From the design to the result, without a project plan.
Example
Design
Picks regions and test length
Launch
Holds out 12 regions in Meta
Run
Live, day 9 of 21
Running
Read
Measures the real lift
Learn
Saves it in the brain, proposes the next budget move
Day 9 of 21
Customer experience from Cassandra’s marketing mix modeling work.
Know what really caused each sale
Incrementality testing uses geographical A/B tests on campaign spend to estimate incremental lift. Cassandra doesn’t stop at the design: it can launch the test, keep it running and read it for you. Agent execution depends on supported channels and account permissions, including Meta access.
A suitable design can support a causal estimate, subject to assumptions, statistical power and uncertainty.
Existing Campaigns
Run a geo-holdout test to understand its true incrementality
Launch a new Channel
Launch new channels discriminating on geos to assess their true incrementality before spending nationally
Detect diminishing returns
Understand how much you can invest in one channel before reaching diminishing returns
How it works
YOU
1. Your question
state your measurement question
Cassandra
2. Region selection
split of geos into test and control
3. Set how long the test runs
Check the test can spot a real lift
YOU + CASSANDRA
4. Implementation
Agents execute within your permissions and approval rules.
Cassandra
5. Read the result
Measure lift and inform the next action
Use the test evidence to prepare the next budget move. Cassandra implements supported changes within your workspace permissions, then checks the observed result and saves what the evidence supports.
Available actions depend on your connected workspace and permissions.
See how the result is checked →
Example execution
Illustrative
Budget plan prepared
Prepared
Required approval received
Approved
Supported budget change applied
Implemented
Confirm the scope and permissions. Request approval where your workspace rules require it.
Compare the observed result with the saved prediction. Keep supported findings and open questions for the next decision.
Causal evidence. A clear plan. Permissioned action. A result to learn from.