Analytics & Attribution

A dashboard is live and interactive: you pick a date range and read the current picture. A report is a saved definition you can re-run, schedule, and send. Use a dashboard to investigate, a report to distribute.

• Analytics — the main screen, with tabs for Ennie AI (ask questions in plain language), Overview, Predictive, Cultural, and Reports.
• Analytics dashboard — a metrics grid, performance charts, a campaign breakdown table, platform comparison, and period-over-period comparison, with a date range picker and an export action.
• Campaign analytics — the same treatment focused on campaigns; individual campaigns also have their own analytics and performance views.
• Reports — saved custom reports you create, edit, run, and delete, with quick starting points for a weekly performance report, a monthly summary, and a campaign comparison.

A saved report can carry a schedule: a frequency, a list of recipients, and an output format. Running one generates and downloads the file, and reports can be exported for use elsewhere.

Attribution is built on customer journeys. A journey is one person, an ordered list of touchpoints, a conversion event with a value, a length in days, a first-touch channel, a last-touch channel, and the path between them. An attribution model is a rule for splitting the credit for that conversion across those touchpoints. Six are available:

• First touch — all credit to the touchpoint that introduced the customer. Good for understanding awareness.
• Last touch — all credit to the final touchpoint. Good for identifying what closes.
• Linear — credit spread equally across every touchpoint.
• Time decay — more credit the closer a touchpoint is to the conversion.
• Position based — 40 percent to the first touch, 40 percent to the last, 20 percent shared across the middle.
• Data driven — credit assigned by a model trained on your historical data rather than by a fixed rule.

Changing the model does not change what happened — it changes who gets the credit for it. That is why the Model Comparison view exists: it puts the models side by side, ranks channels under each, and flags variance. Low variance means the models agree and you can act with confidence. High variance means the answer depends on the rule you picked, which usually points at thin data on that channel rather than a real insight.

1. Open Attribution. It defaults to the last 30 days and the linear model. The four headline numbers are total conversions, attributed revenue, average touchpoints per journey, and average journey length in days, each with a change against the previous period.
2. Use the Attribution Models tab to switch models and see conversions, revenue, and share per model, plus the top channels under the model you selected.
3. Move to Customer Journey to see the flow by stage — awareness, consideration, evaluation, conversion — the common paths people take, and where they drop out.
4. Use Touchpoint Analysis for the channel-level detail: influence score per channel, assisted versus direct conversions, and the interaction matrix showing which channels work together.
5. Before acting, check Model Comparison. If the models disagree sharply on a channel, treat that channel as unproven rather than as a winner or a loser.
6. Create a report from the attribution screen when you need the result to leave the platform. Reports can select metrics, set a date range, include journey and touchpoint detail, and export as CSV, Excel, PDF, or JSON.

AI Measurement is a tab inside Attribution with five panels:

• Modeled Conversions — estimated conversions where direct measurement is incomplete.
• Ennie AI — a chat panel for asking questions about your measurement data.
• Taxonomy — suggestions for naming and structuring campaigns so they group correctly in reporting.
• Discovery — signals surfaced from the data that you did not explicitly ask for.
• Scenarios — a simulator for testing what a change would do before you make it.

Separately, the Insights screen collects automated findings for the active Brand Space. Each is typed as an opportunity, an alert, a recommendation, or an insight, carries an impact of high, medium, or low and a confidence score, and can be acted on or dismissed. Insights are generated on request as well as automatically, and dismissing one is recorded rather than just hiding it.

A brand runs paid social, search, and email. Last-touch reporting says email is carrying everything, so the team considers cutting paid social.

In Attribution they switch from last touch to position based. Paid social moves up sharply, because it is where most journeys start. Touchpoint Analysis confirms it: paid social has a high assisted conversion count and very few direct ones. The interaction matrix shows paid social and email appearing together in the same journeys far more often than chance.

Model Comparison shows low variance between models for email and moderate variance for search. The team keeps paid social, leaves the email budget alone, and treats search as unresolved pending more data. They save the view as a monthly report with the finance lead on the recipient list, so the same argument does not need re-litigating next quarter.

• Attribution only sees touchpoints that reach the platform. Channels with no integration connected in this Brand Space will be absent from journeys, not shown as zero.
• A journey is stitched per person as the platform knows them. Cross-device journey mapping is not part of the current system, so one person on two devices may appear as two journeys.
• Journeys are not tracked in real time. Read attribution over a window — 30 to 90 days is the practical range — rather than as a live feed.
• You cannot define your own attribution model. The six models are fixed; Custom applies to brand adherence, not to attribution.
• High variance between models is a data-quality signal, not a result. It is most common on channels with few touchpoints.
• Analytics and attribution are scoped to the active Brand Space. Cross-client comparison is an agency-level screen, with its own dashboard, client comparison, and scheduled reporting.
• Public landing-page A/B tests are tracked in Google Analytics 4 and do not appear in these screens.