How Culture Code asks and calculates
Instead of asking in general whether people agree, Culture Code asks about specific experiences: in the last three months, how often could employees raise a mistake without fearing negative consequences? The fixed questions use frequency steps. Whether the questions are clear and measure reliably, we check with real participants in the pilot.
- 6
- dimensions
- ≈ 40
- questions, closed and open
- ≤ 3
- follow-ups per interview
- 5
- people per group, at least
Where the method comes from
In work and organisational psychology the approach is called the critical incident technique, or the behavioural interview: you collect the incidents people describe and look at what repeats. Until now that took interviewers and a lot of analysis time. So most organisations send out questionnaires with rating scales instead.
Culture Code runs the interview by machine, so that every employee can take part. The system writes the follow-up questions itself, for example who decided or what happened next, at most 3 per interview. It never asks for names.
Two sources, collected separately and then compared
What the company says about itself
Mission statement, values page, brand book, leadership principles, code of conduct. We record each statement on its own with its source and where it was found. The company approves each one, and only approved statements count as the standard. If the company has no such documents, the management answers seven questions about what should hold. Those answers become the documented standard.
What employees report
A guided conversation with about 40 questions, closed and open, and at most 3 follow-ups. The survey is anonymous and runs in the browser on the employee’s own phone, without signing in. We estimate about 19 minutes.
How the questions are asked
Four closed questions with their answer options, one open question, and an example of the automatic follow-up on the phone. Whoever approves the survey receives all of them word for word beforehand.
The interview itself is in German. The questions here are translated so you can read them.
Openness & trust · How often it happenedWhen you make a mistake, how often can you raise it without fearing negative consequences?
- NeverContradicting
- RarelyContradicting
- SometimesMixed
- OftenAligned
- Almost alwaysAligned
- Does not happen to meNot scored
- I cannot judgeNot scored
Involvement & decisions · How often it happenedHow often can you make as many decisions about your own work as you think sensible?
- NeverContradicting
- RarelyContradicting
- SometimesMixed
- OftenAligned
- Almost alwaysAligned
- I cannot judgeNot scored
Openness & trust · How often it happenedHow often do you keep concerns to yourself because you fear negative consequences?
- NeverAligned
- RarelyAligned
- SometimesMixed
- OftenContradicting
- Almost alwaysContradicting
- I cannot judgeNot scored
Direction & goals · How often it happenedWhen priorities change, how often are you told why?
- NeverContradicting
- RarelyContradicting
- SometimesMixed
- OftenAligned
- Almost alwaysAligned
- Does not happen to meNot scored
- I cannot judgeNot scored
When someone picks an answer at the edge of the scale on a core question, the system asks once about the specific incident. Tap an answer:
Follow-up generated from the last answer
“Never”: What specific case is behind that?
Write here … (free text, names are never asked for)
Imagine someone new starts working with you. What should they know about how things really work here?
A few sentences are enough. Please leave out names.
Follow-up questions on open answers. After an open answer the system asks one follow-up question, for example who decided or what happened next. It never asks for names. There are at most 3 follow-ups per interview. Whoever approves the survey in the company receives their framework word for word beforehand.
On their own phone. The survey runs in the browser, without a login, and can be paused at any time. People who work shifts need neither an appointment nor a computer.
The 6 dimensions
The structure follows well-known models from organisational research. The questions are our own wording, not questions from licensed instruments. Each dimension has three fixed questions, asked the same way in every company, so that the values can be compared.
The fixed questions of each dimension use the same scale: never, rarely, sometimes, often, almost always, plus “I cannot judge”, which counts towards no share. They ask about the last three months. For some questions “often” is the unfavourable answer, for example when concerns are held back; the question catalogue sets this.
One of the fixed questions has no scale: it asks people to name up to three of eleven reasons why they work here. It is not scored and goes into no finding. Another question is open and voluntary: what would have to happen for you to leave? The answers are grouped into themes and counted, not shown word for word.
In each dimension one question is about the person’s direct manager. Anyone without a direct manager, such as the management, does not get these questions. At the start the interview asks whether someone manages staff. That way you can compare how managers and staff experience the same everyday work, without evaluating any group under 5 people.
