SynPopProduct newsEight highlights · April 2025 to October 2026
What we shipped, one feature at a time.
Eight releases that shaped SynPop, from the first version in April 2025 to the pilot running today. Each issue tells the story of one feature. Newest first.
Synthetic employees answer better when they know where they work. The Custom People Engine gives every simulation your organisation's background, from material you approve.
Illustrative · Custom People EngineStatus: ready
LogisticsNetherlands and Belgium2,400 employeesSays “rota”, not “schedule”
Hybrid work policy 2026Policy · 14 passagesApproved
Works council letterLetter · 3 passagesNeeds review
Salary bandsSensitive, not usedExcluded
This run used 3 approved passagesContext recorded with the run
Illustrative. Fictional organisation and materials.
Add a short profile of your organisation and the materials you are happy to share, then approve what SynPop extracts. Simulations draw on the relevant parts automatically, and every result records what it used.
Population Builder
Start from a description, not a data export.
No file to upload yet? Describe the people, the place and the decision in your own words. SynPop drafts a population plan, shows where every characteristic comes from, and generates synthetic profiles only once you approve it.
Illustrative · Population BuilderPlan v1 · awaiting approval
Who do you want to simulate?
Create 1,200 warehouse and logistics staff in the Netherlands to test a new shift-pattern announcement.
Age band18-24 · 25-34 · 35-49 · 50-66Public source
TenureUnder 1 yr · 1-5 yrs · 5+ yrsReviewed assumption
ContractPermanent · AgencyModel suggestion
Source not used. Wrong unit: this dataset counts households, not people.
1,200 synthetic records · 0 real respondentsApprove plan and generate
Every characteristic in the plan shows where it comes from: a public source, a reviewed assumption or a suggestion. When you have real HR data, uploading it stays the better starting point.
Survey Echo
Extend the survey you already ran.
Most pulse surveys hear from a fraction of the workforce. Survey Echo extends the answers you collected to the people you did not reach, and keeps every real answer exactly as given.
Illustrative · Survey EchoHybrid policy pulse
“I understand why the hybrid policy is changing.”
AgreeNeutralDisagree
Observed312 real answers
58%24%18%
Simulated2,088 employees without an answer
49%27%24%
ExtendedAll 2,400 employees
50%27%23%
Real answers unchangedHeld-out check passedReal sample stays 312
Illustrative and experimental. Hatched bars are simulated answers; solid bars are real ones.
Results come in three layers you can tell apart: Observed, Simulated and Extended. Synthetic answers never count towards your real sample, and names or employee IDs are not needed.
Also in the October pilot
Qualitative Lab. Hear the reasons behind a forecast in a moderated synthetic focus group of four to six people from the run.
Tested from the inside, checked against the outside.
Does a forecast hold when nothing meaningful changes, and does it match real people it has not seen? Since September 2026, SynPop tests both and reports them separately.
Illustrative · ValidationOffice days forecast
Internal · stability
Same question, different words
Answer order shuffled
Scale reversed
Run repeated
External · real people
Held-out survey answers within range
Beats simple baselines
One segment gap flagged
Illustrative. Internal stability never raises a result's evidence level on its own.
Internal checks test whether a result holds when nothing meaningful changes. External checks compare it with real answers it has not seen. Only the second kind can tell you a forecast is right.
What-if scenarios
Change one thing, and ask the same people again.
A forecast tells you how a plan will land. A what-if tells you what would make it land better, before you commit to anything.
Illustrative · What-ifSame 1,500 people
What if we asked for three office days instead of four?
SupportUncertainResist
BeforeFour office days
39%26%35%
AfterThree office days
51%27%22%
18%moved towards support
4%moved away
78%stayed put
Illustrative. Biggest shift: Customer care.
Describe the change in a sentence. SynPop puts it to the same simulated people and shows who moved, and in which direction.
Interviews
Ask the workforce why.
Numbers say how many. Interviews say why. Since April 2026 you can question any simulated employee about the answer they gave, or a whole group at once.
Illustrative · InterviewOne person
CCustomer care · 1 yr · answered “Disagree”
You
What would make four office days work for you?
C
Customer care · 1 yr
Honestly, not much. I joined because the role was hybrid, and four days means paying for travel four times a week. Two days I could live with.
You
And three?
C
Customer care · 1 yr
Three, if the company helped with the cost. Otherwise I would start looking.
Illustrative. Synthetic and exploratory: shows possible reasons, not how common they are.
Pick one simulated employee, or a group such as everyone who disagreed, and ask a question in plain language. Interviews show possible reasons, not how common they are.
Probabilistic Response Engine
Every answer, with the doubt left in.
Real people are rarely certain. Since January 2026, every simulated employee gives a likelihood for each answer instead of picking just one.
Illustrative · One simulated employeeOperations · 12 yrs
“Four office days would work for my team.”
Strongly agree6%
Agree28%
Neutral33%
Disagree23%
Strongly disagree10%
Uncertainty keptSame result on rerun
Illustrative. Each bar is how likely this person is to give that answer.
The forecast keeps that uncertainty instead of rounding it away, so a divided workforce looks divided. Run it again and you get the same result.
The first SynPop
Rehearse the question before you ask it.
The first version of SynPop did one thing: put a question to a synthetic version of a population, and show the answer before a single real person was asked.
Illustrative · First releaseResults by group
“Would a four-day office week change how you feel about your job?”
PositiveNo changeNegative
Everyone
31%30%39%
Engineering
38%31%31%
Sales
34%30%36%
Customer care
17%27%56%
Illustrative. Group results, never individual records.
Upload a population, write a question, run it and read the results group by group. In January 2026 we rebuilt SynPop from the ground up, and everything since builds on that loop.
Get the next issue.
One email when a feature ships, with the story behind it.