How Franko works
This page describes Franko for people who run climate, engagement or feedback surveys: what it does, how a survey gets built, what happens when someone answers, what the report contains and what stays private.
What Franko is
Franko is a survey where every question can be followed by a short conversation. The person answers the question as in any questionnaire: they pick a point on a scale, pick an option, or write. Then Franko, the interviewer, asks one or more questions about that answer, one at a time, to find out what is behind it.
The conversations build on each other. When Franko reaches question 8, it knows the answers the person gave to questions 1 to 7 and the words they wrote in the earlier conversations. It can ask about an episode they already mentioned, or notice that an answer seems at odds with an earlier one, without making anyone repeat themselves.
At the end the survey owner gets a report: statistics per question, the points that came out of the conversations, themes, suggested actions and a written summary from Franko.
How a survey gets built
There are four ways to start. Each one ends in the editor, where everything can be changed before the survey is published.
With the guided setup
- Pick what the feedback is about: a person, a team or group, a process or tool, an event or programme, or another topic.
- Give it a name and, if you like, one line of description (“Runs acquisition and onboarding”, “A two-day offsite last month”).
- Pick a length: 3 to 5 questions, 6 to 10, or 11 to 15.
- In the editor, Franko opens a short chat and asks what you hope to learn. You answer in your own words. This conversation stays private: it shapes the questions, and respondents never see it.
- Franko writes the title, the questions, the answer options and the survey context.
From a template
More than 30 survey templates are available, filtered by category, company size and length.
From scratch
Open an empty editor and write the questions.
From a document, with Claude
Franko has a connector for Claude (an MCP server). Once Franko is connected to your Claude account, you can give Claude a questionnaire you already have, in Word, PDF or another format, and ask it to turn it into a Franko survey. Claude creates a draft with questions, answer types, options, chapters and context. A draft collects no answers until you publish it, so you review it in the editor first.
What a survey contains
Whatever the starting point, a Franko survey has this structure:
- Title and description: the description is one or two lines respondents read before they start.
- Subject: what or who the survey is about, with one line of description. For a person, also the pronoun to use.
- Context: a few sentences for Franko about what the survey covers and what matters. Respondents do not see it.
- Chapters, if needed: groups of questions with a title and a one-line description.
- Questions: for each one the wording, the answer type, the options and how far Franko should follow up.
Why the context matters
Franko reads the context before every conversation. It tells Franko what the survey is about and where it may go: the follow-up questions stay on that ground and on what the person just said. A context like “The team got a new manager in March and is reworking its release process” produces sharper questions than a generic one.
The context is confidential. Franko uses it to aim its questions, but does not quote or confirm it. If someone asks who is behind the survey or what Franko was told, Franko says that is not its to share and goes back to the question.
Trying it before publishing
From the editor you can take your own survey as a test, with real conversations, and share the test link with colleagues in your organisation. Test answers never reach the report.
Question types
- Open: the person writes the answer, or dictates it.
- 1 to 4 scale and 1 to 5 scale, with a label for each point.
- Single choice: one option. You can add “Other”, with a free-text field.
- Multiple choice: several options, up to a maximum you set. “Other” is available here too.
- Ranking: 2 to 10 items to put in order, first to last.
Chapters
Questions can be grouped into chapters. Before a chapter’s first question the respondent sees a card with the chapter’s title and description and a button to continue. A chapter can also sit at the start or the end of the survey with no questions, as an introduction or a closing.
Follow-up after each question
For each question you choose how far Franko follows up:
- None: the person answers and moves to the next question.
- Light: up to 2 exchanges.
- Standard: up to 5 exchanges.
- Deep: up to 10 exchanges.
These are ceilings. Near the limit Franko only continues if the person is still adding detail. At the limit it asks nothing new and leaves the person the choice to add something or move on. The person always decides when to go to the next question.
A useful rule: short scale or choice questions with follow-up on. The choice gives the comparable number, the conversation gives the reason.
How Franko interviews
Franko behaves like an outside interviewer with no stake in the answers. These are its rules:
- It does not name feelings. If someone describes something hard, Franko does not comment (“that must have been tough”) and asks the next question.
- It does not praise or thank people for their honesty. No “great point” or “valuable feedback”.
- One question at a time, in one or two short sentences.
- It does not suggest answers. No examples inside the question, and no offering two options to choose from.
- It does not sum up the answer before asking more, and does not read a ranking or a set of picks back.
- It accepts a pass the first time. “I don’t know” and “I don’t remember” count too. It does not push or rephrase.
- Short answers are fine. Three words are an answer.
- It talks like a peer, not like HR and not like a therapist.
