Scenario Planner
Adjust input sliders, downstream metrics recompute based on statistical correlations learned from your historical data. "What happens to CAC if we increase paid spend by $20K?"
Why use it: Make decisions with projected outcomes, not gut feel. "If we add $20K to paid spend, CAC goes from $42 to $51."
Where to find it: Suite → Intelligence → Scenarios (
/scenarios)Who can use it: Any authenticated team member can run and save scenarios. Each metric relationship needs at least 3 common weekly data points (learned from up to 52 weeks of history).
Scenario Planner
"What happens to CAC if we increase paid spend by $20K?" Adjust input sliders, and downstream metrics recompute based on statistical relationships Celiq learns directly from your historical data. Relationships are computed deterministically using Pearson correlation between metric pairs, with impact projected via an elasticity term; Claude Haiku is used only to assign the overall confidence score and write the plain-English narrative summary.
Using scenarios
Suite → Intelligence → Scenarios (/scenarios)
How it works
- Celiq computes Pearson correlations between your metrics from
kpi_snapshots. Only metric pairs with a correlation strength of |r| > 0.4 are kept, and the impact of one metric on another is projected via an elasticity term (r × sdB/sdA). - You enter a metric name, a baseline, and a target value (e.g. paid spend +$20K). A slider appears once a positive baseline is entered.
- Downstream impacts are projected, and Claude Haiku assigns an overall confidence score and a narrative summary.
- Save scenarios to compare options side by side.
Confidence scores
After the impacts are computed deterministically, Claude Haiku assigns an overall confidence score (0–1) from a prompt summarising the input changes and the computed impacts (including each pair's r-value and weeks of data). It is a model judgement, not a deterministic formula. If the AI call fails, confidence defaults to 0.5.
The model weighs signals such as how strong the underlying correlations are and how many weeks of data back them. The UI buckets the result as High (≥75%), Medium (≥50%), or Low (below 50%).
Relationships are learned only where a metric pair shares at least 3 common weekly data points (drawn from up to 52 weeks of history). When no correlated metrics are found, Scenarios shows "No correlated metrics found. Add more historical KPI data to unlock scenario analysis."
Saving and comparing scenarios
Each scenario you run can be saved with a name. The Scenarios page shows all saved scenarios in a comparison grid — projected inputs, projected outputs, and confidence for each.
Use this to present options to stakeholders: "Here are three paths to our Q3 CAC target."
Orion integration
Ask Orion "what happens if we increase X by Y?" and it will automatically run a scenario and show projected impacts inline in the chat.
You can then save that scenario from the Orion response without leaving the conversation.