REJECTED

In Review

Backlog

Next up

Deployed to QA

Approved Through QA/Ready for UAT

UAT Approved

Submitted to CAB

Completed and Working In Production

Coach Pulse Roadmap

Define a regular reporting cadence for leadership on Coach Pulse release activity — what ships each week, QA/CAB outcomes, and upcoming scoped work. Covers format, frequency, and audience.

Reporting cadence for leadership

Harrison

Asked directly in the session following the avatar work shown at SMC. No timing set. Decision needed on what to signal to the field. Source: pilot production testing, Aug 11 2026.

Avatar capability for Train personas

Matt Britton

Store leaders asked to flag a Culture Feed item or an Ask Pulse answer so their own associates see it marked as a store priority. Copy link is the current workaround. Decision needed on whether the pin lives in Culture Feed, in Playbook, or both. Source: pilot production testing, Aug 11 2026.

Manager pin to store for Culture Feed and Ask Pulse answers

Incorrect prices appeared in generated summaries on a Teri shoulder bag and a cherry charm. Immediate mitigation is suppressing prices from generated summaries. The durable control is verifying any price shown anywhere in the application against an authoritative product source before it renders. Source: pilot production testing, Aug 11 2026.

Culture Feed: authoritative price verification before render

Store leader asked whether feed items can be carried into Playbook daily messages. Decision needed on whether that path is supported deliberately or left to copy link. Source: pilot production testing, Aug 11 2026.

Share Culture Feed content into Playbook daily messages

Roughly eighty percent of staff and four in five walk-ins at the reporting Canada location speak Punjabi or Hindi. Neither was in the original language set. Feasibility answer required first, then validation through the Canada testers who volunteered. Not in Aug 22 build. Source: pilot production testing, Aug 11 2026.

Settings: Punjabi and Hindi support for Canada (F20)

Metric questions such as how to boost UPT or AOV route to Ask Rexy. The routing rule stands. The improvement is the response: explain why it is redirecting and offer a rephrasing framed as selling behaviour that Coach Pulse will answer. Handled through training until built. Not in Aug 22 build. Source: pilot production testing, Aug 11 2026.

Ask Pulse: explain metric redirects and offer a rephrasing (F15)

Cantonese is withdrawn from selection for launch because the underlying voice technology does not support it to standard. Mandarin is the interim path. Restore Cantonese as a selectable language once the voice partner confirms support and latency and conversational quality pass QA. Source: pilot production testing, Aug 11 2026.

Train and Settings: restore Cantonese when voice support is confirmed (F9, F19)

Managers cannot tell associates whether an example clip is a real associate recording or a generated rendering. Whenever example audio ships, state its origin on screen and mirror the wording in training materials. Source: pilot production testing, Aug 11 2026.

Train: on-screen origin label for example audio (F14)

Post-session feedback is longer than associates will read. Restructure to lead with score, single highest-impact behaviour, one concrete example, with full detail behind an expand. Sixty percent of staff at one multi-unit work 8 to 12 hours a week. Build alongside the named coach idea. Not in Aug 22 build. Source: pilot production testing, Aug 11 2026.

Train: short-form post-session feedback with expand (F13)

Associates need explicit in-session copy confirming that Train practice sessions are for coaching only and are not counted against performance, evaluation, or their record.Sessions are recorded, retrievable, and visible to a store manager, so the wording needs a read from Coach training and governance before it ships. Placement to be confirmed: pre-session context screen, in-session, or both.

Planned, but No Date

Add "does not count against you" disclaimer to training sessions

normalizeEmployeeId trims, strips domain, and strips leading zeros, but leaves case untouched, and Tier 1 matches employee_id exactly and case-sensitively. On qa, 39 of 51 corporate rows carry lowercase nickname employee ids, so a differently-cased SSO claim misses Tier 1 and falls through to Tier 1b, which compares on a lowered value. No lockouts today because Tier 1b catches every case, including the 7 rows with no federated_alias. This is tier degradation rather than an access failure. Fix is to lower-case within normalizeEmployeeId or add a lowered candidate to employeeIdCandidates, applied to both the TS module and the sidecar JS copy in server/alb-identity/claims.js. Source: verification of PR 189, Aug 16 2026.

