Quartz 4

platform-strategy

15 min read

Platform Strategy

Strategic frame for the Dungeon Books platform work. Source: extended strategy session 2026-05-01, building on platform-survey-2026-05-01. This doc supersedes prior implicit framings; update in place when the frame shifts.

Positioning Reframes

The platform is not a “bookstore OS” or generic “indie retail OS.” It is the operating layer for gaming-and-nerd-culture third-spaces. Evidence converging from multiple angles:

  • Pilot pipeline is gaming-adjacent (Victory Point Cafe, Pixel Labs likely overlap)
  • Brand language is cassette futurism + 80s sword & sorcery + RPG character sheet
  • Carrie’s pitch: “real-world MMORPG with your shop as a quest hub” (see outsiderpg-platform-vision)
  • Discord-as-identity is a subculture filter — natural fit for game stores, comic shops, RPG-leaning bookstores, board game cafés, hackerspaces; awkward fit for record stores, plant shops, cafés-without-gaming
  • Indie RPG/zine ingest thesis is the wedge against Ingram/Hardcover incumbents — same subculture (see 2026-04-23)

Addressable market under this framing: ~8–15k US venues. ~5k game stores (GAMA), ~2.5–4k indie bookstores (RPG/SFF/comics-leaning subset), ~1.2k comic shops, plus board game cafés, escape rooms, LGS event venues, hackerspaces. Comparable scale to Mindbody’s fitness studio TAM at founding.

Stop using “any third space” in pitch. Narrow the messaging. Brand is the moat in community-driven verticals. Embrace the subculture rather than positioning as culture-neutral. This supersedes the broader framing in jc-canvassing-plan target profile.

Pitch

Guild is built for bookstores, game stores, comic shops, and other geek-adjacent rooms whose regulars already use your Discord. It turns those regulars into a membership: they earn XP for showing up and buying things, work through quests you set, and redeem points for things you actually stock. When a new shipment lands or Free RPG Day is coming up, write a quest for it and your members see it the next time they open the app.

The pitch loads three things: (a) subculture filter (bookstores, game stores, comic shops, geek-adjacent rooms self-identify in or out); (b) Discord prerequisite (if your community already lives there, the auth chain is one-click and the loyalty surface is plug-and-play); (c) hyper-local ads as a verb the owner does on a Tuesday afternoon, not a feature in a list. The shop owner is the curator. See discord-orbs-quests for the design lineage.

Hyper-local ads as a Guild surface

The “Quests” primitive (guild-portal-ui-implementation Quest Log, guild-quest-hub-network partner network) is also the ad surface:

  • Self-promo quests: “buy a Mörk Borg supplement this month → 2× XP,” “attend Free RPG Day → 500 pts + a buff.”
  • Cross-shop quests (network play, post-multi-tenant): “visit Victory Point this week → 200 pts redeemable at either shop.”
  • Partner / publisher quests (curator-gated, not self-serve): “pre-order James Islington’s next book before launch → exclusive Mithril-tier ARC slot.”
  • Discovery quests (Marty-personalized): “you’ve been on a grimdark kick — try this Le Guin.” Same engine, just tuned to the member’s purchase history.

What separates this from Discord’s model: the shop owner is the curator, not a sales team. No self-serve advertiser market — that’s how you get spam. Quests are authored by Carrie / Panat / Phil for their own communities, and at network scale by the org running the partner network. Discord-quality design ethos, indie-shop-quality curation.

Product Surface — Platform vs Applications

Distinguish in all future communication:

Platform (the offering sold to partner shops):

Applications (run on the platform; demo material, not the offering):

  • marty — SMS/Discord book chatbot, Dungeon Books-specific persona
  • Manapool integration — MTG marketplace, reference integration showing platform composes with vertical marketplaces
  • Hi.Events integration — events vertical
  • Future: F&B, ticketing, etc.

Integration discipline: only build vertical integrations a paying pilot shop is actively asking for. Never on spec. Hi.Events shipped because Carrie uses it. Manapool shipped because Panat sells MTG. Hold the line.

Pitch frame for partner shops: “Plug in the verticals your shop cares about — events, MTG, indie RPG/zine, F&B. We handle the integration substrate.”

PMF Status

12 launch members at Dungeon Books, 2 months post-launch:

  • 50% engaged with multi-weekly check-ins

  • Points earned across cohort
  • 1 tier upgrade (Bronze → Silver)
  • Zero churn
  • Discord + IRL engagement
  • Top regulars converted to paid Guild members
  • Withfriends fully sunset 2026-05-01

This is real PMF signal at single-shop scale. Not migration inertia. Working loyalty product for top regulars.

This is NOT platform-level signal. That requires partner shop validation. Phil at Victory Point is the next test.

Threshold: 8 of Phil’s top 15 regulars converting to paid Guild within 60 days with Dungeon-comparable engagement = platform thesis validates. <50% of that = Dungeon-specific product, different decision tree.

Caveat: emporium not yet launched. Don’t onboard Phil before Dungeon’s own storefront is stable. Partner shop will judge platform partly by what’s running at Dungeon.

