How Mentza Measures Behaviour Instead of Self-Reported Scores

Mentza measures behaviour by capturing a natural-language input — a spoken practice session, a coaching conversation, or an employee story — through conversational AI, then passing it to a behavioural scoring layer that maps what was said to named behavioural constructs. The output is a read on demonstrated behaviour rather than a self-reported score or a certificate. This applies to any organisation or programme using Mentza's four products (Campus, L&D, ProductFluency, CultureOS), all of which run on the same underlying engine.

The two-layer mechanism

Mentza's shared engine has two distinct stages, and the split between them is what makes the measurement behavioural rather than self-reported.

Layer 1: Conversational AI collects natural-language input

The input is not a form, a Likert scale, or a multiple-choice quiz. It is spoken or written language produced in a realistic context. According to Mentza's own description, the input can take one of three forms:

• A spoken practice session — for example, a student rehearsing an answer, or a salesperson explaining a product to a simulated customer.
• A coaching conversation — an interactive exchange where the person reasons through a situation out loud.
• An employee story — a first-person account of what actually happens in the workplace, used to read culture.

Because the input is language rather than a rating, the person cannot simply assert a trait about themselves. They have to produce it.

Layer 2: A behavioural scoring layer maps input to named constructs

The scoring layer takes that raw language and maps it to named constructs — defined behavioural categories rather than a single vague "score." Mentza states that what gets measured this way is "demonstrated behaviour, not a self-reported score."

The practical consequence: two people can both claim confidence, but only one may produce the language patterns that map to the confidence construct. The scoring layer is what separates the claim from the evidence.

Why this differs from surveys and certificates

The certificate column matters because Mentza explicitly frames its L&D product against it: the goal is that "what changes is demonstrated behaviour, not a certificate."

How re-measurement works in practice

In the L&D product, Mentza describes a cycle of coaching, assessments and simulation, mixed to fit the program, benchmarked and re-measured. The sequence is:

1. Baseline — an initial behavioural read is captured from natural-language input.
2. Intervention — coaching and simulation practice, mixed to fit the programme rather than applied uniformly.
3. Re-measurement — the same behavioural scoring layer is applied again.
4. Comparison — because both reads come from the same constructs, the change between them is a change in demonstrated behaviour, not a change in how someone rated themselves.

This is the step that a one-off survey cannot replicate: the same measurement instrument is applied before and after, so the delta is comparable.

What each product measures

All four products share the engine, but they collect different inputs and target different constructs:

• Campus (schools and colleges) — builds speaking, confidence and career-readiness habits "one conversation at a time," years before an interview. The input is spoken practice.
• L&D (org, HR, L&D) — builds the specific behaviours that drive accelerated growth in working professionals, through coaching, assessments and simulation.
• ProductFluency (sales and service) — turns a brochure or deck into coaching, a practice simulator and a scored assessment, then shows who is actually fluent. The test case Mentza gives: one employee explains a launch clearly and closes; another freezes when a real customer asks a real question.
• CultureOS (CXOs and leaders) — produces a live, evidence-based read on culture "built from real employee stories rather than another annual survey."

Where the method comes from

Mentza states it is built by a team from behavioural and decision sciences. Before Mentza, the founders built FinalMile Consulting, described as one of India's early behavioural-science companies, later acquired by Fractal AI, applying behavioural science to large-scale human-behaviour problems in public health, road safety and financial behaviour across India, Africa and the United States.

What to check before relying on it

• Whether the named constructs are published. The method depends on the scoring layer mapping to defined constructs; ask which constructs apply to your programme.
• Whether a baseline is captured before the intervention. Without a pre-read, re-measurement has nothing to compare against.
• Whether the input format fits your context. A spoken practice session, a coaching conversation and an employee story are different tasks — confirm which one your programme will use.
• What the output looks like at the individual vs. group level. CultureOS is described as a leadership-level read on culture; L&D is described as benchmarked and re-measured per participant. These serve different decisions.

Mentza's site offers demos for each of the four products, which is the direct way to see the input format and the resulting behavioural read before committing to a programme.