Analytics & CRMYour questions, answered.

A key event is an interaction important to the business, such as a completed enquiry or purchase. It is not necessarily the same measure as a Google Ads conversion. Define the action precisely and validate collection before interpreting reports or using it for advertising decisions.

Next step: Document the meaning and trigger of each key event.

Choose outcomes showing meaningful commercial progress rather than every interaction. Clicking a button and successfully submitting a form are different actions. Poor definitions inflate results and can misdirect optimisation even when the tracking software operates correctly.

Next step: Test completed outcomes and distinguish primary goals from supporting indicators.

A CRM records qualification, opportunities and sales beyond what a website session shows. Linking outcomes to marketing can improve decisions, but matching depends on available identifiers, implementation and privacy requirements. A connection alone does not create complete attribution.

Next step: Agree sales stages and map how outcomes return to reporting.

Eligible offline events can be sent using supported methods such as Measurement Protocol, depending on identifiers, timing and configuration. Not every sale can be matched to a visit, and historical data may not update as expected. Review implementation and privacy requirements first.

Next step: Pilot one outcome with technical and privacy review.

Data Import joins supported external information to analytics data according to its processing rules. Measurement Protocol sends events, including eligible server-side activity. They solve different problems and should not be described as interchangeable ways to reconstruct every customer journey.

Next step: Decide whether you need an attribute join or an event before choosing the method.

A form submission is not automatically a qualified prospect. Qualification needs an agreed business rule and a reliable CRM stage. Changing definitions silently makes monthly comparisons misleading, even if the lead count appears accurate.

Next step: Agree qualification criteria with sales and retain definitions alongside reports.

Platforms can use different attribution settings, reporting dates, eligibility and measurement inputs. A discrepancy does not automatically prove faulty tracking. Compare the same event and period and understand configuration before forcing reports to match.

Next step: Audit one conversion's definitions and settings across both platforms.

Attribution assigns credit to recorded touchpoints associated with an outcome. Models change how that credit is distributed but do not prove causation or reveal all influences. Explain the model and missing data instead of presenting attributed revenue as exact incremental value.

Next step: Record the model, lookback window and limitations in reports.

Use them as one consistent view of recorded marketing touchpoints, alongside CRM outcomes and customer feedback. They are not complete proof of which channel caused a sale. Long sales cycles and offline conversations can be missing from the model.

Next step: Compare attributed activity with real opportunities and closed sales.

Cross-domain measurement can preserve a coherent analytics journey across configured domains, such as a website and booking service. It is not universal identity matching and does not guarantee joining cross-device behaviour. Incorrect setup can create unwanted referrals.

Next step: Inventory journey domains and test a real transition between them.

Consent Mode communicates consent states to Google tags so their behaviour can adjust. It is not a banner and does not obtain permission by itself. The data flows and legal suitability depend on the implementation, which needs appropriate review.

Next step: Test both accepted and declined choices through your actual tag setup.

No. A separate interface must collect and communicate relevant choices where required. Consent Mode manages tag behaviour based on signals; it does not itself settle compliance obligations. Avoid choosing a mode solely because it promises more measurement.

Next step: Confirm who owns the consent interface and who configures the tags.

The approaches differ in how tags operate before consent interaction and what data flows occur. In basic implementations tags can be blocked until the relevant interaction; advanced setups have different behaviour. Neither label establishes legal compliance or exact data recovery.

Next step: Ask for a documented data-flow review against current guidance.

Repeated submissions, duplicate tags or overlapping conversion setups can inflate counts. Analytics events do not automatically deduplicate CRM records. Define what counts as one outcome and trace the event sources rather than simply dividing a report total.

Next step: Follow one test lead through tags and CRM and remove duplicate triggers.

Google recommends generate_lead for a submitted form or request for information. Fire it for the relevant completed action, not merely opening the form. It records an interaction, not later qualification or sale, which needs separate business processes.

Next step: Verify the event only fires after successful receipt of the enquiry.

Modelled values estimate outcomes that cannot be directly observed under certain conditions. They are not individually recorded events and should be labelled appropriately. Modelling does not eliminate the need for consent controls, accurate setup or transparent reporting limitations.

Next step: Explain where estimates may be included in stakeholder reports.

External customer value can support analysis where a documented import method and suitable data basis exist. Minimise personal data and check processing and historical behaviour. Not every business needs detailed customer attributes in analytics to make useful decisions.

Next step: Identify the specific decision before importing CRM value data.

A later purchase needs a reliable outcome record and, where lawful and supported, identifiers that connect it to earlier activity. Timing and platform limitations affect matching. Do not assume a form event can automatically identify every eventual sale.

Next step: Map the lifecycle and test a sample lead-to-purchase connection.

Check permissions, account links, auto-tagging and the intended conversion definitions. Importing an event already tracked another way can create duplicate goals. Linking accounts does not make all metrics or attribution identical across platforms.

Next step: Review existing conversions before creating additional imports.

Specify business questions, events, definitions, data sources, owners and known limits. Include consent and quality checks. A plan prevents collecting everything without a reason and makes it easier to diagnose changes when a site or sales process evolves.