Identity resolver: case-insensitive matching at Tier 1

Field asked what good, better and best actually sound like for cultural fluency and reading the moment. Generated reference examples per scoring category avoid the consent question entirely. Decision needed on whether we build a reference library. Source: pilot production testing, Aug 11 2026.

Train: reference audio library per scoring area

Strongest idea of the session. Rather than reading post-session feedback, the associate hears it from a persistent named coach with a chosen voice, building familiarity across sessions. Overlaps with short-form feedback item and should be built together. Source: pilot production testing, Aug 11 2026.

Train: named coach that speaks the feedback

Store noise is captured during sessions and personas respond to speech the associate did not make. Requested directly by store leaders. Add a press to speak mode in the Train voice UI as an alternative to open mic, with per-user preference persisted. Interim: headset guidance in launch training plus voice partner tuning. Source: pilot production testing, Aug 11 2026.

Train: press to speak control (F12)

Playback of another associate audio is removed from the Aug 22 build. Returning it requires a consent system: capture consent at first Train session, record scope of consent, per-associate opt-out honored at retrieval time, and admin visibility into consent state. Defined with Coach Legal and HR. Source: pilot production testing, Aug 11 2026.

Train: session playback consent capture and opt-out (F8)

Revise and roll out an updated grading curriculum for Coach Pulse training sessions — the framework that defines how trainee performance in a session is evaluated, scored, and reported.
What this covers:
<ul>
<li>Updated grading criteria and rubrics — refresh the competencies each session is graded against so they reflect the current curriculum, with clear definitions for each score level to keep grading consistent across sessions and graders.</li>
<li>Scoring scale and weighting — define the scale used per criterion and how individual criteria roll up into an overall session grade.</li>
<li>Session-type coverage — ensure the updated curriculum applies correctly across the different training session types on the platform, including any multilingual sessions.</li>
<li>Feedback output — align the feedback shown to trainees and coaches with the new criteria so scores and comments map to the same rubric.</li>
<li>Migration — decide how previously graded sessions are handled (regrade, leave as-is under the old curriculum, or flag the curriculum version on each grade).</li>
</ul>
Outcome: every training session is graded against the current curriculum with consistent, transparent criteria, and reporting reflects the updated framework.

Updated grading curriculum for training sessions

Coach Pulse has a badge system. It does not have a rewards program. This post scopes what a rewards layer requires and what we commit to for launch.What exists 15 achievement badges, score badges unlocking at 80/90/100%, badge progress on Profile, and an "Earn More" section on Today. All of this is recognition — no cost, no fulfillment, no budget.What a rewards program adds Something of value that an associate receives. That means a funding source, a finite catalog, a redemption path, a fulfillment owner, and per-market HR, payroll and tax review. It is a net-new layer, not an extension of badges.Core decision — earn model Do badges convert into rewards, or do rewards run in parallel on their own criteria?<ul><li>Conversion — badges unlock rewards at a threshold. Reuses the shipped earn logic; cheapest path. Risk: badge design now has money attached, so anything gameable becomes a liability.</li><li>Parallel — rewards run on separate criteria, likely sales or performance rather than training completion. More defensible, but a new earn engine, and it depends on associate sales data in the C360 feed.</li></ul>Most other decisions follow from this one.Open decisions and owners<ul><li>Earn model — FutureProof recommends, Tapestry approves</li><li>Reward catalog and funding — Tapestry</li><li>Redemption and fulfillment path — joint; drives build scope</li><li>Market eligibility — Tapestry HR/Legal</li><li>Day-to-day fulfillment owner — Tapestry</li></ul>Proposed launch position Recognition-only. Badges ship as-is; rewards land once catalog, funding and eligibility are settled.Immediate issue, independent of the above "Earn More" is live in nine languages telling associates they can earn additional rewards. No rewards exist. This copy needs correcting regardless of what we decide here.

