Measure the preferences. Then act on the evidence.
Every engagement I run follows the same shape: map how your people prefer to work, hold that map against the results you care about, and change what the evidence says to change. This page explains the method behind that — and the boundaries I hold it to.
Three steps, in the open
First, we map. Your people take the survey: sixty statements, answered on sliders, in around ten minutes. Each person gets their own myMAP, and each person owns their own data — consent is captured before the survey begins, and the results belong to the individual who produced them.
Second, we hold the maps against your numbers. A talentMAP built from your longest-tenured, best-performing people shows the preference profile behind the KPI. A teamMAP shows where a team clusters and where opinion splits into camps. A cultureMAP shows the whole operation at once. When those maps sit beside your attrition, quality and sales figures, patterns appear that exit interviews rarely surface.
Third, we act — and a human decides. The maps inform hiring conversations, onboarding, coaching and team design. Where matching is used, the figures inform and a person decides: the app applies no threshold, rejects nobody and shortlists nobody on its own — no algorithm screens anyone out. I advise, your managers decide, and the evidence carries the argument or the argument fails.
Four factors. Twelve preference pairs.
Sariio measures preference — how someone is inclined to work — from a person's own answers. Preference is measured against the individual's own answers; a group norm plays no part. And the survey passes no verdict on ability — a map assigns no type and grades no one.
Wary ↔ Trusting · Concern for task ↔ Concern for people · Independent ↔ Interdependent. How someone builds working relationships, and what they protect when work and people pull in different directions.
Problem solver ↔ Risk taker · Similar ↔ Different · Procedures ↔ Options. Where someone's thinking starts: the proven route, or the untested one.
Thinking ↔ Doing · Rational ↔ Intuitive · Certainty ↔ Ambiguity. How choices get made — and how much unknown a person will carry while making them.
Specific ↔ General · Finisher ↔ Starter · Perfectionist ↔ Pragmatist. Where someone's energy sits in the life of a piece of work, from first framing to final polish.
This is what a cultureMAP looks like
Below is a real cultureMAP — twenty-seven completed myMAPs from a client operation, anonymised. Six dimensions, each drawn from the preferences of the people who took the survey, with commentary the AI is only allowed to write where the numbers support the claim.
The full interactive version, with every dimension explorable, is in the app. Explore a live cultureMAP (opens in new tab)
The commentary is gated, logged and audited
The narrative on a Sariio map is written by a language model, and the model works under constraint. Where the numbers give no support for naming a leading preference, the commentary names none. The model works from what the survey recorded; where a gap exists in the data, the gap stays visible. All output is logged with timestamps and version numbers, so any sentence on any map can be traced to the model, prompt and data that produced it.
Two commitments sit above all of this. No automated decision is ever made about an individual — a human being always decides, and that commitment is written into the contract as well as the design. And Sariio takes no biometric input of any kind: no voice, no face, no video. That keeps the platform outside the EU AI Act's Article 5 prohibition on emotion recognition in the workplace — by design.
What preference data cannot do
A map describes orientation: how a person is inclined to work. A map does not measure skill, predict performance on its own, or justify screening anyone out. Where a talentMAP benchmark is used, the app shows how closely a candidate's preferences match it — a fixed count of aligned statement pairs, not an AI judgement — and that match is not a merit ranking: a high figure means similar preferences to the benchmark group, not a better candidate. Selecting only for similarity also selects for more of the people you already have — a benchmark drawn from a homogeneous group can reproduce whatever that group has in common, relevant or not: a risk under the Equality Act 2010. So the match informs, the client decides which lens the team needs — similarity, or deliberately-sought difference — and no threshold, automated screen or auto-shortlist exists to decide it for them. I will say that in the room whenever a client asks for a ranked list.
Where the data lives, and who can see it
All platform data is hosted in Frankfurt on European infrastructure, encrypted in transit and at rest, with access controlled through role-based permissions and organisation-level isolation. Consent is captured before every survey, the data belongs to the person who took it, and deletion requests are honoured on demand.
The full picture — hosting architecture, certifications, sub-processors, GDPR and POPIA positions, retention, incident response — is published in one place for your DPO and procurement team: the trust & AI governance declaration.
See the method against your own numbers
Bring one role and one KPI to the first conversation. Half an hour is enough to know whether preference mapping will tell you anything your operation cannot already see.
Or start with the evidence: explore a sample myMAP (opens in new tab)