Specifications
The long version, for the people who need it.
The technical sheet of the platform. One entry per module: a sentence on what it does and for whom, then the facts, one dash at a time.
Thirty-eight entries in five families: Learn, Produce, Sell, Measure and Run. The page stays in the menu, under Resource Centre, and in the footer, for the day you need to check a number.
In detail
Module by module.
Five families, thirty-eight entries. Open the one you came for.
Learn
01Short Format Learning
Vertical episodes an advisor finishes between two customers.
- One episode at a time, swiped like a phone feed
- A single video, or a carousel of images and clips
- Four kinds: video or photo, article, podcast, survey
- Podcasts carry a cover, chapters and a transcript
- Articles mix text, photos, questions and episode links
- A quiz on any episode, XP and a timer per question
- Question kinds: multiple choice, true/false, poll, open · 5 answers max
- Grouped into quests, or sent alone from HQ
- Targeted to everyone, or to named audiences
- Published, scheduled with a teaser, or switched off
- Served in each advisor's language, subtitles and dubbing included
- Likes and bookmarks, shown or hidden per episode
- 54 content languages · 9:16 adaptive streaming
02Long Format Learning
Programs turn the Maison's documents into chaptered courses, translated for every market.
- An ordered set of chapters, each holding slides
- Six slide types: video, image, PDF, quiz, link, SCORM
- Import a PDF or slide deck as the starting point
- AI proposes the structure: chapters, metadata, chapter quizzes
- Each proposal reviewed before it applies
- Marked mandatory or optional
- A language added three ways: from scratch, metadata only, full translation
- Translated in batches of up to 200 slides, 1 credit each
- Machine output corrected slide by slide
- Dubbing per slide: pick a voice, preview, correct each segment
- Subtitles generated for program videos
- Progress tracked per chapter and per slide
- Default slide duration set per Maison, 5 to 255 seconds
- Uploads 2 GB per file · 20 files, 8 GB per run
- Auto-structure handles up to 300 slides
03Live Session Training
One host leads invited advisors through a playlist, in the room or over video.
- A session drafted, or scheduled for a date
- One broadcast language per session
- Playlist items: episodes, programs, chapters, slides, profiles, products, quizzes
- Invitations by name or email, answers visible to the host
- The host turns the page for everyone at once
- Play, pause, rotate and slide changes reach the whole room
- The stage switches to the host's camera
- The floor granted and revoked person by person
- Participants raise a hand to speak
- One control mutes everyone
- Late joiners land on the exact slide in play
- Quiz answers stream in live, results shared back
- A report opens when the session ends
- Every host action logged, append only
- 2 to 500 participants, 50 by default
- Switched on per Maison, with named authorised hosts
04SCORM
A Maison's existing SCORM packages play inside a program and report back like native content.
- SCORM 1.2 and SCORM 2004 both supported
- A package uploads as a zip, attached to a chapter
- Manifest parsed, assets extracted to object storage
- Standard SCORM JavaScript API exposed inside the player
- Progress saved per person and per slide, suspend data included
- A session resumes on another device from its bookmark
- Completion and score feed the program completion pipeline
- Attempts tracked per person and slide
- Interactions captured, with a per-package report
- SCORM signals map to platform KPIs, presets included
- A package missing a signal still reports what it has
- Assets served through signed URLs
- Progress saves capped at 1 MiB
05The feed
The feed decides which episodes an advisor sees, and in what order.
- An episode reaches the feed through its audience targeting
- Audiences built on market, store, country and role
- HQ hides whole categories of content
- HQ boosts or buries by signed weight
- Five dimensions: kind, subtype, media type, category, quest
- Order: unfinished first, then boost score, then newest
- A preview shows a ranked sample and each score
- The rules that matched each episode, listed
- The preview and the live feed share one ranker
- Zero rules leaves the order unchanged
- 50 episodes per page
06The home screen
HQ composes each advisor's home: hero, push messages and ordered rails, market by market.