High Priority

Rewards for Launch

Work with China team to deploy Train module on their environment

Create new version of Train module gameplay where users sign up with email and complete 1 of 3 available personas. Scores are logged and full scoring reports emailed top user. Leaderboard accessible from admins who name winners./

Tapestry AI Summit Training Module Gameplay

The QA C360 gateway froze after 07-25, so the nightly acquisition job has been writing stale numbers while still reporting success. This cutover points the job at the production endpoint and production client-id.Change: update the client-id in Secrets Manager and register new ECS task-def revisions adding the C360_URL variable. Environment and secret only. No schema change, no frontend deploy.Prerequisites: the production gateway must allowlist our egress IPs 3.93.139.220 and 3.93.243.176. Cutover should happen before Sunday 08-09, after which one week of data becomes unrecoverable.Rollback: re-register the prior task-def revision and restore the old client-id.Dependency: AWS writes require Ajay, as our role is read-only. Full runbook held separately.

Deployed to QA.

C360 endpoint cutover to production

What is happening: Access to the demo application is being cut off for North America and EU store managers. Those users move to the production Coach Pulse application. The demo remains available for Asia, which runs SMC Macau on it.

What to build
<ul>
<li>Gate at login: a user whose market resolves to North America or EU cannot enter the demo application.</li>
<li>Clear message, not a dead end: a blocked user sees a short message pointing them to the production application. No blank screen, no generic "account not set up" error.</li>
<li>Market driven, not store number: scope off the market or region value on the user record so the rule holds for users sitting under a US store record.</li>
<li>Reversible mechanism: preferred implementation is a config driven list of permitted markets that flips back without a code deploy.</li>
<li>Accounts stay intact: access gate only. No account deletion, no removal of existing demo session history or scores.</li>
</ul>

Assumed scope
<ul>
<li>Surfaces: all demo surfaces, meaning coachpulsedemo.com plus vegas, futurenow and gcx subdomains.</li>
<li>Markets retaining access: HK/Macau, Taiwan, Mainland China, Japan, Korea, Singapore, Malaysia, ANZ.</li>
<li>Untouched: the production application, and every Asia market path including the Macau SMC build.</li>
</ul>

Timing: launch is Aug 22. This needs to land before launch so no North America or EU manager is sitting on demo when production goes live. Release window Aug 17 to 21.

Open items
<ul>
<li>Store managers only, or all North America and EU roles including associates.</li>
<li>Whether corporate, admin and FutureProof accounts keep demo access for testing and client demos.</li>
<li>Exact wording of the blocked message, and whether it carries a link to the production application.</li>
<li>Whether EU is defined by the market field or by store country.</li>
<li>Whether the cut off applies to every demo subdomain or only the main demo site.</li>
</ul>

