Geo-lift read-out, fictional example
A lift estimate needs a recorded design, control regions and an interval before it can be read.
- Lift
- +7%
- Interval
- 2% to 12%
- Test window
- Fictional: 4 weeks in 2 matched regions
Stage 05 of 07 · Price & Promote / Marketing mix modeling
See what changes when your budget does.
Bring channel spend and business outcomes into one budget conversation. Explore assumptions, compare allocations and agree what to test next.
Confirm model availability, data access and review criteria together before starting an MMM pilot.
Discuss an MMM pilotExplore the seven-stage workflowIllustrative scenario
Compare a fixed €100,000 budget
Illustrative budget · EUR
Current mix
€100,000
- Paid search
- 45%€45,000
- Paid social
- 35%€35,000
- Other channels
- 20%€20,000
Illustrative budget · EUR
Scenario for review
€100,000
- Paid search
- 35%€35,000
-€10,000 against baseline
- Paid social
- 40%€40,000
+€5,000 against baseline
- Other channels
- 25%€25,000
+€5,000 against baseline
Compare synthetic budget shares at the same total spend. These are planning examples, not model recommendations; changing the view does not change any advertising budget.
Illustration — not a forecast. No revenue effect has been calculated.
Review required inputs and uncertaintyChannel contribution, fictional example
| Channel | Base sales | Incremental sales | Return per euro |
|---|---|---|---|
| Paid search | €42,000 | €6,100 | €1.80 |
| Paid social | €28,000 | €4,400 | €1.60 |
| Other channels | €19,000 | €1,200 | €0.90 |
Channel return over time
Each channel's estimated return moves with season and competition. These figures are fictional and show the layout only.
| Channel | Jul | Aug | Sep | Oct |
|---|---|---|---|---|
| Paid search | €2.10 | €2.30 | €2.00 | €2.20 |
| Paid social | €1.60 | €1.50 | €1.70 | €1.80 |
| Other channels | €1.10 | €1.20 | €1.00 | €0.90 |
Lift calibration: Channel returns are calibrated against the geo-lift read-out above. Until its interval is supplied, read the estimates as provisional.
Budget
Based on the channel estimates above. A suggestion for review; no spend changes.
| Channel | Current mix | Test a different mix |
|---|---|---|
| Paid search | 45% | 40% |
| Paid social | 35% | 40% |
| Other channels | 20% | 20% |
Check the basis before you approve.
Separate underlying demand from marketing effects. Promotions, prices, seasonality and stock availability can all change the outcome.
Hold the total budget constant. Review channel minimums, maximums and operational constraints before interpreting a scenario.
Marketing effects may carry over and diminish with higher spend. Check data quality, model fit, uncertainty and sensitivity; correlation alone does not establish incrementality.
A person reviews the evidence and signs off on any plan. Use suitable experiments and subsequent outcomes to challenge the model; execution remains a separate decision.
Data for a useful budget comparison.
- Consistent historical spend by channel and period
- Aligned revenue or other agreed business outcomes
- Prices, promotions, stock constraints and seasonality
- Geography and experiment evidence where available
History length and granularity depend on the business and model. A public catalog scan alone cannot supply private marketing spend or sales outcomes. Share only agreed, authorized data after pilot scope and access are confirmed.
Review how the estimate was made.
MMM estimates relationships between aggregate marketing activity and outcomes over time. Carry-over and saturation assumptions help compare scenarios. Model-based contribution differs from click attribution; incremental effects require credible identification and validation. Results depend on the supplied data and assumptions.
Keep estimates separate from measured results. Review model uncertainty and validation evidence before using a scenario to plan an experiment.
Human review pending
Check the comparison before interpreting the change
Read the recorded assignment, outcome definition and observation windows. These fictional bundles prepare a human review; they do not estimate campaign impact.
Choose an evidence bundle
Change only the example being inspected. The budget scenario above remains separate and unchanged.
Example controls are preparing
EX-COMP-01 · v1
Read the recorded design
A protocol describes intended assignment. It does not prove who actually received an offer or whether the protocol was followed.
- Protocol reference
- EX-PROTOCOL-01-v1
- Fictional field source: EX-PROTOCOL-01-v1#protocol
- Recorded assignment method
- Protocol states random assignment; adherence unverified
- Fictional field source: EX-PROTOCOL-01-v1#assignment
- Recorded eligibility
- Example opted-in repeat buyers; application unverified
- Fictional field source: EX-PROTOCOL-01-v1#eligibility
Compare definitions and windows
Keep both group windows and the outcome definition visible. Matching dates alone do not establish comparable groups.
- Outcome definition
- Any recorded purchase per assigned example participant during the stated window
- Fictional field source: EX-PROTOCOL-01-v1#outcome
- Outcome unit
- Participant with recorded purchase
- Fictional field source: EX-PROTOCOL-01-v1#unit
- Treatment window start
- 2026-09-01T00:00:00Z
- Fictional field source: EX-PROTOCOL-01-v1#treatmentStart
- Treatment window end (exclusive)
- 2026-09-08T00:00:00Z
- Fictional field source: EX-PROTOCOL-01-v1#treatmentEnd
- Comparison window start
- 2026-09-01T00:00:00Z
- Fictional field source: EX-PROTOCOL-01-v1#comparisonStart
- Comparison window end (exclusive)
- 2026-09-08T00:00:00Z
- Fictional field source: EX-PROTOCOL-01-v1#comparisonEnd
- Time zone
- UTC
- Fictional field source: EX-PROTOCOL-01-v1#timezone
Keep missing evidence visible
Unknown exposure, denominators and missing outcomes remain unknown. Do not replace them with zero or infer a campaign effect.
- Assigned group denominators
- Unknown — not supplied
- Fictional field source: No source supplied
- Actual exposure evidence
- Unknown — not supplied
- Fictional field source: No source supplied
- Outcome completeness
- Unknown — not supplied
- Fictional field source: No source supplied
- Review owner
- Unknown — not supplied
- Fictional field source: No source supplied
- Approval record
- Unknown — not supplied
- Fictional field source: No source supplied
Prepare the next review
The selected bundle highlights an evidence question, not an approval. Ask the responsible analyst to check assignment, delivery and outcome completeness before interpreting a difference.
Design recorded; review still open
The protocol records assignment and aligned windows. Delivery, adherence and complete outcomes remain unverified.
Request assignment adherence, exposure records and complete outcomes.
Example controls are preparing
Return to planning inputs
No experiment is launched, no audience is selected and no budget changes here. Use this review alongside the original planning inputs.
Review planning inputsAgree data, scope and review criteria first.
The public-catalog pilot is a starting conversation. MMM requires a separately agreed scope and suitable historical data; it is not included automatically.
Discuss an MMM pilot