Reading a result

This page explains what comes back when someone finishes a TIGER assessment, what each part means, how to display it, and what it may be used for. It is for the people who own the integration and anyone who acts on a result. Engineers building the display should read sections 5 to 7.

Related pages: what Kirnova delivers and the partner API reference, which covers how a result reaches you. Where this page and the API disagree, the API is correct. Tell us at team@kirnova.com.

Why this page matters#

The most common mistake with a Kirnova result is a screen, not a bug. Five numbers per candidate invite a table sorted by the middle number. That table invents an order the assessment did not produce, because where two people's ranges overlap, the assessment has not separated them. Everything below helps you show only orders the result supports, with overlapping readings shown as level, and leave the decision to a person.

1. What a TIGER result is#

A TIGER result is a reading on five dimensions, each an estimate with a range around it. It adds a small amount of written interpretation: the dimension that stands highest within the person's own profile, the two that stand lowest and are worth asking about, and, on TIGER Core and TIGER Student, two short readings in words about how the person describes themselves under demand and how they meet new ideas.

Show the meaning sentence next to each number instead of paraphrasing it.

The readings come only from the person's answers to a structured questionnaire about their own behaviour and preferences. Scoring never sees a name, photograph, age, school, university or CV.

What a dimension reading is not#

• Not a mark out of a hundred. There is no total or pass mark. The five dimensions are separate readings.
• Not a percentile. Kirnova does not report where a person sits relative to others who sat the assessment.
• Not a measure of ability, intelligence or performance. It reads self-described behaviour and is evidence about how someone is likely to work.
• Not a type or label. TIGER names nobody. There is no code, colour, animal or profile name, and you should not create them.
• Not comparable across people by arithmetic. See section 2.

2. Why every reading has a range#

The range is part of the answer#

Each dimension arrives as a lower point, an estimate and a higher point. Together they form a bar, described by the sentence:

"This bar shows where the score most likely sits."

The estimate is the best single guess, and the ends show how much room surrounds it. An assessment of this length can locate a person on a dimension but not to a single point. Answers vary with how someone read a sentence, what was on their mind or whether they were interrupted. The range is the honest width of what the assessment can tell you.

Never display an estimate without its range. A bare number implies more precision than the instrument has, and readers will trust it.

A wider range is a more honest answer#

When a reading is less settled, because answers were less consistent or gave less to work with, the range gets wider. It is not hidden, and the estimate is not nudged toward the middle. A wide bar is not a defective result. Your interface must make bar width visible. If every dimension is an identical element with a number in it, a wide reading and a narrow reading look the same.

When two people's ranges overlap#

Take two synthetic candidates who sat TIGER Core. On Resilience:

A sortable table puts A above B by nine points. But the ranges share the stretch from 56 to 64. The assessment has not distinguished these people on Resilience, and the nine-point gap sits inside the room the instrument already told you it needs.

Two people whose ranges overlap are not distinguishable by this instrument, however different the midpoints look. Any screen, sort, filter or shortlist that treats the midpoint gap as a difference asserts something Kirnova did not send.

No overlap does not prove a difference#

"When ranges do not overlap, the higher one is really higher" is also wrong. The published range is a display choice that tells a reader how much room to leave. It was not calibrated as a decision threshold. A rule built on non-overlap fires far more often than genuine differences occur and turns a display convention into an automated judgement about people.

Overlap means the readings are level. Absence of overlap is a reasonable prompt to ask about the difference in a conversation.

3. When Kirnova will not call it#

A score is shown on a scale from 1 to 100. The word beside it says where the score sits: Low is 1 to 20, Medium is 21 to 79 and High is 80 to 100. Medium means neither low nor high on this scale. It does not mean the person is average, typical or like most people, because Kirnova compares nobody with a reference group. The range and the confidence word say how sure the reading is.

• A Medium score with a Low confidence word is a reading Kirnova would not call either way: on this evidence it will not say how much weight the number can carry. Ask about it if it matters.
• Kirnova never sends the words "average", "typical" or "median", and you should not introduce them.

An empty section is dropped, not padded. Some parts of a result are present only when there is something to say. The under-demand reading arrives as text when it applies and as nothing when it does not, and an interests section appears only on an assessment that has one. When a field arrives as nothing, render nothing: no placeholder, dash, "no data available" box or generic paragraph that sounds like a finding.