Planned, With Date

Cut off demo application access for North America and EU store managers

Coach Pulse: Globalization &amp; Localization FrameworkToday, Coach Pulse serves one global experience. What we're building is a region-aware architecture where the four core surfaces — Today (Welcome Dashboard), Train (AI Voice Roleplay), Culture Feed (What's New), and Ask Pulse — are each customized per region, so an associate in Tokyo, Riyadh, or Kuala Lumpur opens an experience built for their market, not a US experience translated after the fact.There are three layers to it:1. Regional content buckets. Each region gets its own configuration of the four modules: region-specific Culture Feed content and campaigns, Train personas and scenarios relevant to that market's customers, Ask Pulse knowledge scoped to regional product assortment and selling context, and a Today dashboard reflecting regional priorities and training calendars. Content admins manage each bucket independently through the admin PWA.2. Language selection within region. Region and language are decoupled. Within any regional bucket, associates can switch their display language from a dropdown — so a Malay-speaking associate in a Singapore store and an English-speaking colleague in the same store see the same regional content, each in their own language. The roleplay agent infrastructure already supports this pattern (persona sets exist per language — Korean, French, Arabic, Malay, Indonesian, Thai, etc.), and this extends that model across the full app.3. Automatic regional assignment. No manual setup: on sign-in, the associate's region is derived from their identity — store/location data from the Location Data API and Entra token claims — and they're automatically dropped into the correct regional bucket. Region is assigned by the system; language is chosen by the user.The net effect: one codebase, one platform, many market-correct experiences — with region determined automatically and language personally.4. Region-aware admin. The admin PWA has to mirror the same regional structure so content can be created, translated, and published per bucket — not just consumed per bucket. That breaks down into:<ul><li>Admin scoping by role. The AD group structure already anticipates this: Global Training Admins see and manage all regions; Regional Training Admins are locked to their own bucket. Region assignment on the admin side comes from the same identity plumbing (Entra claims), so a regional admin signs in and lands directly in their region's workspace with no visibility into others.</li><li>Per-region content management. Every content surface in admin — Culture Feed posts, Train persona/scenario assignments, Ask Pulse knowledge scoping, Today dashboard configuration — gets a region dimension. Admins author into a specific bucket, or Global admins can publish globally with regional overrides.</li><li>Language variants at the content level. When an admin publishes a Culture Feed post or dashboard item into a region, they manage language variants of that content (authored or translated), so the associate-side language dropdown has real localized content behind it — not machine-translated UI chrome over English content.</li><li>Regional analytics. Team and training analytics in admin roll up within the regional boundary — a Regional Training Admin sees their region's data only; Global sees cross-region rollups.</li></ul>

Globalization and Localization of Application

Corporate and HQ users whose identity tokens carry no email address and no employee code were dead-ending at login with an "account not set up" message. The resolver now derives the mailbox from the username and adds a gated fallback on the HR id, and the failure screen surfaces the user's own details in all 15 languages.Release dependencies: production is currently on the 07-30 build, so this requires QA promoted to UAT in order to roll production, plus migration 2026-08-03-01 applied to the production database.State: live in dev and QA.

SSO login fix for corporate and HQ users

**What is happening:** The two section labels on the roleplay results screen are being relabeled. This wording is final and supersedes the earlier candidates ("Where to Celebrate", "Where to lean in" in sentence case, "Focus areas", "Wins").

- Strength / Strengths becomes **What to Celebrate**
- Opportunity / Opportunities becomes **Where to Lean In**

Title case exactly as shown, capital I in "In". The new labels have no plural form. The asymmetry is intentional as written: "What" on the first label, "Where" on the second.

**Environments:** Coach Pulse QA and Coach Pulse Demo. Each environment has its own change spec.

**Where it lands**
- Roleplay results screen (Consumer PWA, Train), including the results summary strip, session history rows and the results card preview
- Coach Feedback and session summary copy, including prompt-side wording so generated feedback does not reintroduce the old terms
- Localized strings for every language in the picker. Translations come from each market; Greater China follows the GC AI Localization Guide and lands in Traditional Chinese for HKMO and TW
- My Team "Strengths &amp; Opportunities" label. Code-only in QA where My Team is hidden and locked for launch; user-visible on Demo
- Admin session detail, filter labels and column headers
- Export column headers and reporting dimension display names
- Other demo surfaces, including the session logs pages
- No change to API field names, database columns, scoring rubric criteria or analysis output keys. Display layer only, no backfill and no rescoring
- User guide and training materials

**Out of scope:** scoring and rubric logic, item counts per section, and any other copy change. The in-app disclaimer stays dropped and the user guide carries the explanation.

**Verification:** English results screen, one localized language (must render in language rather than falling back to English), a string sweep of the consumer and admin bundles, admin and export headers, and a pre-change session rendering under the new labels with its scores intact.

**Open items:** fallback behavior while a market translation is outstanding; who signs off each market translation and by when; whether any circulated deck, PDF or guide needs a reissue; release path given the Aug 13-14 blackout, the Aug 17-21 window and the Aug 22 launch.

