Understand how media contributes to business performance

The Attribution page shows how Prescient’s models distribute performance across your media mix for any target type.

→ One pager

Use it to:

• Compare modeled results with channel-reported results
• Understand direct and cross-channel impact
• Review performance by channel, tactic, or campaign
• Analyze how attribution changes over time
• Find room for scaling based on historical data points turned into saturation plots.

---

How this fits together

%%{init: {"theme":"base","flowchart":{"curve":"basis","nodeSpacing":50,"rankSpacing":68,"padding":12},"themeVariables":{"fontSize":"14px","primaryColor":"#ffffff","primaryTextColor":"#111111","primaryBorderColor":"#d4d4d4","lineColor":"#8a8a8a","textColor":"#111111","edgeLabelBackground":"#ffffff","clusterBkg":"#fafafa","clusterBorder":"#ebebeb"}}}%%
flowchart LR
  DC("<b>Data Center</b><br/>connectors · channels · tactics") --> MC("<b>Model Center</b><br/>are the models trustworthy?")
  MC --> ATT("<b>Attribution</b><br/>what your media did")
  MC --> MF("<b>Media Forecaster</b><br/>what to do next")
  AI("<b>Prescient AI</b><br/>ask across all of it") -.-> ATT
  AI -.-> MF
  MF --> EX("<b>You execute</b><br/>in your ad platforms")
  EX --> DC
  classDef here fill:#111111,stroke:#111111,stroke-width:1.5px,color:#ffffff
  classDef accent fill:#f1fcf5,stroke:#1fd25a,stroke-width:1.5px,color:#111111
  classDef choice fill:#fafafa,stroke:#c9c9c9,stroke-width:1.5px,color:#5a5a5a
  class ATT here
  linkStyle default stroke:#8a8a8a,stroke-width:2px

Where the next dollar should go

%%{init: {"theme":"base","flowchart":{"curve":"basis","nodeSpacing":50,"rankSpacing":68,"padding":12},"themeVariables":{"fontSize":"14px","primaryColor":"#ffffff","primaryTextColor":"#111111","primaryBorderColor":"#d4d4d4","lineColor":"#8a8a8a","textColor":"#111111","edgeLabelBackground":"#ffffff","clusterBkg":"#fafafa","clusterBorder":"#ebebeb"}}}%%
flowchart TD
  S("<b>Your spend</b>") --> M("<b>Your MMM models</b>")
  M --> B("<b>Base</b><br/>what you would have sold anyway")
  M --> H("<b>Halo</b><br/>what media caused elsewhere")
  B --> P("<b>Performance Table</b>")
  H --> P
  P --> R("<b>ROAS</b><br/>what the spend so far returned")
  P --> MR("<b>Marginal ROAS</b><br/>what the NEXT dollar returns")
  R --> J("<b>Judge what happened</b>")
  MR --> D("<b>Decide where to grow</b>")
  classDef here fill:#111111,stroke:#111111,stroke-width:1.5px,color:#ffffff
  classDef accent fill:#f1fcf5,stroke:#1fd25a,stroke-width:1.5px,color:#111111
  classDef choice fill:#fafafa,stroke:#c9c9c9,stroke-width:1.5px,color:#5a5a5a
  class MR accent
  class D accent
  linkStyle default stroke:#8a8a8a,stroke-width:2px

Find your way around

Set up the view
Select your view · Date granularity · Comparison periods

Read the numbers
Performance summary · Attribution over time · Performance table · Base and halo flows

Reference
Modeled vs reported · Best practices · FAQs

---

---

1. Select Your View

Use the filters at the top of the page to choose:

• Target: Revenue, New Customers, or any enabled Modeled target.
• Model: One, many, or all applicable models
• Channels: All channels or a selected group
• Date range: The period you want to analyze
• Views: Save your filters and results for quick access

*These filters apply across the page.

---

→ Date Granularity

Attribution Over Time can be rolled up to month or quarter, not only viewed day
by day or week by week.

The finest grain available follows your models. If every selected model serves
daily data, you can select any date range. If any selected model serves weekly
data, date selection is limited to whole weeks — Prescient names the models
responsible in the hint beside the date picker, so you can deselect them to return
to daily.

From that base grain, the charts' own granularity control rolls buckets up:

Rolling up changes how the data is bucketed, not which data is included. The
totals are the same; the resolution is not.

→ Comparison Periods

Turn on comparison and Prescient shows the same metrics for an earlier period
alongside the current one, so every figure carries the change beside it.

Rather than setting the earlier dates by hand each time, choose a preset. Each one
shows the dates it resolves to, so there is no ambiguity about what you are
comparing against:

• Previous period — the window immediately before the one you are viewing
• A month earlier
• A quarter earlier
• A year earlier
• Custom period — pick the dates yourself

The comparison shows the earlier period's own figures, not only the difference,
and the comparison columns are included when you export, so the export matches what
is on screen.