- Several home templates per Maison, each aimed at markets
- One default template as the fallback
- A hero with pinned content and its own cover
- Separate covers for desktop and mobile
- An optional hero slideshow, each slide promoting its own content
- A looping muted hero video, image kept as poster
- Rotating push messages with copy, a label and a destination
- Sections dragged into order and switched individually
- Three section modes: manual, dynamic, objective
- Manual rails mix episodes, quests, products, stores and programs
- Dynamic rails fill themselves: trending over 30 days, or unfinished content
- A suggest bar of quests, content and product categories
- Pinned items first, then a rule fills the rest
- Titles and messages translated per language
- Pull to refresh updates the page in place
07Live Stories
A story posted from the phone tops the team's home, then expires on its own.
- Posted from the app: camera, or the device library
- One photo or one vertical video per story
- A short title, an optional caption
- Visible 1 hour to 7 days, six presets
- Posted to every market, or a chosen list
- An optional push to the audience
- Video encodes in the background, the author can leave
- The cover picked from five frames, changeable after publication
- Viewed and unviewed marked per advisor on the rail
- A tap opens the full episode
- Deleting a story keeps the episode in the library
- Stories stay out of the feed, search, home rails and quests
- Expired stories drop off by themselves
- Posting is a granted right, per person
- Photo to 10 MB · video to 150 MB, 90 seconds
08Community
Advisors talk under the episodes; HQ reads every conversation in one inbox.
- Comments and replies under episodes
- Every comment from every episode, in one list
- Filtered by episode, advisor, period, read state, status
- Free-text search across the inbox
- New comments land in the inbox live
- HQ replies, approves or rejects from the same place
- Rejected comments sit in a trash, restorable with their replies
- Permanent deletion exists, behind its own permission
- Any comment translated for the reader, on demand
- Episodes and regions ranked by comment volume
- A moderation widget on the console home
- Comments up to 5 000 characters
09Objectives and guidance
HQ sets a goal, aims it at an audience, and follows each advisor's progress.
- Business objectives target 1 to 50 products
- Episodes tied to those products surface on their own
- Learning objectives run on a quest or a custom playlist
- Playlists mix episodes, programs, quests, categories and products
- A dedicated flow for finishing a program by a deadline
- Deadline fixed, or counted from each person's assignment
- Aimed at the whole brand, markets, stores, audiences or named people
- New joiners can get their own rolling window, 1 to 180 days
- Priority decides when several objectives apply
- Up to 5 reminder dates before the deadline, push and email
- The advisor's home shows one objective card, with what remains
- A streak counter: days in a row
- Progress per person: not started, on track, behind, urgent, completed
- A dashboard: assigned, started, completed, and a 14-day curve
- The matched audience previewed before publishing, person by person
- Switched on per Maison
10Quests, XP and rankings
Learning earns XP points; points build levels, and levels carry benefits the Maison defines.
- XP points earned on activity, quiz answers included
- Every award lands in a ledger, tied to what earned it
- Levels defined per Maison and per market, one default
- Level names translated per language
- Benefits attached to each level, with their own values
- Benefits switched on or off per market
- Annual XP reset on a date each market chooses
- The reset writes a negative ledger entry, so history stays auditable
- Store and country rankings for market managers
- Team highlights for store managers
- A personal activity page lists finished and unfinished content
- Monthly activity snapshot on the last day of the month
- Quests show each person's current position
11Notifications
One switchboard decides which events reach an advisor, and through which channel.
- Two channels: push on the phone, email
- An in-app feed carries every notice, always on
- Advisors set push and email per event type
- Events grouped by module, then by category
- One reset returns every choice to the defaults
- A kill switch silences everything at once
- The person who caused an event is never notified of it
- Only plausible events appear in each person's list
- Push defaults on, email defaults off almost everywhere
- Failed emails retried five times, up to eight hours apart
- HQ sends its own push campaigns from the console
- 30 event types across 7 modules
12Offline mode
Episodes and programs download to the phone and play with no connection.