4. The quality note#

Some sittings do not meet quality checks. For example, someone clicked through without reading, gave identical answers throughout, or missed the check questions that confirm instructions are being followed.

In that case Kirnova does not delete, hide or silently correct the result. It does three things:

1. The reading widens on every dimension until the ranges reflect how little the answers support.

2. A quality note is included as a single sentence:

   "The answers in this sitting did not meet our quality checks, so this result should not be used for a decision. A new sitting is recommended before relying on it."

   The person's own copy says "Your answers" instead of "The answers".

3. The under-demand reading is withheld entirely. The reading about how the person meets new ideas is still sent. Treat it, like the numbers, as not decision-grade while the note is present.

The numbers remain on screen, because hiding them would hide that a sitting happened, but they are not decision-grade.

Requirement: when a quality note is present, display it in full wherever scores from that sitting appear. Everyone sees the same note in the same words. There is no softened buyer version, no hidden candidate version and no setting that suppresses it. If a screen can show a score from a sitting, it must be able to show that sitting's note. A layout that puts the note on a detail page while scores appear on a list page breaks the requirement, because the list page is where people decide.

Most sittings carry no note. When there is none, nothing is sent and nothing should render.

5. A worked example#

This is real output of the scoring engine, produced from invented answers for a made-up person. It is not a real person's data.

Scenario. A synthetic candidate, Anya Kulkarni, is invited by a partner to sit TIGER Core as part of a hiring process. This is what the partner receives.

5.1 The result#

Copy{
  "engine": "tiger",
  "tier": "core",
  "contract_version": "tiger-report-v1",
  "bar_description": "This bar shows where the score most likely sits.",
  "dimensions": [
    {
      "name": "Trust",
      "value": 81,
      "bar": { "lower": 71, "estimate": 81, "higher": 90, "axisMin": 0, "axisMax": 100 },
      "presentation": "two_sided",
      "meaningLine": "How honestly and fairly they deal with people when nobody is checking."
    },
    {
      "name": "Influence",
      "value": 59,
      "bar": { "lower": 52, "estimate": 59, "higher": 66, "axisMin": 0, "axisMax": 100 },
      "presentation": "two_sided",
      "meaningLine": "How readily they step forward, speak up, and move other people."
    },
    {
      "name": "Grit",
      "value": 48,
      "bar": { "lower": 39, "estimate": 48, "higher": 58, "axisMin": 0, "axisMax": 100 },
      "presentation": "two_sided",
      "meaningLine": "Whether they keep going when work gets long, slow, or hard."
    },
    {
      "name": "Execution",
      "value": 65,
      "bar": { "lower": 55, "estimate": 65, "higher": 74, "axisMin": 0, "axisMax": 100 },
      "presentation": "two_sided",
      "meaningLine": "Whether they deliver finished, careful work — not just effort."
    },
    {
      "name": "Resilience",
      "value": 64,
      "bar": { "lower": 56, "estimate": 64, "higher": 73, "axisMin": 0, "axisMax": 100 },
      "presentation": "two_sided",
      "meaningLine": "How steady they stay under pressure, and how quickly they recover."
    }
  ],
  "strengths": {
    "framing": "Strongest within this person's own profile. This compares them to themselves, not to other people — and it is an ordering, not a measured difference between their dimensions.",
    "dimensions": ["Trust"]
  },
  "to_probe": {
    "framing": "Lowest within this profile — worth probing in an interview. Again an ordering within this person, not a measured gap.",
    "dimensions": ["Grit", "Influence"]
  },
  "adaptive_load_reading": "We asked about several of these twice — once about how they are in general, and once about how they are when other people are relying on the outcome. Their answers did not shift measurably between the two: they describe themselves the same way either way. In practice that usually means what you see of them in an interview is close to what you get under pressure, so there is less to probe here than with someone whose answers moved.",
  "openness_reading": {
    "title": "How they meet new ideas",
    "body": "In their answers, they prefer to work from what they already know rather than go exploring, they tend to stay with proven ways of doing things, they value doing things the customary, expected way, and they give more attention to whether things work than to how they look. Worth knowing when the work depends on stability and consistency. This reading describes how they engage with ideas and change. It is not part of the five scores above and is not something to rank people on."
  },
  "quality_banner": null,
  "interests": null,
  "onet_interests": null
}

This example predates four fields. A current result also carries, on each dimension, a one-word "confidence" (High, Moderate or Low), and the keys what_changed (null until the person's TIGER is read again, then it explains what moved and why), role_families (the kinds of work the five readings point at, null when the sitting fails its quality checks), interests_boundary (null on Core) and decision_report (the report written for the decision the sitting was for). The API reference (section 9.2) defines each field.

