Skip to content

Events, artists and bookings

Page Purpose
/admin/events Event list; new and [id]/edit for setup
/admin/events/[id]/booths Booth layout and allocation
/admin/events/[id]/timetable Schedule
/admin/events/[id]/gallery Event gallery
/admin/artists Artist roster
/admin/signing-config Per-artist signing configuration
/admin/bundles Ticket / package bundles
/admin/artworks Artwork pieces and orders
/admin/sponsors Sponsor listings
/admin/stamp-rally Rallies and vendor directory
/admin/auctions Auctions, with [id]/bids and [id]/winners

Signing configuration is what makes slots bookable: time_slots.slot_kind must be signing, since createHold refuses every other kind. The signing-queue skill covers the QR, marker, language and orientation metadata that hangs off a booking item.

Artist deletion is refused when any package or slot has booking history (including cancelled/refunded items and bundle items without a solo slot), or when an event bundle composition references the artist or their packages. The checks, catalog parent locks, and all deletions run in one transaction; a later constraint failure rolls back cart and hold cleanup too.

Page Purpose
/admin/bookings All bookings
/admin/bookings/[id] One booking: items, payments, status changes
/admin/bookings/signing Signing bookings
/admin/bookings/new-order Admin-only multi-line unpaid signing order and Stripe payment link
/admin/bookings/entry Entry ticket bookings
/admin/grading-bookings Grading add-on bookings

Payment deferral is an owner-approved concession that lets an unpaid booking be signed, scanned and slip-printed while the pending amount stays truthful — the payment_deferred_* columns on bookings are the record of it.

At /admin/bookings/new-order?event=<event-id>, choose an event and enter the customer’s email (required), name and phone. Existing customer details can be loaded from the email match. Add bundle or package lines, choose their slots, and complete the same personalisation fields used for walk-in bookings.

Add every bundle (first open slot) adds quantity 1 for each active bundle with available capacity. When its earliest open slot has only one or two seats, the picker prefers the earliest slot with at least three seats on that same Malaysia calendar day; it never moves to another day just for more seats. Bundles without an open slot are listed as skipped. Required dedications and choose-bundle artist selections still need completing.

Review the subtotal, 4% platform fee and total. Choose a 24-, 48- (default), or 72-hour payment deadline, then create the order. All lines belong to one booking with one Stripe Checkout link. The result links to the booking and shows its total and deadline in MYT. Copy the URL or use Email link to customer; the email includes every line, meeting times, the fee-inclusive total and a Pay now button. Preview it at /admin/email-previews → Order ready to pay.

Refresh link reuses an open matching Checkout or regenerates an expired Checkout; it does not extend the order’s payment deadline. Emailing also checks the link before sending. An email failure leaves the order/link intact and reports the failed delivery; sends and failed attempts are audited.

Cancel order requires confirmation, expires live payment links first, then cancels all signing lines through the existing cancellation procedures to release capacity. A payment in progress prevents cancellation. If cancellation fails partway through, inspect the linked booking before retrying.

In booking details, Outstanding balance now collects the live pending signing balance with settle_outstanding. The server supplies the read-only amount and validates it again when generating the link; this is not the paid-cash correction effect.

Page Covers
/admin/revenue/signing Artist signing revenue
/admin/revenue/booking Booking revenue
/admin/revenue/whatsapp WhatsApp-originated sales
/admin/analytics Cross-event analytics

Routers: dashboard, reports, analytics, admin-bookings.

/admin/sales-update generates a WhatsApp text update for entry tickets or signings, with team and permission-gated stakeholder audiences.

The event defaults to the live event or next upcoming event; an unknown ?event= ID falls back to that default. Live/Upcoming labels and the default refresh every minute.

For Entry tickets, the WhatsApp album compares the selected event with Card Mania 2026 (default) or another event chosen in Compare event. The selected event is excluded from the comparison options. A URL whose compare ID equals the selected event is treated as the default comparison. The calm Convention 2026 booklet-style album contains six 1080 × 1350 PNGs for stakeholders, or four for the team: yellow rounded frames, cream pages, navy headings and date tiles, without rays or halftone.