- The unit is a whole episode or a whole program
- A download carries video, subtitles and poster
- Programs bring their dubbed audio too
- Image and PDF slides come along with the video
- Subtitles and audio in the advisor's language only
- Quality picked automatically, up to 720p
- A server manifest lists exactly what to fetch
- A re-encoded source flags its download as stale
- A download that would fill the device is refused
- A downloads page: items, storage used, slots used
- Delete one download, or all of them
- Offline, a banner and the app's start lead to the library
- 20 downloads per device · iOS and Android
- Switched on per Maison
13Documents
A library of files HQ publishes and advisors open in the app.
- A folder tree with sub-folders, ordering and descriptions
- Upload, rename, move, reorder and delete from the console
- Advisors browse, search and download in the app
- Full-text search across everything an advisor can open
- Signed view links for in-app preview
- Access profiles built from six granular capabilities
- Ready-made profiles: administrator, editor, viewer
- Profiles assigned one by one or in bulk
- Folders and files restricted per market
- An activity log per file: views, downloads, changes
- Logs filtered by action and by period
- A storage quota per Maison, with a usage view
- The same files serve as a picker in merchandising missions
- Files up to 50 MB each
14TU Agent
Advisors ask in plain language and get a sourced answer from the Maison's own knowledge.
- An AI mode inside the app's search
- Answers stream in and cite their sources
- A cited source opens the episode, program or product
- Conversation history kept per advisor
- The Maison names the assistant and writes its instructions
- Knowledge drawn from the Maison's episodes, programs and products
- Episodes join the knowledge base as they publish
- HQ adds documents: text, Word, PDF, images, audio, video
- PDFs read page by page, images through OCR, media transcribed
- Web pages crawled from a sitemap, on a schedule
- Every source scoped to chosen markets
- Reader Digest: a transcript and summary per episode, in the advisor's language
- HQ reads usage, coverage and what advisors actually ask
- A monthly limit for the Maison, a budget per advisor
- Out of budget, the assistant closes and says why
- Switched on per Maison
Produce
15Video Studio
HQ edits video in the browser and exports straight into an episode or the library.
- A video editor embedded in the console, in the browser
- Named projects per Maison, listed and searchable
- A project starts from videos picked in the Media Library
- Projects link to an episode and to a specific slide
- Autosave with versioning: a stale save is rejected
- Export to the Media Library as a Maison-owned asset
- Export can replace a slide's master or add a new slide
- The previous video stays in the library after a replacement
- Replacing a master flags subtitles and dubbing for regeneration
- Progress reported in phases: rendering, encoding, uploading, done
- Exports feed the standard pipeline: frames, audio, transcription, adaptive streaming
- One export at a time per project
- Switched on per Maison, with dedicated permissions
- Default export 1080×1920, 30 fps, MP4
- Imports up to 2 GB · projects 8 GB · exports 4 GB
16Subtitles and dubbing
The platform writes subtitles from the audio, translates them per market, and renders them in the Maison's style.
- Subtitles generated from the audio of any episode video
- Translation into each market's language, on the same timing
- Produced by AI or by a human, order by order
- Each order tracked end to end, proofread before publication
- Episode subtitles sit in the Proofreading queue as their own kind
- Reviewers read the source and correct the target, side by side
- Dubbing runs through the same governed pipeline
- One saved subtitle style per Maison, kept server side
- The app player applies that style at runtime
- Video exports burn the same style into the picture
- Style covers typography, colour, casing, position and background plate
- Without a saved style, the platform default applies
- 7 order states, from ordered to published
17Media Library
One catalogue holds every image, video, audio file and document the Maison owns.
