FROM SURVEY EVIDENCE TO SCENARIO THINKING
Explore what changes
when your audience mix does.
Use a real survey snapshot to explore multichannel campaign use. Change the assumptions, see the expected result, then generate fictional respondents to illustrate the scenario.
Choose an audience mix
The campaign audiences respondents most frequently serve.
Preset order: B2B / split B2B & B2C / B2C.
Moving a slider redistributes the other two shares proportionally. Shares total 100%; displayed values are rounded.
Explore different answer rates Optional
Start with the observed answer rates in each group. Edits are hypothetical assumptions. “No” is the remainder.
Read the expected result
Share answering “Yes” to using more than one media type
Expected under your assumptions
WHAT THIS TELLS YOU
Downloads include the assumptions, source counts and calculation method.
Make the scenario tangible
Fictional examplesGenerate artificial survey rows from these assumptions. This is one reproducible random draw; its result can differ from the expected calculation above.
Your examples will appear here. No real respondent records are included.
| Synthetic record | Campaign audience | Uses multiple media types? | Evidence status |
|---|
More synthetic rows do not mean more evidence. Larger artificial cohorts reduce random variation within this model. They do not resolve uncertainty in the original sample or validate the assumptions.
Evidence, method & limits
Read before sharingA transparent worked example of scenario analysis. The data describe this sample, and do not establish a representative market benchmark.
THE OBSERVED SNAPSHOT
| Campaign audience | n | Yes | No | DK |
|---|---|---|---|---|
| Mostly / entirely B2B | 17 | 11 | 6 | 0 |
| Split B2B & B2C | 34 | 27 | 5 | 2 |
| Mostly / entirely B2C | 19 | 16 | 3 | 0 |
| Total | 70 | 54 | 14 | 2 |
DK = “I don’t know.” Counts include all three answers in the denominator. Unrounded source fractions are retained in calculations.
WHAT WAS ASKED
“Does your organization conduct campaigns that include more than one media type?”
Groups come from: “What type of campaigns do you most frequently run?” They describe reported campaign audiences.
This is the 2026 sample snapshot, not a model fitted to historical trends or a live connection to Power BI.
How the calculation and synthetic examples work
For each answer, multiply each group’s share by its answer rate and add the three results. At the observed mix, this exactly reproduces 54 / 70 = 77.142857% Yes. The expected result needs no artificial respondents.
Synthetic examples allocate whole respondents across groups using the largest-remainder method, with ties resolved B2B, split, then B2C. Answers are independently drawn within each group using the selected rates and a seeded Mulberry32 generator. Expected results for the rounded group counts are shown separately. No names, emails, or other real respondent details are used.
This browser generator is version 1.0.0-js-mulberry32. It uses a different random generator from the earlier Python / Power BI setup kit, so identical seeds do not imply identical rows across those tools. The same assumptions produce the same expected percentages.
What this example can and cannot support
Useful for: understanding how audience composition can change a reported headline, comparing explicit assumptions, and demonstrating a reusable synthetic-data workflow.
Limits: small groups (17, 34 and 19) make the source rates uncertain. Raw “Complete” status is a snapshot inclusion rule; final eligibility, independence and population representativeness are unverified. This is not a validated forecast, significance test, causal effect, or recommendation to change marketing spend.
Seven of the 54 Yes respondents selected only one named channel in a later checklist. They remain in this example; the outcome models the stated gate answer, not verified channel combinations. Observed zero “Don’t know” counts in B2B and B2C do not prove that this answer is impossible. Rates are unsmoothed defaults and can be varied above.
The expected calculation treats the selected rates as fixed. It does not quantify uncertainty about those rates. Synthetic data add no independent observations and cannot make this sample statistically significant.