2.1Landing page

After signing in, you land on the All Projects page. From here you can search existing projects or start a new one. Projects are listed as tile cards, paginated at 10 per page by default.

2.2Creating a project

Click Create New and give the project a name.

2.3Project tile cards

Once created, a project appears as a tile showing the brand, year, and creator. Click any tile to re-open that project and continue where you left off. With many projects on file, use the search bar above the grid to jump straight to the one you need.

4.1Trend charts

Pick one primary variable (typically the KPI) and up to 50 secondary measures to plot against it, over a chosen time window. Click Get Trend to render the chart.

• Add variables on the fly — the left panel lets you extend the comparison after the first render, without starting over.
• Save a trend — click Save and give it a name; a correlation table appears alongside the chart for every secondary measure.
• All saved trends — click the Trend chevron at any time to return to the list of everything you've saved.
• Bulk actions — from that list, filter, download, or delete trends individually or via the checkboxes for a bulk download.

4.2Time comparison

Compares a set of variables across up to four periods at once.

1. Select up to 50 variables under Classify Measures.
2. Choose up to four periods — Monthly, Quarterly, Yearly, or a Custom range you define yourself.
3. Click Compare.

The chart sits above a data grid showing the support for each period plus the percentage change between consecutive periods, ending in an overall change column (e.g. 2022 vs. 2020). With more than four variables or periods, a scrollbar lets you pan across the rest. Save the comparison the same way you'd save a trend, and revisit every saved comparison from the Time Comparison chevron.

5.1Create model & setup

Click Create New Model from the (initially empty) Modelling landing page. In Model Setup, confirm the dependent variable and the model duration — by default, the full data history.

The Advanced link inside Model Setup exposes three further choices:

Click Confirm to move to Variable Selection.

5.2Variable selection

The left panel lists every classified variable; tick individual variables, whole groups, or use search. Turn off the Hierarchy toggle to browse a flat list instead of the grouped view.

Each selected variable can be set to one of four states:

Mutually exclusive variables

When two forms of the same signal shouldn't sit in one model together — impressions and spend for the same channel, say — mark them mutually exclusive so the system tests them in separate iterations instead of risking multicollinearity.

1. Answer Yes to "Are there mutually exclusive measures in the list below?"
2. Click Add Custom Group and name the group.
3. Classify the competing variables into #1 / #2 (use the + control if there are more than two forms).

5.3Report format

Defines the period structures available later on the model's "due to" and output pages, plus which one loads by default. The system pre-populates recent yearly, monthly, and quarterly periods; you can also define your own.

Custom periods — pick Custom granularity, set a length in weeks/months/days and a start date, then Add. The platform slices the full data range into consecutive periods of that length automatically (e.g. twenty 20-week periods across a two-year cube).

Once periods exist, choose Period 1 and Period 2 on the right, click Add, and mark one pairing as the default with the radio button.

5.4Transformations, priors & qualifying criteria

The Advanced panel (top right of Modelling) turns on three more configuration tabs, plus the choice between an auto and a manual run:

Variable transformations

Pick a transformation type per variable — adstock, gamma, log, lag, and so on — and either an exact parameter value or a range to iterate across. Turning on Show Saturation layers a saturation curve on top of the transformation for the variables that need one.

Priors

Two ways to encode prior belief about a variable's effect:

• Contribution % range — set a minimum and maximum share of the dependent variable this variable is allowed to explain.
• Coefficient — fix an exact coefficient and standard deviation.

Auto-run models default to a 0–80% contribution range per variable, which keeps incremental variables from coming out with a negative coefficient.

Qualifying criteria

Thresholds an iteration must clear to count as a good model:

Only iterations meeting every enabled criterion land in the Qualified tab of the results.

5.5Run & model results

Click Run to see a summary of every edit about to be applied — dependent variable, duration, mandatory/optional variable counts, transformations, qualifying criteria, and the number of model outputs that will be produced with an estimated processing time. Click Run again to execute.

Each full run is a batch; a batch can contain one output or many, depending on how many optional variables, transformation ranges, or mutually-exclusive combinations you allowed. Batches in progress show live status on the Modelling landing page, with a View Details log and the option to jump into partial results before the batch finishes.