- One card per asset, federated across the whole platform
- Filter by kind: image, video, audio, document
- Filter by dominant colour, orientation and aspect ratio
- Filter by tag, episode of origin, or date range
- A used versus unused filter, platform wide
- Full-text search on filename, episode and linked product
- Duplicates detected by checksum, grouped under a copies badge
- Dominant colour and palette read from every image
- Tags derived from the content each asset is attached to
- Video frames extracted and browsable as reusable stills
- A detail drawer lists sources, usages and technical metadata
- A prepare-use flow puts an existing asset on new content
- The same picker serves the console: library, frames, brand assets, upload
- 12 colour families · ratios 1:1, 4:5, 16:9, 9:16
- Signed links: thumbnails 5 minutes, original downloads 10
18Proofreading
HQ follows the human review of every translation from one queue and one dashboard.
- Covers programs, merchandising missions, quests, products, stores and episodes
- One review per content and target language, not per slide
- Progress tracked field by field inside each review
- A coverage matrix by content type and language
- Episode subtitles reviewed in the same queue
- External reviewer accounts provisioned from the console, by email
- Reviewer access scoped by language and content type
- The workspace shows source and target side by side
- Reviews are validated, and can be reopened
- A validated review turns stale when the content changes
- Baseline mode declares existing content reviewed, with a mandatory note
- Bulk requests cover content never reviewed before
- Deep links open the native editor of each content type
- Every validation, reopening and baseline writes an audit entry
19Content Requests
The Maison commissions content from ToldUntold here and follows it through to delivery.
- A request starts as a draft, submitted when ready
- Five content types: video, carousel, podcast, survey, article
- A structured brief: description, key facts, reference, publication date
- Requested languages, a primary language and target markets
- Reference media uploaded with the request
- ToldUntold answers with an estimation priced in credits
- Episodes, cost per episode and language variations, totalled live
- Validation debits the Maison's credit balance
- Commercial proposals offered as an alternative to estimations
- A threaded conversation per request, attachments included
- Unread counters, read markers and per-request mute
- Episodes attached as deliverables, each with its publication state
- Change requests, restarted production and approval per delivery
- A readiness matrix per episode and language: metadata, subtitles
- Requests close on their own after approval, 7 days by default
- 12 statuses from draft to closed
20Brand Voice
The Maison writes down its voice: identity, glossary, banned terms, and instructions per task.
- One profile per Maison, enabled by the Maison itself
- Brand DNA: values, codes, positioning
- History: founding, key dates, the story the Maison tells
- Tone of voice: register, rhythm, signature phrasing
- Writing rules as explicit do and don't pairs
- Locale conventions per language, formal or informal address included
- A glossary of house terms with definition and usage note
- Imposed equivalents per language for each glossary term
- A term without equivalents stays untranslated
- Banned terms carry a reason and a warn or block severity
- A lexical check reads any text against the banned list
- Per-task instructions override the default composition
- A preview shows what each task would carry
- Composed per task, within character budgets
- Banned terms are never trimmed from the brief
- 10 task recipes, from translation to quiz drafting to chat
21Admin agent
HQ asks a question in plain language and gets an answer from the Maison's live data.
- A conversational panel in the console header
- Answers built from live data, not from documents
- Analytics: overall summary, market performance, market detail
- Episodes: lists, detail, recent comments
- Programs: detail, plus learners lagging behind
- Quests and content pages, orphan pages included
- Catalogue: products and low stock
- Live sessions: registrants and reports
- Workforce: attendance, pending requests, idle stores
- Platform how-to answered from a built-in guide
- Every tool checks the asker's own permissions
- Non-administrators see only their markets and audiences
- Tools appear only where the module is active
- The conversation carries its history across turns
- Reserved to administrator roles, switched on per Maison
- 22 tools across 8 domains · 20 requests per minute
Sell
22Product-Focused Selling
One sheet gives the advisor the product, the content around it and its availability.
- The product sheet in the app: gallery, price, description, reference
- Short looping clips play in the gallery, up to 5 per product
- Episodes about the product, on the same sheet
- Related products follow below
- Size and colour variants, chosen on the sheet
- Live availability per variant: in stock, low stock, out of stock
- Variants and stock appear where the Maison switches them on
- The catalogue browses by the Maison's own category tree
- Refine by price, attribute, size, availability or new in
- Sort by newest, or A to Z
- One search spans products, episodes, programs, quests and stores
- Bookmarks keep a product to hand
- Product views feed the Maison's analytics
23Product catalogue
The Maison brings its products in by spreadsheet, by URL or one by one.
