How to run a what-if scenario simulation in DataLens
Charts describe what happened. Simulation asks what happens if something changes — on your data, not a template.
What you will do
You will pick a target metric, adjust the variables that drive it, and see the projected effect — plus forecast it forward and test how it behaves under a shock.
The tools range from a straightforward scenario to comparison, impact and shock analysis.
When this is useful
- A decision needs a defensible estimate rather than an instinct.
- You want to compare two strategies before committing to one.
- You need a forecast from your actual history rather than a spreadsheet extrapolation.
- Someone asks what happens if demand drops 20% and "we would cope" is not an adequate answer.
Before you start
- A clean dataset with the relevant metrics — The simulation is only as good as the data behind it.
- A date column, for forecasting — FutureCast needs a date column and a metric to project.
- A specific question — "What happens if unit cost rises 10%?" gives you something to configure. "What might happen?" does not.
Steps
Open Analyse → Simulate & Decide
Select Analyse in the navigation, then Simulate & Decide, and choose your dataset.
Several tools are available, from a conversational starting point through scenario, comparison, impact, forecast and shock analysis.
The available simulation tools listed.
Start with Ask Ari if you are not sure where to begin
Ask Ari takes a question in plain language and points you at the right tool.
It is the entry point when you know the question but not which screen answers it.
Set up a scenario
Open Scenario Lab. Choose a Target Metric — the thing you care about — then Select Variables and Adjust Variables to change the inputs.
Change one variable at a time first. Changing three at once and getting a surprising result tells you nothing about which one caused it.
The projected effect of your adjustment on the target metric.
Compare strategies
Strategy Comparator sets two or more scenarios side by side, which is more useful for a decision than looking at each in turn.
Impact Radar shows how a change propagates across metrics rather than only to the one you targeted.
Forecast forward
FutureCast projects a metric into the future. Set the Date Column, the Metric to Forecast and how many Periods Ahead.
Choose the method: Linear Trend for steady movement, Moving Average to smooth noise, Exponential Smoothing to weight recent periods more heavily.
A projection for the periods you specified.
Test a shock
Shock Simulator models disruption. Choose the shape: a Sudden Drop/Rise as a step change, a Gradual Change as a ramp, a One-off Spike, or Sustained Pressure.
Matching the shape to the real event matters — a supply disruption is sustained pressure, not a spike, and the two produce very different answers.
Reach a decision
Decision Copilot brings the analysis together into something you can act on.
What happens next
Simulation output supports a decision; the decision still needs writing down. The document generation screen produces a Scenario Analysis Summary and a Decision Support Brief.
Where the underlying data needs to be current for the simulation to be worth running, CDC sync keeps it so.
Example
A supply chain team asks what a 15% cost increase does to margin. Scenario Lab shows margin falling below target in two of five product lines. Strategy Comparator sets a price rise against a volume shift. The board discussion starts from two modelled options rather than two opinions.
Tips
- Change one variable at a time before combining them. Attribution is impossible otherwise.
- Match the shock shape to the real event — step, ramp, spike and sustained pressure answer different questions.
- Pick a forecast method deliberately: linear trend for steady series, moving average for noisy ones, exponential smoothing when recent periods should count for more.
- Simulations are only as good as the history behind them. Clean the data first.
Limitations
- Forecast methods are linear trend, moving average and exponential smoothing. Sophisticated econometric or machine-learning forecasting is not what these are.
- A simulation projects from your data under assumptions you set. It is a structured estimate, not a prediction.
- Forecasting needs a usable date column and enough history to establish a pattern.
- Results reflect the relationships present in the data supplied. A driver absent from the dataset cannot be modelled.
Related how-to guides
Chart and pivot a dataset
Bar, line and scatter charts, histograms, a correlation matrix and a pivot table over your data.
Generate a data document
Produce an architecture document, a model spec, a mapping or a runbook from what you have already built.
Clean a messy dataset
Drop nulls, fill gaps, remove duplicates and cap outliers — with a live preview before anything is applied.
Connect an AI model
Add a model provider with your own credentials, choose which models are allowed, and set a default.
Related questions
More of these on the DataLens FAQ page.
Can DataLens do what-if analysis?
Yes. Simulate & Decide under Analyse offers scenario modelling on a target metric, strategy comparison, impact analysis across metrics, forecasting, and shock simulation with step, ramp, spike and sustained-pressure shapes — all against your own data.
What forecasting methods does DataLens use?
Linear trend, moving average and exponential smoothing. You choose the method, the date column, the metric and how many periods ahead to project. These are transparent statistical methods rather than black-box models — appropriate for planning, not a substitute for specialist forecasting where that is warranted.
Try this in DataLens
DataLens is in private beta. Request access and work through this guide on your own data.
Request beta access