Next step: Write a measurement plan before adding new tracking.

Users describe identified visitors under the tool's rules, while sessions group activity into visits. Identification and session definitions vary and are affected by consent and devices. Neither number should be assumed to represent an exact count of unique people.

Next step: Document the definitions used in your reporting tool.

A customer may use several devices or be unmeasured, while one account may represent several people. Identification limits and consent also affect totals. Use sales records for customer counts instead of relabelling website users as buyers.

Next step: Compare confirmed customers with analytics activity without assuming one-to-one matching.

GA4 engagement rate uses sessions meeting its engaged-session criteria. It is a platform-defined indicator, not proof that someone valued the content or became a lead. Interpret it alongside task completion, page context and your configured events.

Next step: Review the configured engagement and outcome definitions before interpreting changes.

A landing-page report groups visits by their initial page under the reporting rules. It helps examine entry journeys, not the full influence of each page. Consent, redirects and configuration can affect what is recorded as the starting page.

Next step: Compare important entry pages with their qualified outcomes.

Use consistent channel rules and campaign labels that distinguish known evidence from inference. Tools may categorise the same visit differently. Do not invent a source where tracking is missing or treat direct traffic as proof of brand loyalty.

Next step: Document your channel definitions and audit incorrectly labelled campaign links.

Direct commonly includes visits without a recognised referral source, not just typed URLs. Missing campaign labels, apps and privacy restrictions can contribute. It should be reported as limited source information rather than an exact account of visitor intent.

Next step: Review tracking gaps before drawing conclusions from direct traffic growth.

Agree consistent sources, mediums and campaign names and keep a shared register. Case differences and inconsistent labels can fragment reports. Never place personal or sensitive information in tracking URLs that may appear in logs or shared links.

Next step: Create a short naming guide and check links before publication.

Test important events and reconcile relevant outcomes with authoritative records. Inspect duplicates, missing values and consent behaviour. A dashboard loading successfully does not establish that collection, definitions or attribution are correct.

Next step: Run a test enquiry and verify exactly one meaningful recorded outcome.

Use suitable exclusion or identification methods and test them carefully. Remote work and changing network addresses can limit simple IP-based approaches. Keep diagnostic traffic distinguishable without accidentally excluding customers or relying on a method that no longer applies.

Next step: Check exclusion behaviour from office and remote test journeys.

Tools filter some automated activity but detection is imperfect and definitions differ. Suspicious visits are not automatically customers or evidence of AI visibility. Avoid drawing sales conclusions from traffic patterns without corroborating human engagement or enquiry evidence.

Next step: Investigate unusual spikes using available technical and outcome data.

A data dictionary explains each metric's source, definition, period and limitations. It prevents teams using the same label for different calculations. Keep it current when goals, connectors or reporting rules change.

Next step: Write definitions for leads, qualified leads, revenue and conversion rate.

Use comparable periods and account for seasonality, working days and data delays. A partial month should not be presented as equivalent to a full month without qualification. Label both periods so readers understand exactly what changed.

Next step: Compare like-for-like elapsed periods or clearly disclose the difference.

Cohort analysis follows groups sharing a starting event or period, such as first purchase month. It helps separate customer ageing from overall growth. Define the group and observation window clearly and avoid comparing new cohorts with mature ones unfairly.

Next step: Compare customer cohorts with the same time since acquisition.

A funnel report shows recorded progression through defined stages. Missing events or different journey paths can distort it. Use it to locate possible friction, then verify the behaviour before assuming every apparent drop represents abandonment.

Next step: Test the recorded steps against a real completed journey.

Scroll events indicate recorded depth or movement, not that content was read or understood. Page length and device dimensions affect comparisons. Combine the data with meaningful actions and usability observations rather than treating deep scrolling as automatic engagement quality.

Next step: Compare scroll depth with page purpose and relevant outcomes.

A heatmap visualises collected interactions such as clicks or scrolling across a page and device context. It can reveal patterns but not motivations. Layout changes, viewport sizes and incomplete collection must be considered before interpreting clustered activity.

Next step: Review a consistent page version and device group alongside usability evidence.

Collect only appropriate interaction information and avoid capturing sensitive field contents or identifying details unnecessarily. Review consent, notices and retention for the actual implementation. A colourful visualisation does not remove the responsibilities associated with its underlying data.

Next step: Check collection payloads and ensure private form information is excluded.

Use the corresponding layout and viewport context so interaction positions match what visitors saw. Combining unlike devices can produce misleading overlays. Historical interactions also need caution after a page redesign changes element positions.

Next step: Compare each device's data against its actual page layout.

Call tracking associates eligible phone interactions with campaigns or journeys using supported methods. It has technical and attribution limits and may involve personal data or recordings. Clarify notices, routing and business ownership of numbers before implementation.

Next step: Test routing and qualification reporting before advertising a tracking number.

Use clear entry and exit criteria based on observable progress, such as qualification or proposal acceptance. Stages should match the actual sales process. Vague labels let different staff classify the same opportunity differently and weaken forecasts.