On the retrieval route, this object arrives inside an envelope with the session identifier, the respondent's identity as you supplied it, the completion time and an engine field. In a signed webhook delivery, the same object arrives under a singular result key. The exact envelopes are in the API reference.

5.2 Field by field#

engine identifies the shape of result you hold. Branch on it before reading anything else, because a partner account can hold more than one assessment.

tier is core or student for a partner-commissioned sitting.

contract_version is the version of the result shape. Store it, and do not branch on it unless told to.

bar_description describes what the range means. Render it verbatim on the screen where the bars appear. It is the line that stops readers treating the estimate as exact.

dimensions has the five readings, always all five, always in this order.

• name: one of the five. Do not translate or relabel.
• value: the estimate as a whole number. Do not add decimals by averaging, scaling or converting.
• bar.lower, bar.estimate, bar.higher: the three points of the range. bar.estimate equals value. Draw from bar.
• bar.axisMin, bar.axisMax: the scale ends, always 0 and 100. Draw each bar against the full axis. Do not auto-scale the axis to the data, because that exaggerates small differences.
• presentation: how to phrase the headline. two_sided reads as the estimate. at_or_above and at_or_below occur when the range would otherwise pass the end of the scale, and read as "at or above 82" or "at or below 18". Handle all three.
• meaningLine: the plain sentence for the dimension. Show it beside the bar.
• confidence: how solid the reading is, as High, Moderate or Low. Show it beside the bar. Never convert it to a number or percentage, use it as a score, or use it to place one person above another.

Note: casing is not uniform in this payload. Outer fields are lower-case with underscores (bar_description, adaptive_load_reading), while fields inside bar and meaningLine are camelCase. This is kept for compatibility. Copy the names exactly and do not normalise them.

What you can conclude about Anya. Trust reads high, with a range of 71 to 90, which is the one dimension clearly toward the higher end. Influence, Execution and Resilience read somewhat above the middle, with ranges roughly seven to ten points either side. Grit reads 48 with a range of 39 to 58. The range covers the middle of the scale, so Kirnova would not call Grit either way, and your screen should not either.

What you cannot conclude. You cannot say she is "an 81 on Trust". The estimate is 81 with room from 71 to 90. You cannot say her Execution (65) is meaningfully above her Resilience (64), because those ranges nearly coincide. You cannot say she is above or below another candidate without comparing ranges, and even then see section 2.

strengths and to_probe. Exactly one dimension is named a strength and exactly two are named worth probing. A dimension never appears in both. Two of the five carry no label. Both come with a framing sentence, which you must show with the names. The sentences say the list is an ordering of the person against themselves, not a measured difference between dimensions and not a comparison with anyone else. Anya's Trust is her strength because it stands highest in her own profile. Grit and Influence are worth probing because they stand lowest in hers. That is not a claim that her Trust is better than someone else's or that her Grit is a weakness. Use to_probe to design interview questions, not to reach conclusions.

adaptive_load_reading is words only, with no number, and is present only on sittings that administer the questions behind it. Several questions are asked twice: once about how the person is in general and once about how they are when others rely on the outcome. This reading says whether the answers moved. When they do move, the reading says so and says it cannot tell adaptability from strain. Show that sentence. This field is empty on some sittings, including every sitting with a quality note. When it is empty, render nothing.

openness_reading has a title and body in words, describing how the person engages with ideas and change. It is not a sixth dimension: it has no number or bar and sits outside the five. Its closing sentence says it is not something to rank people on. Do not add it to your scoring, filters or comparison views. It arrives on every TIGER Core and TIGER Student sitting, including those with a quality note.

quality_banner is null here because Anya's sitting was clean. If present, it carries the sentence in section 4 and the display requirement applies. Check this field before rendering anything else.

interests and onet_interests are null because TIGER Core has no interests block (section 9).

5.3 The same person after a poor sitting#

For contrast, here is what the same scenario produces when the answers do not meet quality checks, with the same synthetic person and instrument:

The quality note carries the sentence in section 4, and the under-demand reading is withheld.

