Skip to content
7Analytica

Method Six steps

How a score is made.

7Analytica reads news reports about world events and turns them into one record you can check. Follow it in six steps, then move a calculator to see how a score is built.

Read About this page

Picture first, no background needed

7Analytica turns a pile of news reports into one record you can check, in six steps. On this page: the steps as pictures, then a calculator to play with. No background needed.

Illustrative

01 The pipeline

Six steps from reports to a sealed record.

The numbers in these figures are made up to show the idea. They are not scores of any real outlet. In plain words: 7Analytica gathers many reports about one event, picks out what each one claims, and grades how well the evidence holds up.

The six stages in order: Collect the reports, Pull out the claims, Score how well it stands, Check six questions, See who it touches, Seal it and share it.

Read The six stages in pictures

The pipeline

  1. 01Collect
  2. 02Extract claims
  3. 03Score
  4. 046-axis audit
  5. 05Impact
  6. 06Dossier
Figure 1. From a pile of news reports to one sealed, checkable record. Select a stage to jump to it.
  1. Stage 1: Collect the reports

    Wire agencyDaily paperLocal blog14:05 IST

    7Analytica gathers reports from many outlets, stamps the time each was seen and merges copies. This step is called .

    Example: 11 reports about one event arrive. Seven only repeat the others.

  2. Stage 2: Pull out the claims

    port closed TuesdayCLAIM 1Port closed on Tuesday

    Each report is split into single checkable statements, the claims, so each can be followed back to where it started (its ).

    Example: “Port closed Tuesday” becomes Claim 1, with four independent origins.

  3. Stage 3: Score how well it stands

    Reliability82Corroboration64Recency91Documents55Framing71

    Five readings are combined into a (how well the evidence stands) and a (how far the wording leans). Each input is shown.

    Example: 82, 64, 91, 55 and 71, weighted, give trust 72 out of 100.

  4. Stage 4: Check six questions

    WHOWHATWHENWHEREWHYHOW

    The asks who, what, when, where, why and how, and shows how sure the record is on each, and which are disputed.

    Example: “When” is 62% sure and marked contested. “What” is 90%.

  5. Stage 5: See who it touches

    LowMedHighSupply chainFinancialLegal

    The crosses areas of life (such as supply chain, finance, law) with how much each matters for this event.

    Example: supply chain is high, finance medium, legal low.

  6. Stage 6: Seal it and share it

    DOSSIERSEALEDclaimsscoressourceshistory

    Everything is packaged as a . A freezes it, so later changes appear as new entries, never rewrites.

    Example: the dossier is sealed at 14:05 IST and can be handed on.

All numbers in these figures are made up to show the idea. They are not scores of any real outlet.

02 Try it

Move the inputs. Watch the score move.

A four-input version of stage 03, with sample inputs. It is not a live score of any real outlet.

Trust 75 out of 100. Bias 40 out of 100, tier 0, Nominal.

Start from a sample

Custom inputs. Move any slider, or pick a sample to start from.

Trust: who is telling it
Source longevity120years

How long the outlet has published. Age earns credit slowly, and is capped.

Corroboration70/100

How far independent origins agree. Repeats of one origin do not count twice.

Ownership conflict30%

How much the owner has a stake in how the story is told. Higher costs the score.

Veracity track record80/100

How often the outlet's earlier claims held up when checked.

Bias: how it is told
Accountability55/100

Low when passive wording hides who did what.

Intensity40/100

Density of loaded, high-arousal wording against a neutral baseline.

Framing45/100

Whose voice anchors the telling, and who is cast as the actor.

The fourth bias axis, veracity, is not set by hand. It is read from the Trust result: the less a source can be trusted, the more unverified claims it carries.

Readings, built live

Illustrative

Trust

75

Bias

40

Tier 0 · NominalPerspective Slider off

This source reads as dependable on the record (trust 75), and the wording reads as plain (bias 40). The bias is under 65, so the audit would not ask for a second account by default. A low trust score does not say a claim is false, and a high one does not say it is true. It says how much weight the record supports.

Worked example, line by line

Illustrative

All lines shown

  1. 01

    Source longevity

    50 + 120 / 4, held between 30 and 99

    80

  2. 02

    Corroboration

    as set = 70

    70

  3. 03

    Ownership conflict

    100 − 30 = 70

    70

  4. 04

    Veracity track record

    as set = 80

    80

  5. 05

    Trust score, the average of 01 to 04

    (80 + 70 + 70 + 80) / 4 = 75, rounded

    Each part against the average: +5 longevity, −5 corroboration, −5 ownership, +5 veracity.

    75

  1. 06

    Accountability, inverted

    100 − 55

    45

  2. 07

    Intensity

    as set = 40

    40

  3. 08

    Framing

    as set = 45

    45

  4. 09

    Veracity axis, read from Trust

    100 − 75, held between 28 and 80

    28

  5. 10

    Bias score, the average of 06 to 09

    (45 + 40 + 45 + 28) / 4 = 39.5, rounded

    Tier 1 above 48, tier 2 above 62, tier 3 above 78. The Perspective Slider turns on above 65 (the dark tick).

    40

Read What the calculator does

Build a score yourself

Move the sliders and watch the and readings assemble, with the arithmetic beside each step. This is a simplified four-input version of stage 03 with sample inputs, not a live score of any real outlet.

03 How to read it

A score describes the evidence. It does not decide.

  • Trust: the teller
  • Bias: the telling
  • Tiers: attention
  • Never a verdict

Read Plain terms, stated limits

Plain terms, stated limits

Trust is about the teller
Four readings about a source are averaged with equal weight: how long it has published, how far independent origins agree with it, how little its owner stands to gain from the telling, and how often its earlier claims held up.
Bias is about the telling
Four axes read the wording of a report: is the actor named, how loaded are the words, whose voice anchors the account, and how much is attributed but unverified.
Tiers: how much attention
A bias reading above 48 raises a -1 mark (elevated), above 62 tier 2 (critical) and above 78 tier 3 (systemic). Above 65 the audit offers the Perspective Slider, which shows other outlets' accounts of the same event side by side.
What it does not say
A score describes how well the evidence stands. It does not say a claim is true or false, and 7Analytica describes rather than advises. The weights are shown so you can see what weighing them differently would do.

The whole method in one document, with a worked example and the limits, is the methodology whitepaper. Read it on the page or download it as a PDF in light or dark. To cite it, or to see which method version a score was built under, use how to cite 7Analytica.

Longer write-ups are in the library: how a score is built, reading the six axes and all methodology articles. To see scores used on a whole record, read a sample investigation.

Next

Now see a whole record.

Open the live demo, or ask for access and we will walk through your own use case.

In brief

How does 7Analytica score media bias and trust?

7Analytica turns reports into a record in six steps: collect, extract claims, score, audit six questions, map impact, and seal. Trust estimates how well the evidence stands, using source reliability, corroboration, recency and primary documents. Bias reads how far wording leans. Every input is shown, and no score is a verdict of true or false.

Last reviewed · See the glossary