership
We confirm the incident, assess impact and route it to the appropriate owner.
- 04
Repair and retest
After the change, the control is rerun and the cause and resolution are documented.
Example control areas
Events and conversions
Availability, volume, duplicates, drops and relationships between funnel stages.
Parameters and identifiers
Required-field completeness, value formats, transaction_id and source mapping.
Consent and tag firing
Consent state, tag behaviour and the impact of CMP and GTM configuration changes.
Freshness and completeness
Delays, missing periods, schema changes and breaks in source delivery.
Result reconciliation
Agreed comparisons between GA4, backend, CRM, BigQuery or sales systems.
Dashboards and access
Refresh, filters, permissions, KPI consistency and Power BI or Looker Studio views.
Repairs may be delivered by your team or as part of agreed support. Responsibility is defined before monitoring begins.
Identify critical dataA tangible process
What a useful alert looks like.
This example shows the communication structure. Thresholds, sources and ownership are agreed for each environment.
High priorityPurchase below the agreed range
- Signal
- Event count fell below a threshold that accounts for normal variation.
- Impact
- Risk of incorrect revenue reporting and campaign optimisation.
- Verification
- Compare backend orders, dataLayer behaviour and recent GTM changes.
- Owner
- Analytics confirms the signal; remediation is routed to the agreed team.
- Closure
- Repeat the purchase test, validate parameters and document the cause.
- Status
- Open, investigating, assigned, fixed or under observation.
Engagement model
Monitoring matched to data criticality.
Not every metric needs a daily alert. Scope and rhythm follow the cost of failure and the way the data is used.
Recurring review
Regular checks of the agreed scope with a summary of risks and remediation priorities.
For stable environments with fewer changes.Continuous monitoring
Rules and alerts for critical signals with an agreed triage and escalation process.
For e-commerce, lead generation and intensive campaigns.Ongoing support module
Monitoring combined with analysis, change QA, recommendations and remediation support.
As part of ongoing analytics support.Fit
When monitoring makes sense — and when to start elsewhere.
Monitoring is justified when
- conversions and revenue drive active campaign decisions,
- the website, GTM, CMP or integrations change frequently,
- multiple sources feed BigQuery, Power BI or Looker Studio,
- errors are found manually or only during reporting.
A different first step is better when
- the current configuration is unknown — start with a GA4 audit,
- core events and the dataLayer are missing — implementation comes first,
- KPIs do not yet have agreed definitions and systems of record,
- no one can receive alerts and coordinate a response.
Honest limitations
Monitoring reduces risk. It does not remove it completely.
Controls are always limited by available data, agreed rules and known scenarios.
- Not every error is an anomaly.A change may fall within a normal range or require business context.
- An alert can be false.Thresholds need tuning as seasonality, campaigns and the service evolve.
- Detection is not remediation.An owner, system access and agreed delivery responsibility are still required.
- Controls require maintenance.New events, reports and integrations require monitoring-scope updates.
Measuring the effect
How the process is evaluated.
I do not promise outcomes without a baseline. We first agree what is critical and how failures are currently detected.
Control coverageshare of critical KPIs, events and sources covered by agreed rules
Detection timetime from the problem occurring to incident confirmation
Response timetime from confirmation to assignment, repair and retest
Alert qualityuseful signals, false alarms and recurring-problem counts
One accountable consultant
I connect diagnosis, QA and business context.
I lead the work personally, combining GA4, GTM, campaign, BigQuery and BI knowledge with the question of which data truly affects decisions and who should respond to a failure.
20+years of experience
150+analytics and data projects
50+clients
Figures refer to the broader analytics and data practice, not only monitoring engagements.
Before we talk
Frequently asked questions.
Does monitoring work in real time?
It depends on source availability and signal criticality. Some checks can run frequently; others make sense daily, weekly or after a specific release. The rhythm is agreed before launch.
Who receives the alerts?
Every control should have an owner and escalation path. Depending on the problem, an alert can go to analytics, marketing, IT or the dashboard owner.
Does an alert mean you will fix the problem?
Not automatically. The scope may cover detection and triage only or include remediation support. Responsibility and system access are agreed in the engagement model.
Does monitoring replace a GA4 audit?
No. An audit establishes the baseline and identifies configuration problems. Monitoring protects agreed elements afterwards. When the environment is not understood, an audit is the better start.
How is the service priced?
Pricing depends on the number of sources, critical KPIs, control frequency, alerting approach, and responsibility for triage and remediation. After qualification, I propose a starting scope and maintenance model.
First step
Identify the data whose failure costs the most.
I will follow up with questions about sources, change frequency, data use and the people accountable for responding.
Before monitoring begins