Starting point
A leaver survey is only worth as much as people's candour. People answer candidly when they are sure HR cannot trace an answer back to them. In small departments with only a few leavers a month, a promise in the cover letter is not enough: a single additional answer that shifts the figures gives away who wrote it.
The consultancy wanted to offer its method as a product of its own that shows each client company under that company's own logo, with the questionnaires that suit it, in German and in English, and with an evaluation HR can read for itself. And when HR already has a list of the people leaving, nobody should have to create those accounts one by one.
Solution
Each client company is an area of its own on the platform, with its own accounts, its own logo and the questionnaires the consultancy assigns to it. A questionnaire can be assigned to several companies without being copied. The company's logo replaces the platform's in the navigation, so leavers see the name of the employer they are leaving rather than that of a software product. The consultancy maintains companies, logos, questionnaires and videos in an administration interface of its own.
Anonymity is a rule on the server, not a display setting. The server only releases a questionnaire's evaluation once at least five people have completed it, and after that only in full steps of five: with 37 participants, 35 count, and the other two wait for the next three. A single departure therefore never changes the figures on its own. Free-text answers are stored but kept out of the aggregated evaluation, and each HR department only sees the leavers of its own company.
Every question and every answer option is held in one record in German and in English. Because both language versions share the same identifiers, German and English answers end up in the same evaluation. Leavers answer a questionnaire one question at a time, with a progress bar and an estimated duration, by choosing an option or in free text. It is submitted once and in full, and the server checks that every question has exactly one matching answer. Anyone who stops halfway leaves no partial answers behind.
Next to the questionnaires sits a video library for each language, with thumbnail, duration and a note of what has already been watched. Leavers can rate each video from one to five stars and comment on it; the ratings go to the consultancy.
HR invites leavers one at a time or from a CSV or XLSX file, with sample files in both formats. The file is checked row by row before the import: missing or invalid e-mail addresses and missing roles are reported with the row number, duplicate addresses within the file are rejected, and invitations only go out once the whole file is correct. For people who already have an account, HR chooses whether it stays as it is or is updated from the file. The server repeats every check.
- One area per client company: its own accounts, its own questionnaires, its own logo
- Results only from five participants and only in steps of five, enforced on the server
- Free-text answers stay out of the aggregated evaluation
- Questions and answers in German and English in one record, evaluated together
- One question at a time, submitted once and in full, checked on the server
- A video library per language with watched status, star rating and comment
- Invitations by CSV or XLSX, checked row by row, with a choice for existing accounts
A look inside the product
Illustrative sample data; an anonymised product example that deliberately does not show the real product.
Result
The platform was built in 2022 in about three and a half months: the first screens in mid-February, the connection to the server in early March, the anonymity rule, invitations and file import by the end of March, the last adjustments at the end of May.
It was designed so that the consultancy could provide its method to each client company under that company's own logo, and HR could read the results for itself, without going through the consultancy and without any single answer becoming recognisable.
Three decisions carried this: building anonymity as a rule on the server rather than a promise in the cover letter, keeping both languages in one record rather than in two separate questionnaires, and checking the import so strictly that a faulty list never turns into invitations.
- Built in 2022 in about three and a half months
- Anonymity as a rule on the server, not a promise in the cover letter
- One evaluation for both languages
- Invitations from a checked list instead of person by person