Skip to content
cube4devcube4dev
Explore & search Contact
EN
DeutschEnglishEspañolFrançaisItaliano简体中文日本語
Minimal Organization: SOVIRA, Social network – the site in the model from above at an angle

Simulation · Minimal Organization · Digital platforms

Social network

A social network as a walkable model: how a report moves through triage, moderation, statement of reasons and appeal, language by language.

Start with passwordRequest the password if you are interested →

All data are fictional – SOVIRA does not exist. Fitted to your platform's real data, the model shows what nobody sees at a glance today: report volumes by language and category, or decision times and appeal rates per team.

The content moderation process, step by step in a model. At SOVIRA, a fictional social network, content is made in seconds and reviewed by people, language by language. The main process follows a report through automated triage, priority for trusted flaggers and moderation to a statement of reasons with an entry in the transparency database. If someone objects, it continues to the internal appeal and to out-of-court dispute settlement.

The site has three halls for product, recommendation and ads, the moderation centre, the compliance cube and a data centre. Before an election five times as many reports arrive for three days, and you can see where moderation stops keeping up. Alongside, videos go online, and the annual risk assessment runs as a mandatory cycle with an external audit.

The steps follow the notice-and-action procedure of the EU Digital Services Act. The 24-hour target is set by the model, not a general legal deadline. This page is for anyone at a platform who is responsible for moderation, compliance, product or communications.

Content is reported

  1. 01

    Users report content: A hundred users report a post as hate speech.

  2. 02

    Automatic triage: Classifiers sort by severity and language.

  3. 03

    Priority: trusted flaggers: A moderator reviews at once.

  4. 04

    Moderation reviews: Moderation reviews: illegal, against the rules or neither.

  5. 05

    Statement of reasons and transparency database: Those affected receive a statement of reasons; the decision goes into the public database.

  6. 06

    Internal appeal (not only automated): An appeal is reviewed by people, not only machines.

  7. 07

    Out-of-court dispute settlement: An independent body decides.

  8. 08

    Done.

How a platform can use the model

  • Internal communication

    The diagram ‘Where people work’ shows where in the company content is reviewed. The moderation contractor in another city is part of the model as an outside party.

    For whomLeadership, internal comms, works council

  • Onboarding

    New teams in product and recommendation see what happens after a report: triage by classifiers, moderation, statement of reasons, appeal.

    For whomNew teams in product, integrity and moderation

  • Training

    New moderators play through ‘Language team unavailable’: for two weeks a third of their colleagues are missing. They see which queues grow first.

    For whomModerators, shift leads

  • Explaining to customers

    Show users, creators and the public what happens to a report and where a statement of reasons comes from. ‘How a case flows’ shows the seven steps with volume bands.

    For whomCommunications, creator programmes, regulators

  • Strategic overview

    Better triage or more moderators? ‘What each control does’ weighs technology against people, and the crisis event before an election shows where the limit lies.

    For whomBoard, product, head of moderation, crisis team

What is shown – why – for whom

What is shown Why For whom
Diagram ‘How a case flows’: from report to dispute settlement WhyShows the legal procedure step by step. For whomCompliance, regulators, the public
Moderation backlog (Resources view) WhyShows where the 24 hours are missed. For whomHead of moderation, workforce planning
Diagram ‘Where it gets tight’ during a crisis event before an election WhyShows the limit before it is reached. For whomCrisis team, management
Diagram ‘What each control does’: triage versus moderators WhyWeighs technology against people. For whomProduct, head of moderation
Diagram ‘Where people work’ WhyMakes visible the moderation work nobody else sees. For whomCommunications, works council
Data view: classifiers and transparency database WhyShows what the procedure depends on. For whomIT, compliance
Process ‘The annual risk assessment’ WhyShows the mandatory cycle alongside daily business. For whomCompliance, board

Departments in the model

  • Board

    Leads product, ads and integrity.

  • Feed and video

    Builds what users see.

  • Creator programmes

    Shares revenue so content gets made.

  • Recommendation and ranking

    Decides who sees what first.

  • Integrity

    Classifiers against spam and illegal content.