Next step: Define criteria and owners for each pipeline transition.

Use careful matching and merge rules that preserve relevant history and preferences. Shared addresses or names do not always indicate the same person. Keep a review route for uncertain matches and avoid overwriting opt-outs during imports.

Next step: Test merge behaviour with representative contacts and permission records.

Lead scoring ranks contacts using agreed signals of fit or intent. It is an aid to prioritisation, not proof someone will buy. Validate criteria against outcomes and avoid unfair or inappropriate use of personal information.

Next step: Compare high-scoring leads with actual qualification and closed sales.

Define the start event and what counts as a meaningful response, distinguishing automated receipt from human handling. Account for stated business hours and service expectations. Fast acknowledgement alone may not resolve the customer's need.

Next step: Report enquiry receipt to meaningful first response with consistent rules.

Set rules for routing, handover and reassignment so each enquiry has a responsible person. Include absence and escalation handling. An automated notification is not proof that somebody accepted responsibility or responded.

Next step: Test routing with the usual owner unavailable.

A forecast estimates likely future outcomes using stage, value, timing and assumptions. It is not booked revenue. Review historical close rates and stale opportunities so optimistic stage labels do not inflate expected sales.

Next step: Reconcile forecast assumptions with mature opportunity outcomes.

Record reasons using customer evidence where possible and distinguish them from internal guesses. Compare by offer, stage and source. Losses may reflect fit, process or timing rather than automatically demonstrating a failed marketing channel.

Next step: Review a sample of recent losses with the sales team.

Define the authoritative source for each field, identifiers and supported update directions. Handle retries and conflicts explicitly. Synchronising everything indiscriminately can duplicate records or spread incorrect permissions and outdated preferences.

Next step: Test creation, correction and opt-out changes across both systems.

It means an agreed authoritative record for a particular fact, not necessarily one system containing everything. Revenue may come from finance while qualification comes from CRM. Document ownership and reconciliation rather than blending incompatible sources silently.

Next step: Assign an authoritative source for each key business metric.

Export only what is needed to an authorised destination and protect access and retention. CSV files can escape normal system permissions once downloaded. Review recipient rights and data minimisation instead of treating export capability as unrestricted permission.

Next step: Approve the purpose, fields and secure handling before export.

Set justified periods for each purpose and record deletion or anonymisation processes, considering legal obligations. Keeping all records indefinitely is not a default strategy. Ensure backups and connected tools are considered rather than deleting only visible dashboard rows.

Next step: Create a retention schedule with responsible owners and review dates.

Verify and assess the request under applicable rules, then coordinate relevant systems and exceptions. Do not promise instant deletion of every record where legal obligations or other grounds apply. Keep an appropriate response process and seek qualified advice for uncertainty.

Next step: Map where relevant information is held and follow the documented request procedure.

A snapshot stores results as captured at a particular time for later reporting. It can make dashboards faster and reduce repeated external requests. Show freshness and source limitations because a cached result is not a live view of every connected service.

Next step: Label the last successful refresh and flag failed updates.

Providers process results on different schedules and may revise recent data. Sync failures and quotas add further delays. A dashboard should disclose freshness instead of showing an old snapshot as though it reflects the current moment.

Next step: Check source dates and the last successful refresh before comparing results.

Distinguish unavailable, not connected and genuinely zero results. A zero can imply measurement succeeded when it did not. Provide a clear reason and next step where known rather than inventing values to fill an attractive chart.

Next step: Audit empty cards and label their actual data state.

Give each audience only the information appropriate to its role and enforce access where data is retrieved. Private CRM details and internal analysis may require additional restrictions. Hiding a tab alone does not protect the underlying information.

Next step: Test client access to private agency records and other clients' reports.

Record the change and date, then compare relevant outcomes while noting other influences. A later improvement does not automatically establish causation. Use technical verification for the fix itself and commercial evidence for any claimed business effect.

Next step: Annotate the change and assess comparable before-and-after data cautiously.

Define the outcome, denominator and test conditions before making changes. Consider lead quality and commercial value, not just submission volume. Small samples, seasonal shifts and tracking changes can make apparent improvements uncertain.

Next step: Choose one test with a meaningful outcome and explicit review criteria.

Incrementality concerns outcomes caused by an activity beyond what would otherwise have happened. Attribution assigns credit under rules and is not the same thing. Controlled experiments can help estimate incrementality, but practical designs still have limitations.

Next step: Separate attributed results from evidence of additional outcomes.

State missing data, modelling, lag and assumptions beside the relevant conclusion. Use clear language without burying every qualification in a footnote. Uncertainty does not make reporting useless; it helps teams choose decisions proportionate to the evidence.

Next step: Add source and limitation notes to your most consequential metrics.

Review outcomes, completed work, unresolved issues and the next priorities using connected data and sales context. Separate evidence from recommendations and assign owners. A monthly review should produce decisions, not only a longer collection of charts.

Next step: Agree three next actions with owners and a review date.