Start a digital marketing dashboard with the decisions it must support and an agreed definition beside every number. Record the channel dimension, attribution model, reporting period and joining keys before comparing campaigns. Then reconcile a small sample with its source reports. This gives the team a practical way to investigate differences before reallocating effort.
A dashboard brief often starts with a request to put everything on one screen. A better starting point is a sentence such as: “We need to decide which landing page deserves the next content improvement.” That decision needs a different view from an enquiry follow-up queue or a campaign spending review. Sketch the decision first and choose the measures afterwards.
For a Zambian business combining website enquiries, advertising and sales follow-up, use a definition register that someone can maintain. Give each measure an owner, a source report and a plain-language interpretation. Keep the date of the definition beside the date of the data. A chart can be technically refreshed while its business meaning has quietly changed.
Which channel definition belongs beside the number?
Google Analytics defines attribution as assigning credit along the path to an important action. GA4’s reporting attribution model determines how credit is assigned in Analytics reports. Changing that model applies to historical and future data. Both points come from Google’s attribution settings documentation, checked on 24 October 2026. Save the model name with any exported comparison so another reader can reconstruct its basis.
The scope of the traffic dimension matters as much as its label. User- and session-scoped traffic dimensions are unaffected by reporting attribution model changes. That includes Google’s examples of Session source and First user medium. Do not put a session acquisition table beside an event-credit table and give both the unqualified heading “channel performance”. Name the question each table answers.
GA4’s key event lookback window determines how far back a touchpoint is eligible for attribution credit. Changes to that lookback window apply going forward. These are separate controls from the reporting attribution model, with different effects over time. Record both rather than treating a single screenshot of the attribution settings as a complete history. The same primary document explains those distinctions.
For the proposed definition register, make the following entries mandatory. They are editorial choices for a useful operating document, rather than additional GA4 settings.
| Decision view | Definition to record | Reconciliation evidence | Action owner |
|---|---|---|---|
| Acquisition | Exact user or session dimension | Matching source report and dates | Marketing lead |
| Key event credit | Event dimension, model and window | Saved attribution settings | Analytics owner |
| Advertising activity | Account, campaign identifier and reporting date | Platform export at agreed grain | Campaign owner |
| Enquiry progression | Agreed business stage and counting rule | Sample records with identifiers | Sales owner |
| Combined view | Join fields and expected row relationship | Before-and-after totals | Dashboard maintainer |
Write the counting rule as an example that a colleague can apply. If one person submits two forms, does the operational view count two submissions, one contact or one qualified opportunity? Choose the measure that serves the decision, then keep the other measures clearly named. This proposed distinction is a business definition; configuring a key event alone does not supply the sales team’s qualification criteria.
How can a blend change the apparent story?
Google’s current documentation calls the reporting product Data Studio, previously Looker Studio. Its blend documentation, updated on 22 October 2026, describes mechanics that deserve a place in the dashboard review. Use the name your team recognises in the project brief, with the current product name alongside it.
Blends are embedded in the report where they are created. Copying that report also copies its blends, but this is different from maintaining one reusable blend across reports. Underlying metrics become unaggregated numeric dimensions in a blend. Consequently, inspect the resulting chart’s aggregation rather than assuming the original metric carries its entire reporting behaviour into the combined view.
Data Studio groups and aggregates each table before joining it, using the included dimensions. Google’s explanation adds that identical rows can collapse when those dimensions lack a unique record identifier. Include an appropriate identifier when the intended analysis requires individual records to survive that preparation. This should follow the actual reporting purpose: adding every available dimension can also produce a table at a finer grain than the intended comparison.
A join condition uses fields present in each table to link records. The fields need compatible data, even when their names differ. Choose identifiers deliberately; a campaign’s display name is convenient to read, but your proposed design should document how name changes and duplicate names will be handled. Test the chosen keys against the actual exports before deciding they are suitable.
A cross join returns every possible combination of rows from its two tables. That documented operator is useful to recognise when investigating unexpectedly expanded results. Separately, Data Studio allows a date range or filter on individual tables in a blend. Review those table controls as well as the visible chart controls. A beautifully aligned chart is still hard to interpret if its inputs describe different periods.
A worked reconciliation sheet for one campaign
Consider a hypothetical exercise with campaign C7. A campaign export contains two daily rows: day one has 12 clicks and day two has 8. An enquiry export contains three submissions across the same period. These figures are invented solely to demonstrate the worksheet; they are not campaign results.
First write the intended output grain: one row for campaign C7 across the two-day period. The expected click total is 20 and the expected submission count is three. Save those independent totals before opening the blend editor. Keep the two exports unchanged as the exercise inputs, and record the date field and time basis used in each.
Next draw the proposed join on paper. If the enquiry input has multiple rows for C7 and the campaign input has multiple rows for C7, a campaign-only key needs closer examination. Decide whether both sides should first be summarised to campaign and day, or whether the report should display the entire period. Work through that design before accepting the chart’s total.
Add worksheet columns for source total, prepared-table total, joined total and displayed total. Put the row count beside each value. If a discrepancy appears, locate the earliest column at which it appears. A preparation discrepancy points to selected dimensions or filters; a later discrepancy directs attention to the join or chart aggregation. This is a proposed diagnostic sequence to run, not a claim that these checks have been executed on your data.
Finally record the decision permitted by the reconciled view. The example can support a discussion of clicks and submissions under its chosen definitions. Assessing qualified demand needs the agreed qualification measure and a suitable connection to the enquiry records. Keep that next requirement visible in the brief so the dashboard grows through explicit questions.
Give the register a change log with the old definition, new definition, effective date and affected views. Review it whenever a key event, attribution setting, campaign naming convention or business stage changes. Ask the owner to demonstrate one number from source to chart during handover; this makes future discrepancies much easier to investigate.
Connect the reporting brief to your SEO work, Google Ads activity and online store where relevant. To scope a dashboard around the decisions your team actually needs to make, start with a project brief.
By Spout Team. AI-assisted drafting by Codex using the linked primary documentation.