  • Moderation

    Reviews reported content – language by language.

  • Ads

    Sales, auction, ad library.

  • Compliance (independent)

    Risk assessment, reports, audit – independent of product.

  • Legal

    Orders, proceedings, reporting duties.

  • Platform operations

    Data centre, media, resilience.

All processes in the model

Content is reported

  1. 01Users report content: A hundred users report a post as hate speech.
  2. 02Automatic triage: Classifiers sort by severity and language.
  3. 03Priority: trusted flaggers: A moderator reviews at once.
  4. 04Moderation reviews: Moderation reviews: illegal, against the rules or neither.
  5. 05Statement of reasons and transparency database: Those affected receive a statement of reasons; the decision goes into the public database.
  6. 06Internal appeal (not only automated): An appeal is reviewed by people, not only machines.
  7. 07Out-of-court dispute settlement: An independent body decides.
  8. 08Done.

A video goes online

  1. 01Creator uploads a video: A thousand creators upload videos.
  2. 02Transcoding and hash matching: Transcoding, matching against known illegal content.
  3. 03Moderation reviews: A match goes to moderation.
  4. 04Rank into feeds: Recommendation sorts the video into feeds.
  5. 05In the feed.

The annual risk assessment

  1. 01Identify systemic risks: Compliance assesses risks: illegal content, fundamental rights, elections, minors.
  2. 02Measures in the product: The product changes: defaults, reach limits.
  3. 03External audit: External auditors give an opinion.
  4. 04Report to the Commission: The report goes to the Commission.
  5. 05Commission request for information: The Commission asks for more.
  6. 06Closed.

What you can change

Controls

  • Report volume: After an event many more reports arrive.
  • Moderators: More moderators, shorter waits – and higher cost.
  • Reach limit for new accounts: Less spam and fewer reports, slower growth (model assumption).
  • Better triage: Clear cases are reviewed faster (model assumption).

Disruptions

  • Crisis event before an election: For three days five times as many reports arrive.
  • Regulatory order: For two months every review takes longer: new rules, new training.
  • Language team unavailable: For two weeks a third of the moderators are missing.
  • Classifiers fail: For two hours nothing is triaged.

What the model gives a social network

  • 01

    The procedure as a path

    Seven steps from report to dispute settlement, each with its own department: walkable, instead of hidden in a legal text.

  • 02

    Load before the election

    The simulation runs queues at every step. Five times the reports for three days show where the 24 hours break – before it really matters.

  • 03

    Duty next to daily work

    The annual risk assessment, with external audit and a report to the Commission, runs in the same model as daily moderation.

  • 04

    Work people rarely see

    The moderation centre and compliance cube appear as buildings full of rooms and people, which helps explain their work inside and outside the company.

  • 05

    Designed, not assembled

    An artist and developer designed the model: a calm world of blocks, in the browser and on large projections.

Basis and limits

Fictional social network · model based on researched industry structures · no forecast for real companies

1 case = 100 reports, 1000 videos or 1 risk assessment · 1 figure ≈ 50 employees

All volumes, durations, rates and capacities are model assumptions, not values of a real platform.

The 24-hour target is a model goal, not a general statutory deadline.

Documented: notice and action, statement of reasons and public database, internal appeals, dispute settlement, risk assessment and audit under the Digital Services Act.

Public sources the model is based on:

  • Gesetz über digitale Dienste, VO (EU) 2022/2065
  • DSA-Transparenzdatenbank der Kommission
  • Koordinator für digitale Dienste in Deutschland

Your platform as a model

Would you like to show your notice-and-action process, language teams or risk assessment like this? Get in touch in writing or by phone and we'll work out which figures belong in the model. We'll send the password that opens SOVIRA on request.

Start with password

Request the password if you are interested →

Ask about this

Minimal Organization: Soziales Netzwerk

What you enter goes to this site alone and serves only to answer you.

Similar business models

← Minimal Organization

All data are fictional – SOVIRA does not exist. Fitted to your platform's real data, the model shows what nobody sees at a glance today: report volumes by language and category, or decision times and appeal rates per team.