**Docs:** CoachPulse_QA_ChangeSpec_v2.docx and CoachPulse_Demo_ChangeSpec_v2.docx. v1 of each is superseded by the celebrate-label correction.

**Related, Demo only:** ELX framework scores currently render with decimal points and should render as whole numbers. That change is specified in the Demo doc and is tracked separately from this relabel.

UAT Approved.

Roleplay results relabel: What to Celebrate / Where to Lean In

What is happening: My Team is being hidden and locked in Coach Pulse QA. The section will not be available to users at launch. This is a pause, not a removal. Code, data pipeline and underlying scoring all stay in place; My Team is switched off at the surface only and will be switched back on later. No date is set for that.

<h3>What to build</h3>
<ul>
<li>Navigation: remove the My Team entry from navigation for all roles.</li>
<li>Route: the My Team route should not resolve for a user who reaches it directly. No partial view and no empty state that signals a hidden section.</li>
<li>Code retained: no deletion and no rollback. This is a visibility switch, not a teardown.</li>
<li>Preferred mechanism: a configuration flag that can be flipped back without a code deploy, so re-enablement does not consume a release window.</li>
</ul>

<h3>This is the only change</h3>
Hiding My Team is the entire scope of this ticket. Nothing else in the application moves with it.
<ul>
<li>No other section changes: Today, Culture, Train, Ask and You are untouched.</li>
<li>No copy changes: the Strengths and Opportunities relabel is a separate item and is not bundled here.</li>
<li>No data or scoring changes: collection, scoring and reporting continue exactly as they are.</li>
<li>No cleanup work: no refactor, no deletion, no rollback taken on the back of this change.</li>
</ul>

<h3>Timing</h3>
<ul>
<li>Launch August 22, 2026. Must be hidden before that date.</li>
<li>Release window August 17 to 21. Blackout August 13 to 14.</li>
<li>Re-enablement has no date.</li>
</ul>

<h3>Open items to confirm</h3>
<ul>
<li>Whether the hide applies globally or is market scoped (global assumed).</li>
<li>Whether any admin role retains a view, or the hide is absolute.</li>
<li>The flag name and where it is configured, so re-enablement is a single step.</li>
</ul>

Spec doc: CoachPulse_MyTeam_Hide_Spec_v4.docx

Hide My Team for launch - QA (temporary, code retained)

Surface: Consumer PWA · Culture Feed · video card. Environment: Production. Severity: P1.
What is happening. Culture Feed videos begin playing with audio off. Tapping the volume control turns audio on and the video keeps playing, as intended. Tapping the volume control a second time — the associate’s attempt to turn audio back off — pauses playback rather than re-muting. Mute state and playback state are being changed by the same gesture.
Why it happens. The volume control is nested inside the video card, and the card carries its own tap-to-play/pause handler, so a tap on the volume icon runs both the mute toggle and the playback toggle. The first tap looks correct only because the unmute path calls <code>play()</code> explicitly (browsers block unmuted autoplay, so unmute has to re-kick playback), which immediately undoes the pause the card handler just applied. The mute path has no equivalent <code>play()</code> call, so on the second tap the pause is the only change the associate sees.
Steps to reproduce.
<ol>
<li>Open the Culture Feed in the Consumer PWA and scroll to a video post.</li>
<li>Video autoplays muted. Tap the volume control — audio turns on, playback continues.</li>
<li>Tap the volume control again to turn audio off — the video pauses instead of muting.</li>
</ol>
Proposed fix. Stop the volume control’s event from reaching the card: call <code>stopPropagation()</code> and <code>preventDefault()</code> in the mute toggle handler, and add <code>nativeEvent.stopImmediatePropagation()</code> if the card’s play/pause listener is attached natively rather than through React (React 17+ delegates at the root, so a native ancestor listener fires first and is not stopped by React’s <code>stopPropagation</code> alone). In the same pass, remove the <code>play()</code> call from the unmute path so mute state and playback state are fully independent.
To confirm during the fix.
<ul>
<li>Whether the card’s play/pause listener is a React <code>onClick</code> or a native <code>addEventListener</code> — that determines whether <code>stopImmediatePropagation</code> is also required.</li>
<li>That the volume control is not nested inside a card-level link or anchor wrapper, which would introduce a third handler on the same tap.</li>
<li>Verification on a physical iPhone in the installed PWA — iOS Safari can dispatch pointer and click events separately on some button configurations and reproduce the double-toggle even after propagation is handled.</li>
</ul>
Reference: CoachPulse_CultureFeed_VideoAudio_QA_Findings_v1, finding F1.