Every range here spans most of the scale. The system has produced five wide bars, not five mediocre scores. A screen that showed only the estimates would show a row of numbers near 50 and suggest an unremarkable candidate, which is the opposite of what the result says. This is the strongest reason never to print an estimate without its range.

6. How to display a result#

A correct screen#

A correct presentation of a TIGER result contains at least:

1. The quality note first, when there is one, above everything else and in full.
2. All five dimensions, never a selection.
3. For each dimension, a bar drawn against the full 1 to 100 axis, with the lower point, estimate and higher point legible. Kirnova's own axis labels are Lower, Our estimate and Higher.
4. The sentence describing what the bar means, once, near the bars.
5. The meaning line and the one-word confidence reading beside each dimension.
6. The strength and the two dimensions to probe, each with its framing sentence, presented as prompts for conversation.
7. The written readings where present, and nothing where absent.
8. One person per screen. A layout that holds only one person cannot become a leaderboard.

Four screens to avoid#

The bare number.

Trust 81 · Influence 59 · Grit 48 · Execution 65 · Resilience 64

Five numbers with no ranges publish precision the assessment does not have. A reader sees 65 and 64 and perceives a difference that is not there. They see 48 and read "below average", when Kirnova said it would not call Grit either way.

The ranked list.

1. R. Menon — 71 · 2. A. Kulkarni — 64 · 3. S. Iyer — 61 …

A candidate table sorted by dimension ranks people on a midpoint whose range it discards. The order is largely an artefact of rounding. It presents an order as a finding when the ranges overlap, and it implies an instruction to start at the top and stop when you have enough, which is an automated shortlisting rule. An order that shows overlapping readings as level, keeps every range visible and leaves the decision to a person is a different screen.

The pass/fail badge.

Trust: pass · Influence: pass · Grit: flag · Execution: pass · Resilience: pass — PASSED (4/5)

A threshold applied to the estimate and shown as a verdict. The threshold is yours, and nothing in Kirnova's output supports one. There is no pass mark, cut-off or minimum in TIGER, and no setting creates one. A badge turns a reading with a range into a binary about a person and leaves the decision-maker nothing to decide.

The hidden quality note.

Scores on the summary card, with the note on a detail page, in a collapsed panel, or shown to the candidate but not the recruiter.

The note is the only thing on screen that says the numbers must not drive a decision. Placed where the decision-maker will not read it, it stops working, and a poor sitting becomes indistinguishable from a set of mid-range readings.

7. Requirements when you display a result#

These pass through to your interface.

1. Never show an estimate without its range.
2. Never order people on the middle number alone. Where two people's ranges overlap, show them as level.
3. Always show the quality note when present, wherever the scores appear, in the same words for everyone.
4. Never publish a percentile, standard error, probability or coverage figure, and never turn the one-word confidence reading into a number. Kirnova does not send these figures, so do not compute substitutes.
5. Never derive a threshold, pass mark, cut-off or total from any part of the payload.
6. Show framing sentences with the things they frame: the bar description, the strengths framing, the probe framing, the meaning lines and the closing sentence of the openness reading.
7. Interests are never a hiring signal (section 9).

8. What a result may be used for#

TIGER informs. Humans decide.

A TIGER Core result may support and inform any employment decision, provided a person makes the final decision and the organisation follows applicable law. That includes hiring, promotion, retention, pay and compensation, performance, development, management, disciplinary decisions, termination and other employment decisions.

It may never make the decision. The organisation using a result remains responsible for reviewing the intelligence, considering other relevant information, applying human judgement, complying with employment law and making the final decision.

As requirements on your integration:

• No pass mark, cut-off or minimum score. None exists in Kirnova, and none may be built on top of it.
• No order that treats overlapping readings as different. Where you place people in order, readings with overlapping ranges are shown as level.
• No automated rejection, filtering, screening-out or advancement driven by any part of the result. This includes a person clicking a button on a list that the numbers ordered.
• Nothing configurable into a pass mark. A suggested threshold that your users can set is a pass mark.

A TIGER result is one input beside the CV, interview, work sample, references and everything else the decision-maker knows. It is most useful for deciding what to ask. The two dimensions named worth probing form an interview agenda.

9. Interests#

TIGER Core has no interests block, so on a Core result the interest fields are null. Interests appear on a TIGER Student sitting, where a block about what a person is drawn towards is administered alongside the five dimensions. It arrives as its own section, beside the dimensions and never inside them.