- It stays on the current question. It does not change topic, jump ahead to later questions, or add questions the questionnaire does not have. The wording and order of the questions belong to the survey owner.
- It stays out of private life. If health, family or private matters come up, it says a few words (“understood”) and brings the conversation back to the survey’s topic.
Franko interviews in its own voice: curious and never judging.
How a person answers
- They open the link and read the welcome page, with the number of questions and a privacy note.
- They sign in with Google, LinkedIn or a link sent by email. No password is needed. Signing in prevents duplicate answers and lets them pick the survey up on another device.
- They answer one question per page, typing or speaking. After each answer, if follow-up is on, they talk with Franko.
- At the end they see a thank-you page.
A survey can be in Italian, in English or in both. Respondents read it in the language it is written in, or in their own if the survey has it.
Who can answer
By default a survey is open to anyone with the link: anyone who has the link and signs in can answer.
Enterprise organisations can also make a survey invite only: you upload a staff list as a CSV with an email column, pick people from it and send the invitation by email. Only someone who signs in with an address on the list can answer.
The other CSV columns (team, site, tenure, for example) become segments in the report. When an invited person starts the survey, Franko copies their CSV row onto their response. The report reads only that copy and never goes back to the list to find out who answered. If the CSV is replaced later, earlier answers stay in the segment they were given in.
In the invitation list the survey owner sees who was invited and whether the email went out. They do not see who answered.
The report
The report updates as new answers arrive. It contains:
Statistics per question
- Scales: the spread of answers and the average.
- Single and multiple choice: a bar for each option, as long as the share of people who picked it, with the most picked option marked. In multiple choice each bar is measured against the number of people, since one person can pick several options.
- Ranking: the items in order of average position, with the one placed highest marked.
Points from the conversations
For each question Franko reads every person’s answer and conversation and draws out the distinct points. Next to each point a meter shows how widely it is shared, without a number. A point made by one person is kept, because it can matter: its meter is empty, and when there are broader points it sits under “Show all”. Each point is marked positive, negative or mixed. Above the points is a one or two sentence summary of the question.
Franko counts only what the person said in their own words: an idea Franko offered that the person did not take up does not count. Doubts, corrections and who was responsible stay the way the person put them.
Themes, actions and Franko’s take
- Strengths, tensions (qualities that also carry a cost) and areas to watch, drawn from every question.
- Actions: what to do now, what to do later, and what to keep doing, with an effort estimate.
- Franko’s take: a written reading of the whole.
Participation statistics
How many started and finished, the average time, and how much people talked with Franko on each question. A table shows one row per person and one column per question, with the number of exchanges. The rows are anonymous: they show whether a drop in engagement comes from a question or from a few people.
Segment filters
In an invite-only survey the report shows a menu for each CSV column. Picking a segment recomputes the statistics and the points over just the people in that segment. Franko’s written summary stays the one about everyone.
Groups smaller than 5 people cannot be opened, and the report does not say how many people they hold. A value also stays closed if opening it next to the others would let someone work out a smaller group, for example by comparing the total with a segment or two overlapping segments. In a small organisation, use broad columns such as site rather than columns that single out a few people.
Read-only link
From the report you can create a public read-only link for people without an account. It shows the report as it is, with nothing that can be changed. The link can be turned off or replaced, and the old one stops working at once.
Reading the report from Claude
With the Claude connector you can read the report from Claude: statistics per question, points with how many people made each one, themes and actions. You can ask Claude, for example, to compare two questions. The connector does not say who answered. In enterprise organisations it can list the invited people’s email addresses and whether the invitation went out, as the invitation list in Franko does.
Anonymity
Answers are anonymous: the person who created the survey doesn’t know who answered.
Respondents sign in with an account so that each person answers once and can pick the survey up again. The survey owner gets the report: statistics per question, the points from the conversations with how widely each is shared, themes, actions, Franko’s summary and participation statistics with anonymous rows. Groups smaller than 5 people stay closed.
Where the data lives
According to the privacy policy, the application and its database run with a hosting provider in the European Union. The service is provided by Useful Systems OÜ, based in Estonia.
Answers and conversations are processed by AI models to write the follow-up questions, the surveys and the report. According to the privacy policy, these are the providers that process them:
- Google Gemini, reached through OpenRouter: follow-up questions, survey generation, the report. OpenRouter is based in the United States, Gemini runs across the EU and the United States.
- Deepgram and OpenAI: transcription when someone answers by voice. United States.
- Langfuse: technical logging of model requests, only when enabled, with short retention. Germany.
Transfers to the United States rely on the EU-US Data Privacy Framework or on Standard Contractual Clauses. The full list of providers and the retention periods are in the privacy policy.