Culture Feed video: second tap on the volume control pauses playback instead of re-muting

The manual Add Corporate User form now requires an Employee ID and persists it in normalized form (PR 189, shipped Aug 16 2026). The bulk CSV path at /api/admin/users/corporate/bulk still explicitly rejects an employee_id column and derives keys from the email local part instead. The two paths now behave differently on the same field, which matters ahead of the roughly 8,000-employee bootstrap. Decision needed on whether bulk accepts the column, and if so whether it is required there as it is on the manual form. Source: scoped out of PR 189 deliberately.

Corporate provisioning: accept employee_id on the bulk CSV path

Applies to Coach Pulse identity services and the sync-employees job, with an authoring surface in coach-pulse-admin-ui. Repository attribution to be confirmed before QA.

**Problem.** Coach Pulse runs a two layer identity architecture. SSO through Cognito and AD handles authentication, but the application only trusts the profiles table, which is populated by the sync-employees job reading the Tapestry HR feed. Corporate and test accounts provisioned directly in AD but absent from the HR feed authenticate successfully and then resolve to nothing: blank name, default role display, no store assignment, no C360 data. Corporate, non store manager users have been the group most affected.

**What already exists.** The July 29 identity and access rebuild shipped backfill migrations setting federated aliases for 38 corporate records and keying 50 unkeyed employee IDs, written against dev and QA, so production requires its own backfill pass. The rebuild also made the system fail closed: identities outside approved corporate domains or the HR roster now receive an "account not provisioned" denial rather than a wrong but working session. A break glass provisioning path exists behind QA_IDENTITY_BREAKGLASS and is auditable.

**The gap.** Backfill migrations and break glass are both exception handling. Neither is a supported path. The provisioning gap recurs every time Tapestry stands up a new tester or corporate user outside the roster, and the July diagnosis carried a recommendation that FutureProof build a manual seeding path for non HR accounts. That path has been identified and not yet built.

**Recommended action.** Build a supported, audited admin path to seed and maintain profile rows for corporate and non HR-feed identities, so the fail closed behavior stays intact while legitimate corporate users can be provisioned without a migration or a feature flag.

**Acceptance criteria.**

1. An authorized admin can create a profile row for an identity that authenticates through SSO but does not appear in the HR feed.
2. A seeded profile carries name, role, region and store assignment, and renders identically to an HR sourced profile across Coach Pulse surfaces.
3. Seeded profiles are marked with their provenance so they are distinguishable from HR sourced records in audit.
4. A subsequent sync-employees run does not overwrite or orphan a seeded profile.
5. Every seed, edit and revoke is written to the audit log with actor, timestamp and before and after state.
6. Fail closed behavior is unchanged for identities that are neither seeded nor on the roster.
7. Break glass remains available but is no longer the routine path for corporate provisioning.

**Dependencies and out of scope.** Production backfill pass is a prerequisite and is tracked separately. The open SSO token issue affecting corporate accounts is a separate defect and is not resolved by this work. The compliance question around approximately 4,500 indirect market users accessing outside SSO, with Sarah looping in Michael Pearson, is related but distinct and is out of scope here.

Supported provisioning path for corporate and non-HR-feed accounts

Switch Anthropic API key from FutureProof to Tapestry.