- Spreadsheet import from a generated template, CSV or XLSX
- Templates built for the Maison's chosen languages and markets
- Name, short and long description written per language
- HTML allowed in the long description
- Category, thumbnail, carousel, SKU, price in EUR or USD
- Imports checked line by line, then run in the background
- A row carrying a known ID updates that product
- Web Import turns a product or collection URL into drafts
- Drafts reviewed by hand, or accepted above 0.90 confidence
- A brand-defined category tree, each category shown or hidden in menus
- Attributes per product, values suggested from the existing range
- A product reaches only its markets and audiences
- Bulk duplicate, activate, deactivate, delete
- Variants generated as an axis matrix, up to 500 combinations
- Stock per location, fed by the Maison's systems in batches
- Availability computed: in stock, low, out, stale when feeds go quiet
24Clienteling
The advisor keeps each client relationship in one file the Maison owns.
- A 360° client file: identity, contact, home store, status, value tier
- Search that forgives typos and other alphabets
- A portfolio of followed clients, one referent advisor per client
- Consent held per channel: email, SMS, WhatsApp, call, post, profiling
- Every interaction on one timeline
- Notes marked team, managers or private
- An erased note purges for good after 30 days
- Structured preferences, captured by the advisor
- Playbooks turn the Maison's rules into a prioritised task list
- Opportunities flagged: tried without buying, rare item, reactivation, restock, life event
- Scores built from readable rules: value, conversion, attrition, contact cap
- Solicitation capped by tier, platinum 8 to bronze 3 per 30 days
- CRM, POS and e-commerce identifiers held on the file
- Two files merge only after a human confirms the match
- 20 audited action types, plus a right-of-access export
- Activated Maison by Maison
25Visual Merchandising
Missions leave HQ as briefs and come back as validated photographs.
- A mission targets markets or named stores, in ordered blocks
- Photo tasks and polls: checkbox, yes or no, multiple choice, free text
- A poll can require the guideline to be read first
- Submissions move pending, review, validated, rejected or reopened
- Unanswered and skipped counted too, so completion rates stay honest
- Co-assignees add photos and comments, every item attributed
- AI reads each photo first: a compliance score, 0 to 100
- Four category scores, an issues list and product checks
- Annotation on the photo: boxes and text
- A conversation per submission, with mentions and photo-anchored comments
- Reference photos from the store, brand references beside them
- Short video on a submission, 60 seconds, where the Maison allows
- A furniture library with dimensions and weights, assigned per store
- Store categories, colour coded, filter the gallery
- Assignments imported by CSV, 5,000 rows at a time
- Validated photos gather in a cross-mission gallery
Measure
26Learning Analytics
HQ reads how every market, store and advisor learns, from one dashboard.
- One summary: users, completion, quiz scores, XP, live content
- 12 months of trends: completion, watch rate, quiz score, active users, XP
- Engagement sorted into four tiers, Expert at 90 and above
- Top and flop leaderboards, completion first, XP as tiebreak
- Drill down by market, country and store
- Content performance across episodes, programs and quests
- Every store placed on a map
- A live counter of who is online
- Per-person detail with a paginated activity log
- Filters: store, job title, city, country
- Inactive stores included or excluded with one toggle
- Figures clamped to the viewer's own markets and audiences
- Fed nightly · cached 10 minutes · online counter 30 seconds
- A banner announces when the analytics store degrades
27Business Intelligence
Sales crossed with learning data, so HQ sees whether training moved the business.
- The Maison supplies the sales data
- The sales gap between trained and untrained teams, in points
- Cross-lag analysis: impact traced from 3 months before to 6 after
- Correlation strength at every offset, to see when impact peaks
- Also tracked: completion rate, variance explained, activation window in weeks
- An impact index ties each program's completion to product sales
- Completion and business impact side by side, for every program
- One dot per store: engagement against sales
- Engagement distribution has its own view
- Filters: markets, points of sale, product types
- Results export for analysis elsewhere
28Trends
A weekly score ranks the Maison's products on outside interest and inside engagement.