Two rules apply, and they answer different questions:

• Access. A buyer, including a hiring buyer, may access a person's interest results.
• Interpretation. Interests are never a hiring signal. They are not evidence of suitability for a role or a measure of ability, and they must not be weighed in a hiring decision.

One rule is about who can see something, and the other is about what may be concluded from it. Interests are useful for guidance, development and conversation.

Your screen must visibly separate the interests section from the five dimensions, with a different block, heading and boundary line. The boundary sentence travels in the payload as interests_boundary:

"What someone is drawn towards. This is not a measure of how suitable they are for a role, and it must not be used as evidence in a hiring decision — no part of it belongs in that judgement. The five dimensions above are the part of this sitting that may inform one."

Render it from the payload, above the interests block. On results stored before this field existed, it is absent, so add the same sentence yourself.

Further points:

• Interests produce an ordering, not scores. Themes come back grouped, with level themes in the same group. There is no per-theme number, and you must not construct one. Someone who answered flatly gets one large group and a statement that Kirnova cannot separate their themes, which is a valid answer. Do not present a theme as a job title verdict or as something to avoid.
• Interests feed none of the five dimensions. Nothing a person says about what they enjoy moves Trust, Influence, Grit, Execution or Resilience.
• Licence and attribution text must be rendered. Where an interests section carries attribution or licence text from a licensed third-party instrument, that text ships in the payload and must be shown with it. It is a licence condition (API reference, section 9.4). Where a licensed instrument permits career exploration only, that restriction passes to you, and the interests may not be presented as an input to an employment decision.
• Do not render the interests section's own quality statement. One sitting has one quality verdict, and the result's quality note already states it.

10. What Kirnova does not send#

No result on any route contains:

• A percentage of any kind.
• A percentile, or any statement of where the person sits relative to others.
• A standard error, variance or other dispersion figure.
• A probability, likelihood or band probability.
• Any number attached to the word "confidence", or any coverage figure. The confidence reading Kirnova sends is one word: High, Moderate or Low.
• A reliability figure or psychometric coefficient.
• A formula, weighting, item-level score or description of how the five dimensions are derived.
• A pass mark, threshold, cut-off or rank position.
• A profile name, type or archetype.

This is deliberate. A probability or standard error on a screen invites precision the assessment does not have and becomes an input to a spreadsheet, then a threshold, then a rule. The range does the same job in a form a human reader can only use as intended. Also, what a buyer gets from Kirnova is more decision intelligence, not access to the engine, so the formulas, weightings and inference rules behind a reading are not shared.

11. What a result does not tell you#

• Whether someone can do the job. It reads behaviour and preference, not skill, knowledge, experience or capability. A work sample and a structured interview remain the best evidence.
• How they will perform. It is evidence about how a person is likely to work, weighed by a person alongside everything else known.
• Where they sit relative to other candidates. Kirnova publishes no comparison to a population, and you should not construct one.
• Whether a shift under demand is adaptability or strain. Where the under-demand reading appears, it says the same shift can mean the person adjusts well or that meeting the demand costs them something, and this evidence cannot tell which. That is a question for a conversation.
• Why. No part of the payload explains the cause of a reading.
• Who wins a close call. If two candidates' ranges overlap, the assessment gives no tie-break.
• A fixed characteristic. It reads one sitting, and a repeat sitting can read differently. Where movement between sittings is described, it is in words and usually says Kirnova would not call the movement either way.
• Anything about anyone else. Results today include no team reading, comparison of two people or organisation-level view.

How long you can see a result#

A result you have received stays readable through the routes it arrived on, with no timer on your access. A result read inside a signed-in Kirnova account follows a separate window described in commercial, data and terms (section 6). That window is a platform rule and not your integration's access window.

The person who sat the assessment keeps their own result in their own Kirnova account. Kirnova does not publish a retention period, so do not state one to your users.

12. Summary#

Five dimensions. Each is an estimate with a range and a one-word confidence reading, and the range is the honest part. Never show a score without it. Never order people on the middle number alone: overlapping readings are level. Show the quality note whenever it is present, in the same words, to everyone. A reading that covers the middle means "we would not call it", not "average". Interests may be seen and may never be weighed in a hiring decision. TIGER informs. Humans decide.

Questions, corrections or a mismatch between this page and the API: team@kirnova.com. Include the session identifier, roughly when it happened and what you received.