Reading Model Results

Open Model Results to see every batch as a tile card summarizing its inputs and outputs. Drilling into a batch sorts its iterations into four tabs:

Select up to three iterations and click the view icon to open a dashboard — single-model or side-by-side. Each model view includes Model Fit, Decomposition of KPI, Spend vs. Contribution with ROI, Contribution, Response Curves, Due-to charts, ROI, and Effectiveness. Click any chart header to drill into its expanded view; the Levels dropdown moves between Level 0 (Base/Incremental), Level 1 (your defined groups), and Level −1 (individual variables). Export Model downloads every output to Excel; Export Data gives you both the transformed and raw data behind it.

5.6Update an existing model

To update the model in the platform, re-run it: any change to the data cube, transformations, saturation settings, priors, or variable properties takes effect in Model Output only after the model runs again. To revise an existing iteration, open Edit Config, adjust the relevant setting, and run again — this is also the standard fix when a run gets stuck at a low % or a batch errors out. Remember to publish or save any iteration you want to keep, since iterations are automatically cleared after 30 days.

6.1Publishing to reports

Before a report can exist, the iteration behind it must be saved. From Modelling → Model Results → Saved Iterations, toggle View in Report on the iteration you want published — a report is generated automatically the first time you do this.

From the Reports list you can rename a report, delete it, pop it out into its own window, or check its box to send it on into Simulation.

6.2Reporting outputs

Reports package the published model for exploration. What you lean on next depends on your role.

Reports include everything from the Modelling dashboard, plus:

• Marketing Spend & Marketing Revenue — headline totals for the selected period, with the percentage change against the initial values.
• Budget Allocator and Optimized Revenue charts — computed by default across a 50%–150% constraint range around current spend.

For Supply Chain, use Reports to understand volume contribution and drivers — not media budget optimization.

• Contribution — Base vs Incremental share of volume.
• Decomposition — how volume moves over time.
• Due To — what drove volume change between two periods.

Skip Marketing Spend / Revenue scorecards and Budget Allocator / Optimized charts in this path. Continue into Planning for week-level forecasts and non-media assumptions.

7.1Sessions & scope

A published report opens with a Default Scenario covering the most recent 52 weeks — run it as-is for a quick first read, or click Create New Session to define your own.

1. Name the session (no special characters) and add an optional description.
2. Pick the Report to simulate from.
3. Set the Duration — any timeframe within the report's data.
4. If the data spans multiple dimensions, set a Dimension Filter: the aggregation level to simulate at, and optionally a subset of dimension values to focus on.

Click Next to save the session and move into Data Validation.

7.2Data validation

Before any scenario runs, review the base data the simulation will use — weeks of execution, spend, support, unit cost, unit type, and base period for every variable. Variables with zero executions in the session window are flagged so you can assign them a different base period or drop them.

The Advanced panel unlocks four further controls:

7.3Select groups, response curves & scenarios

Select Groups sets the aggregation level the response curve renders at. Response Curves then shows Average and Marginal Return curves per variable — click into a curve to see its flighting pattern, and use the dropdown to switch between variables. Financial Inputs lets you confirm or edit the coverage factor, gross margin, and revenue multiplier for the session period before reviewing the summary and clicking Create Session.

Every session tracks: session name, created-on date, model used, base period, and dimension aggregation. Inside a session, click View Scenarios to see all scenarios created so far (a new session opens with a default "Base Scenario"), and New Scenario to build another.

7.4Simulate & optimize

Simulate tab — set an expected value or drag the slider for any measure or group, and watch Spend, Customers, Revenue, Profit, and ROI update live in the panel on the right. Values can be entered as a percentage or an absolute number, capped at 300%.

Optimize tab — for when you know the goal but not the ideal mix:

1. Choose the objective: maximize revenue or maximize profit.
2. Optionally set upper/lower limits per variable (default 0–300%) and a total budget constraint (default locked at 100–100%).
3. Use Apply to All to push the same constraints to every driver except total budget.
4. Click Run Scenario.

The system reallocates spend within your constraints and returns the plan with the strongest outcome.