- The Maison picks the products to follow, up to 50
- Outside signal from search interest, inside signal from advisor activity
- A fresh score every Monday
- Any product refreshes on demand, once per 24 hours
- 90 days of history behind each sparkline
- A regional breakdown for every market
- Top and rising topics attached to each product
- Scores and the Top 10 shown or hidden, per Maison
- Episodes, programs and quests ranked by the same weekly run
- History kept one year
29Atlas
One map shows what every team knows, and where to act first.
- Rows by market, store, audience or a custom grouping
- Columns by category, quest, content type or objective
- Three readings of the same matrix: grid, opportunities, mosaic
- A swarm view, one dot per person, brush selection, up to 5,000
- Individual profiles: completion, next content, objectives, boost history
- Objectives paced ahead, on track, behind or urgent
- A 14-day completion curve per objective
- Never-activated people counted exactly, invitations resent in one action
- Custom groupings weighted and versioned, 12 active at most
- Opportunities rank the cells worth acting on first
- Boosts target a cell, a row, a column or a picked list
- Push or email, with a custom message and first name
- Quiet hours kept: sends run 09:00 to 19:00 store time
- 3 reminders per person per week, 72 h cooldown
- A holdout group proves the effect at day 7 and 14
- Completed means 90% viewed, quiz passed when there is one
30Workforce Management
Managers plan the store week; the team clocks in, trades shifts and requests leave.
- Weekly rosters per store, drafted then published
- A finished week duplicates into the next
- Shifts created singly or in bulk
- Shift types carry a colour, default hours and a severity
- Clock in, clock out and breaks, from the app
- An optional location stamp on clocking
- Time entries adjusted by hand, always with a reason
- Absence requests drawn against per-type balances
- Managers approve, and a pending request can be withdrawn
- Shift trades: a direct swap, or an open market bid
- Tasks assigned with a due date, completion tracked
- Compliance flags long days, short gaps and runs of consecutive days
- Those rules warn the manager, they never block a roster
- Labour cost, attendance and coverage dashboards
- Switched on per Maison
Run
31Accounts and sign-in
Every advisor signs in through the Maison's own identity provider, or with a password.
- Single sign-on over OAuth 2.0 and OpenID Connect with PKCE
- Identity provider set per Maison: Microsoft Entra ID, Okta, Infomaniak, generic OIDC
- Sign-in buttons derived from the Maison's allowed email domains
- First SSO login can create the account, or require an existing one
- Password login stays available as a per-Maison fallback
- Accounts arrive by invitation, admin creation, or self-registration
- Self-registration checks a domain allow-list, then waits for admin approval
- Email second factor for admins: 6-digit code, valid 10 minutes
- Short-lived access token, 15 minutes, refreshed silently
- Refresh token 10 days, adjustable per Maison
- Password reset by emailed link, valid 60 minutes
- Every authentication route throttled per user and per IP
- Expired tokens pruned nightly
32Roles and permissions
Roles decide what a person can do; Maison, market and store decide how far it reaches.
- 230 named permissions in the core registry
- Six role presets shipped, administrator down to advisor
- A Maison edits a preset or builds its own role
- Permissions checked on the server, route by route
- Modules bring their own permissions under the same registry
- Every read is filtered by the active Maison
- Market-bound data filtered by the viewer's own markets
- Store scoping through the store manager relation
- Privileged roles hidden from audiences and user lists
- Preset sizes: administrator 164 permissions, market manager 64, advisor 29
33Maisons, markets and audiences
One platform serves several Maisons, each with its own markets, stores, branding and languages.