Weekly organizations compare whole weeks. If your models serve weekly data, a
year-earlier comparison shifts by whole weeks rather than to the same calendar
date, so both windows start on the same weekday and stay comparable.

2. Performance Summary

The Performance Summary provides a high-level overview of the selected date range. The metrics displayed vary depending on the selected Target (e.g., Revenue or New Customers, other targets).

→ Revenue Target

Spend

Displays the total media spend for the selected period.

It includes:

• Daily Average: Average spend per day.
• Channels: Number of channels included.

MMM Revenue

MMM Revenue is Prescient’s estimate of the incremental revenue associated with media.

It includes:

• Base Revenue: More direct media contribution
• Halo Revenue: Revenue influenced across another channel, retailer, or model
• Channel Revenue: Revenue attributed directly to marketing channels.

MMM Revenue %

The percentage of total revenue attributed by the Marketing Mix Model (MMM).

It includes:

• Base %: Percentage from direct media contribution.
• Halo %: Percentage from indirect media effects.
• Channel Revenue %: Percentage attributed directly to marketing channels.

MMM ROAS

MMM ROAS compares modeled incremental revenue with media spend.

It may differ from channel-reported ROAS because advertising platforms and Prescient use different attribution methods.

It includes:

• Base ROAS: ROAS from direct media contribution.
• Halo ROAS: ROAS from indirect media effects.
• Channel ROAS: ROAS based on channel-attributed revenue.

→ New Customers Target

Spend

Displays the total media spend for the selected period.

It includes:

• Daily Average: Average spend per day.
• Channels: Number of channels included.

Reported New Customers

New customers reported directly by advertising platforms.

• Daily Average: Average new customers per day.
• Channels: Number of reporting channels.

MMM New Customers

Prescient's estimate of the incremental new customers generated by media.

It includes:

• Base: New customers directly attributed to media.
• Halo: New customers influenced through indirect media effects.

MMM New Customers %

The percentage of total new customers attributed by the Marketing Mix Model (MMM).

It includes:

• Base %: Percentage from direct media contribution.
• Halo %: Percentage from indirect media effects.

MMM CAC

MMM Customer Acquisition Cost (CAC) measures the average media cost to acquire a modeled new customer.

It includes:

• Base CAC: Cost per directly attributed new customer.
• Halo CAC: Cost per indirectly influenced new customer.
  

---

3. Attribution Over Time

The Attribution Over Time chart shows how selected metrics change over the selected date range, helping you identify trends and understand how media performance evolves over time.

→ Compare View

It lets you compare two metrics across the selected period to help identify trends and understand how marketing activity relates to business outcomes.

Examples include:

• Spend versus MMM Revenue
• Spend versus reported revenue
• MMM Revenue versus channel-reported revenue
• MMM ROAS versus channel-reported ROAS

Use this view to understand whether changes in spend were followed by changes in modeled performance.

→ Breakdown View

Break down a single metric to understand which components are driving performance over time.

Depending on the selected metric, you can view results by Channel or Tactic.

Additional display options include:

• Absolute: Displays the actual metric values.
• % of Total: Displays each component as a percentage of the total metric.

The summary cards above the chart update to show the total value and the contribution of each selected component for the chosen date range.

---

4. Performance Table

Use the Performance Table to compare results at different levels of your marketing hierarchy.

Choose a tab between Channels, Tactics and Campaigns to change the level of analysis of the table.

The table's header stays pinned below the filter bar as you scroll, so the column names stay in view however long the list is. It also scrolls sideways with the table.

Channel

Compare performance across advertising platforms and media sources.

Tactic

Review groups such as awareness, prospecting, retargeting, or branded search.

Campaign

Analyze individual campaigns at the most granular available level.

For each item, you can compare:

• Spend
• MMM Revenue
• MMM ROAS
• Channel-reported revenue
• Channel-reported ROAS

Marginal ROAS and Marginal CAC

Marginal ROAS is the return on the next dollar, not the average return on every
dollar so far.

ROAS tells you what the spend you already made returned. Marginal ROAS tells you
what adding to it would return, read off the channel's saturation curve at its
current spend level. The two answer different questions, and they routinely
disagree:

• A channel with strong ROAS and weak Marginal ROAS has been working well and is near saturation. Adding budget returns less than its history suggests.
• A channel with modest ROAS and strong Marginal ROAS has room. It is the one to grow.

Marginal CAC is the same idea for acquisition targets: the cost of the next
customer, rather than the average cost of the customers you already acquired.

Use ROAS to judge what happened. Use Marginal ROAS to decide where the next dollar
goes.