7.5Reading the output

Optimization output pairs the original base plan (grey) against the optimized plan (blue) across Spend, Customers, Revenue, Profit, and ROI. Use the Filter dropdown to view results at the Incremental, Total, or Media level. Supporting charts include:

• Budget Allocator — current vs. base spend by channel.
• Revenue and ROI — current vs. optimized, with the uplift called out.
• Response Curves, Spend vs. Contribution with ROI, and Strategic Quadrants (Optimize / Sustain / Shift / Expand) for channel-level prioritization.
• Marketing Spend and Marketing Contribution donuts, default vs. forecast.

View Details expands into a Grid View — the same metrics as a table, with Default, Forecast, and Difference columns for Spend, Contribution, Revenue, ROI, and Profit side by side, and column filters to focus on what matters for the analysis at hand.

8.1Choosing a planning approach

Every plan starts with one question:

When you select Build a Plan Manually, a second question appears: which type of forecasting to use.

8.2AI-Generated Plan

After selecting AI-Generated Plan, three further inputs appear before you click Forecast:

1. Select duration — choose the forecast horizon from the dropdown. Options include 12 months, 6 months, next quarter, or a custom start–end date within the available data range.
2. Target KPI (optional) — toggle Yes and enter a sales volume target. The AI optimizer will build a plan that tries to meet it.
3. Marketing Budget (optional) — toggle Yes and enter a total spend cap. The AI will stay within this constraint while maximizing the objective.

Click Forecast. The system reads historical trends for every variable, builds min–max assumption ranges automatically, and finds the plan that best meets your stated objective within your budget. You land directly on the Forecast Plan results screen.

To inspect or tune the AI’s assumptions after seeing the output, click Edit Inputs at the bottom of the results screen. This opens the Input Assumptions table, where you can override any variable before re-running the forecast.

8.3Manual Plan — Plan Details

After selecting Build a Plan Manually → Causal Forecasting, you reach the Plan Details screen, which has three steps:

1. Choose Report — select the published report to source the model from. The most recently published report is pre-selected. Use the dropdown to choose an earlier one if needed.
2. Select duration to create Forecasted Plan — set how far forward the system should generate values for all key variables. The current plan shown covers 12 months (05 Apr 2025 – 28 Mar 2026).
3. Select Assumption Period — use the date picker to define a historical reference window. The system pulls actuals from this period to seed default values for every variable. Non-media variables (price, distribution, competitive spend, events, macroeconomic indicators, etc.) especially need a starting assumption — without a reference period they would have no default. You can override any default in the next step.

Click Next to move to Input Assumptions.

8.4Input Assumptions

The Input Assumptions screen has two tabs: Model Variables and Additive Variables. The search bar and Hierarchy toggle at the top right help you navigate large variable lists.

Each row exposes the following columns:

Change By options

Updating the CPP

If the cost of execution has changed since the assumption period — a new rate card, different media market conditions, updated agency fees — click the pencil icon next to Forecasted CPP and enter the new rate. The system uses this to recompute spend from support (or vice versa) in the final plan.

Week-level editing — Edit Input Details

Click Edit Input Details for any variable to open a time-series panel for that variable. This is especially useful when execution doesn’t follow a flat percentage change — for example, a burst media schedule, a promotional event confined to specific weeks, or a price change that takes effect mid-year.

The panel shows:

• Time Period — each week in the forecast horizon (YYYY-MM-DD format).
• Historical Support Values — the actual value from the assumption period for that week.
• Change By % — editable per week. A week where Historical Support is 0 will show “Infinity” when you try to apply a % change — in that case, enter an absolute value directly in the Forecast Support Values column.
• Forecast Support Values — auto-computed from historical + change, or directly overrideable.

The chart at the top renders the historical series (blue) and manual forecast (orange) side by side so you can visually validate the shape of the plan. Use Save Changes to commit edits to this variable, or Reset to revert.

Export/Import is also available within this panel, for offline editing of a single variable’s week-by-week schedule.