- The Maison is the tenant: every record belongs to one
- A person can belong to several Maisons and switch between them
- Markets under each Maison, countries and stores beneath
- Store directory with managers, teams, country and category
- Per-Maison branding: hostname, app name, logos, login splash
- Per-Maison legal terms, or the platform's own
- Licence per Maison: seats and credits, expiry checked daily
- Audiences built from market, store, country or role
- Custom audiences combine predicates with explicit include and exclude lists
- Memberships recomputed in the background and fed to analytics
- Privileged roles never land in an audience
- Per-market settings for notifications and analytics exclusion
34Languages and localisation
Content is written once and read in each advisor's own language.
- One original, many language versions, stored together
- Each Maison sets its default language
- A Maison's language list follows its markets
- Each advisor's profile carries their language
- 54 locales on file, each with its native name and country
- Quiz questions and answers translated with their episode
- Reward level names translated like any content
- Machine translation, language detection, subtitling and dubbing run through the AI gateway
- The Client API mirrors content with its translation map
35Open API and integrations
A Maison's own systems read the platform through one documented REST API, and hear back through webhooks.
- One API key, scoped by domain, resource and action
- Five domains: analytics, stores, content, catalogue, webhooks
- Versioned under /v1, changes stay additive
- Signed webhooks on product create, update and delete
- HMAC-SHA256 signatures, timestamp tolerance 5 minutes
- Failed deliveries retried with backoff, then parked for replay
- Endpoint policy: HTTPS only, public addresses, redirects refused
- Idempotency keys on creation calls, kept 24 hours
- Async exports in CSV or XLSX, signed downloads, kept 7 days
- Media returned as signed short-lived URLs
- Rate limits: 120 requests/min per key · 240 per Maison
- Integrator guides and OpenAPI 3.1 contracts, English and French
- The developer portal lives at /developers
- Keys issued and scoped by ToldUntold
36Exports and data portability
Whatever the platform measures can leave it, and so can a person's own data.
- 16 analytics export endpoints: metrics, learning, stores, users, markets, content
- Per-user detail: logs, publications, product and category comfort
- Runs on demand or queued, with a signed download
- CSV or XLSX, files kept 7 days
- Advisors export their own activity, under its own permission
- User lists and campaign recipients export from the console
- Account deletion requested by the person, decided by an admin
- An approved deletion archives the profile before the record is erased
- The archive keeps preferences, quiz responses, language and picture
- Both outcomes reach the person by email
- The Client API carries the same exports for machine use
37AI governance
Every AI call on the platform runs through one gateway that logs and prices it.
- 31 routed capabilities, translation, subtitling, transcription and image analysis among them
- Routing set per capability, Maison and region, with documented precedence
- Per-Maison overrides without a code change
- Automatic fallback to a second engine when the first fails
- A provider outside a region's allow-list is refused, never substituted
- Usage ledger: one row per attempt, with latency and outcome
- Cost recorded per call, raw and re-billed
- Billing currency per Maison, exchange rate locked at write time
- Audit trail: who ran what, when, from where
- Governance access needs a named operator designation, revocable at once
- Credentials write-only: status and last four characters shown
- Connection test per credential, batch test per provider
- The whole configuration exports encrypted under a passphrase
- Health view: failure rate, p95 latency, queue depth
- Coverage view shows how far each module is wired
38Support and reporting
Anyone in the console reports a problem where it happens; the support desk takes it from there.
- Report a problem from any console screen
- Screenshot, description and page context captured together
- Capture from the page itself, a shared tab, or an uploaded image
- Annotation on the capture: mask, rectangle, arrow, text
- Personal data and secrets scrubbed from the captured context
- Screenshots re-encoded with metadata stripped
- Categories bug, question, idea · severities blocking, major, minor
- A reference like SN-00042, and a "My reports" status list
- Delivery to the support desk retries for 48 hours
- One-time 60 second code opens the external desk, after 2FA
- The desk never holds a platform token
- Duplicate reports fingerprinted for reviewers
- Ticket counts surface on the console home
Training. Intelligence. Content Production. Visual Merchandising. AI. One platform built for retail performance.