Explore Performance Trends

Expand a row to explore the selected channel, tactic, or campaign. Switch between three views:

• Saturation: the item's Saturation Plot, showing how its returns change as spend changes. An expanded row opens on this view by default.
• Compare: two metrics over time. You can change the reporting interval (Day, Week, or Month).
• Attribution: how the selected item contributes over time.

If a row has no saturation curve for the current target and model selection, it opens on Compare instead.

Saturation Plot

The Saturation Plot helps you understand how changes in spend have historically influenced incremental revenue for a selected campaign.

It uses historical model data to estimate how performance may change at different spend levels, making it easier to identify opportunities to scale investment or recognize when returns begin to diminish.

How to read the chart

The chart includes several elements:

Historical Model – The green curve represents the modeled relationship between daily spend and incremental revenue.
Historical Data Points – Each gray dot represents an observed historical spend level and the modeled incremental revenue generated at that spend.
Confidence Band – The shaded green area shows the model's confidence range. Wider bands indicate greater uncertainty in the estimate.
Current Spend – The dashed vertical line marks your current average daily spend.
Scenario Spend – Adjust the slider below the chart to test a different daily spend scenario.

Explore Spend Scenarios

Use the Explore Spend Scenario slider to simulate different spending levels.

As you move the slider, Prescient updates the estimated results based on the historical model.

The summary below the chart includes:

Current Incremental ROAS – The modeled ROAS at your current average daily spend.
Current Average Daily Spend – Your historical average daily investment.
Scenario Incremental ROAS – The estimated ROAS at the selected scenario spend.
Estimated Incremental Revenue – The expected increase or decrease in incremental revenue compared with your current spend.
Model Confidence – Indicates how reliable the estimate is based on the amount and quality of historical data available.

Seasonal Forecast

Seasonal Forecast incorporates expected seasonal patterns into the spend curve. It is on by default. Turn it off to see the historical model on its own.

When it is on, the projected relationship between spend and incremental revenue reflects both the historical model and seasonal effects, helping you evaluate scenarios during periods of unusually high or low demand.

If no seasonal forecast is available for the selected item and models, the switch is off and disabled. Hover its info icon to see why.

ROAS axis

The ROAS switch adds a second axis showing ROAS at every spend level, so you can read efficiency straight off the curve. It is on by default. For targets measured by cost, such as customer acquisition, the switch is named after that metric (for example CAC) instead.

Best Practices
Compare scenarios close to your historical spending range for the most reliable estimates.
Consider both Incremental ROAS and Estimated Incremental Revenue when evaluating budget changes.
Use the confidence band to understand how certain the model is at different spend levels.
Treat scenarios far outside your historical spend range as directional guidance rather than precise forecasts.

Tip: The Saturation Plot is based on historical modeled performance. It is designed to help evaluate budget scenarios, not to predict exact future results.

---

5. Base and Halo Flows

The Base & Halo Flows visualization shows how media channels contribute to revenue across your business outcomes.

• Base represents the direct contribution of a media channel to an outcome.
• Halo represents the indirect contribution, where one channel influences performance in another channel, retailer, or model.

Available Views

Use the controls to explore the data in different ways:

• All flows
• Base contribution only
• Halo contribution only
• Results by model
• Halo breakdowns
• Combined contribution

→ How to read the diagram

1. Flows start from the contributing media channel on the left and connect to the outcome on the right.
2. Thicker flows represent larger contributions.
3. Selecting a node or flow updates the details panel with a breakdown of that contribution.
4. Use the view controls to compare direct (Base) and indirect (Halo) impact.

Select a channel or model to explore its relationships in more detail (right panel).

This view is especially useful for identifying media that creates demand but does not receive credit from traditional last-click attribution.

---

Compare Modeled and Reported Results

Differences between Prescient and advertising-platform reporting are expected.

Channel reporting answers:

Which conversions did this platform claim?

Prescient attribution answers:

What incremental contribution was associated with this media investment across the business?

A channel may appear stronger or weaker in Prescient because the model accounts for baseline demand, delayed impact, and cross-channel Halo Effects.

---

Frequently Asked Questions

My ROAS is strong but Marginal ROAS is low. Which do I act on?

Marginal ROAS, if the decision is about adding budget. A low marginal figure means the channel is near saturation — the next dollar returns less than the average dollar did.

Why can't I pick an arbitrary date range?

Because one or more of the selected models serves weekly data, so ranges snap to whole weeks. The hint beside the date picker names those models; deselecting them returns the page to daily.

Do the totals change when I switch to monthly?

No. Granularity changes how data is grouped for display, not which data is included.

Best Practices

• Compare similar date ranges when reviewing performance changes.
• Evaluate both Base and Halo contribution.
• Review channel-reported and modeled metrics together.
• Use tactic and campaign views to investigate channel-level results.
• Confirm model health before acting on unexpected findings.
• Use Media Forecaster to test a future budget change.