  1. Scoreboard — paid passes sold, the comparison by this day, days to doors, today’s sales so far, and the biggest observed presale day.
  2. Forecast — most likely final passes (P50, rounded to the nearest 100), the P10–P90 range and reference final, chance of beating the reference, and cumulative actuals plus the forecast fan from T-14 through the last event day.
  3. Spikes & tiers — grouped daily bars for 14 complete presale days plus today, the three biggest current-event days in that window, and up to four paid-unit tier rows. Extra tiers are combined into Other; missing tiers count as zero.
  4. Fans from… N countries — both audiences: how many countries buyers come from, the international share (with the reference event’s share at the same moment), and every country with its flag, buyer count and share (Malaysia included). Country = phone calling code, else the Stripe card country; unknown and walk-in buyers are left out and counted in the footnote. It is not nationality or residence. For marketing claims, GA4 (where people browsed from) can list different countries. Below the list, +N more countries visiting the site shows flags for countries with 5+ engaged GA4 visits to the event pages (last 120 days) but no buyer yet. It is network location, so VPNs count; the big number stays buyer countries only.
  5. Attendance outlook — stakeholders only: likely people (rounded to 100), a 40-bin simulated-attendee histogram, dashed Low/Likely/High markers and an orange reference actual-attendance line. Stickers show the chance of at least 5,000, 10,000 and 20,000 people, capped at 99%+ / <1%.
  6. Day by day — stakeholders only: every calendar day from the earlier first paid-sale offset through today, including zero-sale and event days. Columns show the TCGKL date, daily paid online + desk passes, change from the previous day’s count (signed arrows, not colour alone), reference full-day paid passes at the same offset, and cumulative TCGKL paid passes. Today is so far; its reference column remains a full day. More than 16 rows use two groups. Beyond 32 rows, the first date/cumulative anchor plus the latest 31 days fit without reducing the 30px type or resetting totals.

Countries come from buyer phone country codes, not nationality or residence. International share is known non-Malaysia buyers divided by all known-country buyers; Unknown is excluded from that percentage but included in the donut’s total. No usable phone and walk-ins without accounts are Unknown, never assumed Malaysian. Account buyers are deduplicated per event; accountless walk-ins are counted per booking, so this is not a count of unique attendees. Only paid, confirmed, non-complimentary entry items contribute. Current buyers are all-time; reference buyers are cut at the aligned MYT instant without the matched on-sale lower bound. Below five known-country buyers, the card shows Not enough data yet. No names or phone numbers appear.

Previews stack on mobile, use two columns at md and four at 2xl. Card text is at least 30px in the PNG; the internal-only timestamp footer is 28px. Each card includes its n/6 (stakeholder) or n/4 (team) position.

Tap Share to WhatsApp, then select WhatsApp in the phone’s share sheet to send the complete audience-specific album together. The share title is <event name> · Entry update, using the selected event’s name. PNGs are prepared before the tap to preserve mobile sharing permissions. If file sharing is unavailable or fails, all six/four PNGs download instead; cancelling the share sheet does not download. Attach the downloaded cards together in WhatsApp.

The share also carries the WhatsApp text update shown in the preview, so the cards and the text arrive together. Some WhatsApp versions drop the text when several images are shared, so the text is copied to the clipboard as well: paste it into the chat if it is missing. Both come from the same live data, but they are fetched separately, so a sale landing in between can make them differ by a pass or two.

Download 6 PNGs (stakeholder) / Download 4 PNGs (team) downloads the complete album and reports the actual number downloaded. The browser may ask you to allow multiple downloads. Each card’s Copy copies that PNG, falling back to a download with a notice if clipboard access is unavailable. Attendance and day-table filenames end in -5-attendance.png and -6-days.png. Exports stay at 1080 × 1350, below WhatsApp’s 1600px resizing threshold.

Regenerate refreshes the album with live comparison data and simulations; this fetch is logged. If PNG preparation fails, Retry images rasterises the current album again without fetching or logging another report. Team = units only, never RM. Permission-gated stakeholder cards also show collected RM (no decimals). Projected RM (a Low–High range with the likely value) appears only after the first door sale, because the door price mix is unknown before then.

Events align by MYT calendar days to doors: T-7 means seven days before each event; D1 is opening day. The scoreboard’s “× Card Mania” compares both events over the same days on sale (from whichever event started selling later) and up to the same MYT clock time today, so it doesn’t jump at midnight or favour the event that went on sale earlier. “Today so far” and the hatched so far bar are partial; on door days the reference figure next to them is also cut at the same clock time.

The forecast projects the rest of today from the reference’s share of that day still to come at the same clock time, so door-morning albums don’t drop the day’s remaining sales. Likely is the median of 5,000 simulation runs; Low–High holds 8 in 10 runs. Same data gives the same numbers (the simulation is seeded from the data, not the time). Pace uses 7 complete days, excluding today. Event-day demand starts from a damped pace multiplier (elasticity 0.5) with door uncertainty (σ 0.4). After each complete door day, it blends in our own door sales vs the reference’s on the same days (half weight after D1, 80% from D2), so a slow or strong opening updates the rest of the event. After the event, card 2 shows the final count instead of a forecast. This is based on one reference event: earlier on-sale dates or early-bird pricing can pull door buyers into presales, so a presale lead need not carry fully into door sales.