- Direction & goalsbased on mission: goals, priorities, pace
- Involvement & decisionsbased on involvement: who decides, who is asked
- Collaborationbased on coordination: working across areas
- Change & customerbased on adaptability: customers, learning, change
- Openness & trustbased on psychological safety: mistakes, disagreement, concerns
- Fairness & recognitionbased on organisational justice: recognition, promotion
How a finding is made
Every statement of the company that can be asked about as behaviour becomes a question. A finding comes from two questions: is the standard documented, and how is it lived day to day? That gives seven types, each with an action:
- Lived aspiration (preserve): described in company documents and experienced day to day.
- Emerging aspiration (strengthen): described in company documents and partly experienced.
- Unwritten strength (make visible): not described in company documents, experienced as helpful.
- Aspiration rarely experienced (review aspiration): described in company documents and rarely experienced.
- One-sided practice (rebalance): one behaviour dominates and pushes other aspirations back.
- Contradictory practice (examine causes): not described in company documents, runs counter to an aspiration.
- Differences between groups (examine differences): not described in company documents, experienced differently by group.
The report orders the findings into strengths, gaps and risks, and within each by size and strength of evidence. Each finding shows how it is spread across the groups and ends with a question for the conversation. The questions on openness and trust are not a finding of their own: they are the ground, and say how far the other answers can be relied on. Fixed thresholds decide the type, not a language model.
The system checks every sentence written by machine. Are its figures and quotes in the evidence? Does it read each answer the way it was counted? And does it judge something the finding does not support? Some things the questions leave open, for example whether one role should decide more than another or whether customer contact is part of it. Then the finding carries a reading note.
What the methoddoes not allow
These rules are fixed in the software and cannot be switched off in a project.
A finding with too little evidence is not printed. Nor is it worded more cautiously. The report lists it among the withheld findings and gives the reason. That way nobody reads the gap as an all-clear.
When the figures change, the text goes out of date. For every written finding the system stores which state of the figures it rests on. If they change, the text counts as out of date and is no longer used.
A group below the floor is not evaluated. A group of fewer than 5 people is merged with a larger one, and the method section of the report says which. This already happens in the database query. A group that is too small is not hidden when it is displayed; it never appears in the data.
A dimension the mission statement says nothing about gets no invented claim. If the company’s documents say nothing about a dimension, the report says so. It can only become a finding of its own once the collection of documents has been explicitly closed.
Free text is deleted once the report is complete. Quotes appear in the report only if each one was anonymised and approved beforehand.
Limits of the method
What Culture Code does not do. Worth knowing before you commission it.
The core questions have not yet been validated empirically. The questions are an expert draft. Cognitive interviews, pilot data and a check of how well they measure across roles and areas are still to come. So the evaluation is not a scientifically validated culture score.
The interview runs in German today. There is no English interview yet, and the questions on this page are translated for reading only. Anyone who does not speak German well cannot answer in their own language. In a company with many other mother tongues that skews the result, and it is worth weighing before the survey.
The findings section measures. Recommendations are a separate, marked section. The findings section shows where the standard and everyday work diverge, and by how much. Survey data does not show why, so it derives no measures. Recommendations can be booked as an add-on and appear in a section of their own at the end. The system drafts them from a finding and checks them against fixed rules, then a person releases them. Each recommendation names who that was. What follows from it is the company’s decision.
A comparison with other companies only from ten. The core questions are the same in every company. Once a report is complete, the core values per dimension go into a data pool, without names and without any link to the company. A comparison appears in the report only once at least ten companies with the same core questions have contributed, and then as a quiet line beside the company’s own value.
What is collected is what people report. People asked how often something happened answer less often the way they are expected to than people asked whether they agree with a statement. It is still an account of their own experience: of the 32 fixed questions, 14 ask for their own assessment and 17 ask how often something happened in the last three months. The report shows how many people answer alike, not what happened in a single case.
Whether the mission statement is lived says nothing about whether it is good. A company can keep to its mission statement exactly and still work badly. A team can have good reason to depart from an outdated one. The method shows whether the standard is lived and how differently groups experience everyday work. Whether a pattern makes the work easier or harder is something we discuss in the joint review.
A survey shows one point in time. It shows the state at the time of the survey. Whether anything has changed only a second survey with the same questions can show.
The sample report
A complete report for a fictional company, in German, with 9 findings, the figures beneath each one and the method section. You can check the method against it before commissioning anything.
How the pilot runs and who approves itSample approval document (German)
What happens in your organisation on Monday morning?
Let’s find out together. A short conversation is enough to see whether Culture Code fits your situation and where a first pilot would help most.