Export / Import

Use Export at the bottom of the Input Assumptions table to download the full plan as a spreadsheet template. Edit it offline — useful for bulk changes across many variables, or when inputs are prepared externally by a media agency, finance team, or trade marketing team — then Import to upload. The system validates the file and applies all values at once.

When all inputs are set, click Forecast to run the plan.

8.5Forecast Plan output

After the plan runs, you land on the Forecast Plan screen. The Data and Chart buttons at the top toggle between the two views. The KPI summary cards are always visible.

KPI summary cards

Five headline metrics sit across the top, each showing Forecast value, the historical comparison figure, absolute change, and percentage change (color-coded):

• Sales — total volume
• Revenue
• Spend — total marketing investment
• Incremental ROI
• Profit

The Forecasted Duration and Historical Duration date ranges are shown top-right so it’s always clear what periods are being compared.

Data view — the contribution table

The table breaks down every driver’s contribution to the plan:

The Variables column is fully expandable. Click the arrow next to any group (Base, Baseline, Price, Media, Non-Media, etc.) to drill down to sub-groups, then to individual line items, then to year → quarter → month → week. This is particularly useful for validating seasonality and sharing a week-level view with other teams before the plan is confirmed.

Contributors dropdown — filter the table to show only the Top 5 or Bottom 5 contributors, or set back to All.

Periodicity dropdown — switch between Weekly, Monthly, and Quarterly aggregations on the fly.

Filters — click the funnel icon to show or hide specific columns.

Hierarchy toggle — collapses variable-level rows into channel-group summary rows.

Export / Import — download the results table or bring in an updated plan file.

Chart view — trend

The Trend chart plots the historical sales series (blue) and the forecasted series (orange) on the same axis. The break between the two lines marks the start of the forecast period. Use the Periodicity selector to smooth to monthly or quarterly if the weekly view is too granular. This chart is the quickest way to check that seasonal patterns carry forward sensibly and that planned bursts or promotional periods look directionally right against the historical shape.

Saving and iterating

Click Save at the bottom of the Forecast Plan screen to persist the plan. Saved plans are listed under the Planning sidebar and can be reopened, copied, or used as the basis for Actualization.

To adjust assumptions — whether you used the AI or Manual flow — click Edit Inputs. This returns you to the Input Assumptions table with all your previous entries intact. Make changes and click Forecast again; the output refreshes immediately.

8.6Actualization

Once a plan is saved and new actuals become available for part of the plan period, the Actualization workflow lets any team reconcile plan vs. reality and replan the remainder with full visibility into what caused the gap.

1. Upload new actuals data via the incremental data upload (same process as the Input module).
2. Open the saved plan and click Refresh. The system updates with the newly available actuals.
3. The Actuals vs. Plan chart appears — select the plan version from the dropdown at the top left.

Gap vs. Plan chart

The chart shows:

• Bars (blue) — the planned sales value for each period.
• Line (orange) — the actual sales that have now been reported. Percentage labels at each point show whether actuals are ahead or behind plan.
• YTD — top-right, the cumulative total for the actualized period with % vs. plan.
• Forecast — top-right, the full-year plan total with % vs. plan.

Due To analysis

The Due To panel below the gap chart decomposes why actuals differed from the plan. Each factor is shown as an absolute contribution and a percentage due-to:

Period selector — use the date picker at the top of the Due To panel to focus the analysis on a specific sub-period (e.g. a single quarter) rather than the full YTD window.

Groups — use the dropdown to change the aggregation level of the waterfall chart.

Hierarchy toggle — expand the data table below the waterfall to show individual variable-level due-tos instead of the grouped view.

Replanning the remainder

Once you understand the gap, create a new scenario for the remaining months:

• The actualized portion (weeks with confirmed actuals) is locked — it cannot be edited and serves as the fixed baseline for the replan.
• The remaining forecast period is fully editable using the same Input Assumptions interface (§8.4).
• Any team can update only their own variables while leaving everything else locked or at the existing plan values.
• Model the adjustments needed to close or offset the gap — increased media burst, promotional depth change, updated competitive assumptions — and click Forecast to see the projected impact before committing.

8.7Planning vs. Simulation — quick reference

Both modules use the same underlying causal model but serve different purposes and audiences.