Attendance uses reference admitted passes / issued passes, including comps: non-revoked entry_ticket_passes rows joined to the reference event. A pass with scanned_at set counts once as admitted, regardless of repeat scans; this is neither a scan count nor the booking line’s quantity. Card Mania’s legacy reference is 3,730 / 4,404 = 84.7% (4,182 paid + about 222 comps). The payload returns reference.attendance (issued, admitted, showUpRate) and reference.compsFinal.

In each of the same 5,000 seeded forecast runs, final comps are current comps plus reference event-day comps times that run’s event-day scale. Attendees are round((paid final + final comps) × show-up), with show-up sampled from Beta(admitted + 1, issued − admitted + 1). The attendance random substream preserves the paid-pass draws. forecast.attendees supplies P10/P50/P90, 40 bins whose counts sum to the run count, and the three threshold probabilities. This remains a one-reference-event assumption, not a guarantee or a unique visitor count. No issued reference passes means a null show-up rate and no attendee simulation; card 6 explains the missing reference. After the event, there is no attendee forecast.

With no reference (or if the reference is the selected event itself), the scoreboard comparison is hidden and the forecast card explains that a reference event must be set. If charts are disabled, the album is replaced by an Unable to generate charts: … error box containing the server’s flag-off message.

/admin/vouchers sits in the Engagement group and settles against bookings, so discount reconciliation questions usually start there. /admin/lucky-draws and /admin/merch-purchases sit in the Events group and do not touch bookings: they are about on-site merch receipts and stage draws. Both pages need manage_lucky_draws, which admin and payment_staff hold.

/admin/lucky-draws is one table of every draw, newest first, with live draws sorted to the top. The status strip doubles as a status filter (with counts). Draws can be searched by draw or event name and filtered by event (?event=all|none|<id>); all filter state lives in the URL. Each row has a next action: Stage, Set up or Results. Stage opens in a new tab.

Create Draw asks for a name, a description, how people enter, an event and (for merch-purchase draws) the minimum purchase. The event is optional except for merch-purchase draws.

/admin/lucky-draws/[id] has:

  • a Next step panel, with a progress stepper and a readiness checklist that mirrors the server’s open/activate checks; the main action stays disabled until every item is done
  • a stats strip: prizes, entrants, spins, awarded
  • prizes, spins and entries (searchable via ?q=, voidable) in the main column
  • a sticky rules rail for the event, secret code and T&C

While the draw is a draft, each prize row has Edit. It opens the add-prize form inline, prefilled, and saves through upsertPrize. Ranks are unique per draw. Prizes, T&C, code and event lock once the draw leaves draft.

How the draw works underneath (entry rules, spin order, retry safety, the 72-name wheel) is in Lucky draw flow.

A draw enters people one of two ways:

  • Secret code. Attendees open /lucky-draw/[id] (QR on the stage screen), sign in, type the code the MC reveals, and tick the T&C. One entry per person, phone required, staff accounts blocked.
  • Merch purchase. Activate snapshots receipts ≥ the minimum into entries, one per receipt.

Event-day flow:

  1. Set up (admin, before the event). On /admin/lucky-draws/[id], add prizes (rank 1 = grand prize, spun last), the T&C (include photo consent) and the code. All three lock once the draw opens.
  2. Open entries about 30 minutes before the draw, then Close entries before the first spin.
  3. Open stage (/admin/lucky-draws/[id]/stage, opens in a new tab) on the projector laptop, on a phone hotspot rather than venue Wi-Fi. Admin or payment staff can run it.
  4. Spin. The server picks the winner and the wheel lands on them. Ask for the last 4 phone digits shown on the card, then tap Present. If nobody comes up, tap Absent: that person is out and the same prize is spun again. A re-sent spin returns the same winner; it never picks twice.
  5. Complete draw. Any prize left without a winner is recorded as unawarded.
  6. Proof photos. Afterwards, upload them on the draw page (admin accounts; the upload route does not admit payment staff). They appear on the public results; use Hide if a winner asks.

Where attendees find it: any draw that’s live, or completed within the last 30 days, is listed at /lucky-draw, appears under Play in the site header, and shows as a promo card on its event’s page (if it’s linked to an event). Drafts never show publicly.

Every open, close, spin, outcome, void and photo change is audit-logged. Design and decisions: docs/lucky-draw/implementation.md.

/admin/merch-purchases records on-site merch receipts. These receipts are what merch-purchase lucky draws snapshot into entries.

  • Toolbar. Event picker and search (customer name, email or receipt reference), both in the URL (?event=, ?q=). With no ?event=, the first event is preselected.
  • Stats strip for the selected event, including an “Eligible (≥ RM100)” count.
  • Rows show the customer, amount, lucky-draw eligibility chip, receipt reference and recorded time. Record purchase (admin, payment staff) adds one. Delete is admin-only and labelled per row.
  • The list loads only the latest 100 purchases for the event and says so when it is capped. Search filters those 100 in the browser.

Deleting a purchase is refused once it has been entered into a completed draw, or once one of its entries has won.