Tenant Boundary Decision

Pilot shops onboard as additional Orgs under Script Wizards LLC’s Stripe and infrastructure. Use existing multi-tenant plugin (guild-multitenancy). Do not build separate deployments per pilot — multiplies operational burden by N, forecloses cross-merchant identity layer architecture, wastes integration work.

Migration trigger (when pilot shops graduate to own infrastructure): when partner-shop Stripe revenue materially exceeds Dungeon’s own revenue, when a partner shop wants own Stripe for tax reasons, or when capital raise requires entity cleanup. Not now.

Brian buyout question is upstream of entity structure. Don’t pre-decide entity for platform infrastructure. Use Org-under-Script-Wizards-Stripe model and revisit when Brian’s situation resolves.

Cross-Merchant Identity Layer — Parked

Confirmed parked until 3–5 partner shops are live. Reasons:

  • Cannot validate network thesis without multiple shops
  • Architectural decisions will be wrong without real cross-shop usage data
  • Indie ingest tool is the wedge that makes platform indispensable to partner shops without requiring network effects
  • 2027 build, not 2026 build, unless a partner shop explicitly requests it

Network geography heuristic: When eventually built, network value is highest when pilot shops share a customer base — same subculture, geographically clustered. JC + Hoboken + Manhattan gaming-adjacent third-spaces is a tighter network than 3 random NJ indie retailers. Plan pilot geography accordingly. Supersedes geography-agnostic framing in cross-merchant-identity and guild-quest-hub-network.

Platform Engineering Priorities — Updated Order

Supersedes the priority order in roadmap.

  1. Emporium launch close-out. PRs #10–13, real Open Library covers, related products on blog posts, sitemap, branded 404, Payload-driven footer. Blocks everything else partner-facing.
  2. Indie ingest tool Phase 1. GTIN-less long-tail catalog wedge. Owned by no incumbent. Differentiator vs Bookmanager/Edelweiss/Square. See 2026-04-23.
  3. Domain routing for multi-tenant. Highest-leverage unbuilt platform piece. Phil cannot pilot meaningfully on dungeon.club/victory-point. Unblocks pilot onboarding. See guild-multitenancy “Domain Routing (planned)” section.
  4. First partner shop onboarding (Phil at Victory Point most likely).
  5. Observability hardening. Cross-tenant Grafana/Datadog before second pilot, not after. Pi-crash-class incidents (see guild-kiosk-v2) at partner shops destroy trust because their staff lack operational context.
  6. Auto-discount automation at Square POS. Manual breaks at ~40 members. Solve before 100 members across 3 shops.
  7. Eval infrastructure for Marty + indie ingest tool. Currently nothing in vault. Required before AI outputs ship to partner shop customers.
  8. Cross-merchant identity layer. 2027.

Auth Chain — Subculture Dependency Acknowledged

Discord-as-identity via Better Auth + Authentik is a subculture filter, not a bug. Decision: accept it as the platform’s beachhead-vertical assumption. Gaming/nerd-culture third-spaces have Discord-native customer bases. Record stores, plant shops, and non-gaming cafés are out of scope for the beachhead vertical, can be revisited post-PMF if ever. See discord-integration.

Discord as local-community substrate

The Discord bet is sharper than “our customers use Discord.” Discord has become the de facto local-community substrate for the under-40 cohort in dense urban areas. Jersey City Socials, Victory Point’s server, Dungeon Books’ server, Hudson County Tabletop, smaller TTRPG-specific servers all run in parallel; members move between them.

Empirical check, 2026-05-01. Of Dungeon Books’ 12 launch members, 9 are on Discord (75%). Of those 9:

  • 5 of 9 (~56%) are already in Victory Point’s server.
  • 2 of 9 (~22%) are in Jersey City Socials.
  • 1 of 9 is in Hudson County Tabletop.
  • 1 of 9 is in a smaller TTRPG-specific server.

The 56% VP overlap is the load-bearing number. The cross-shop substrate thesis with Phil as first pilot is validated at the resolution that matters: the customer overlap already exists on Discord; Guild’s job is to overlay loyalty data on it, not to create the network.

JCS at 22% has a different role. Wide-and-shallow. Reads as a member-acquisition channel (sponsored welcome quest in JCS to drive signups at participating shops), not the cross-merchant graph itself.

Substrate fragility caveat. Hudson County Tabletop is currently rotting (spam, bots, low engagement, weak admin custodianship; Panat has left). It was the obvious “local tabletop server” before, and the platform thesis was leaning on it as exemplar. It’s no longer that. The lesson: a Discord server is only substrate if its admin is a good custodian. Not every metro server is healthy. Server-quality due-diligence belongs in the metro-density rubric.

Three substrate layers, not one.

  • Broadcast servers (large, metro-scale, low curation): JC Socials at thousands of members. Role = member-acquisition channel. Value of a sponsored welcome quest scales with membership; value as a shared social graph is low because the graph is too sparse.
  • External curator-run servers (small, focused, high curation): TTRPG HQ at ~70 members, JC-local, 18+, questionnaire-gated, mission-statement-on-entry, system-diversity over D&D-only. Role = an external cross-shop social graph if a relationship exists. Cold approach is not the move; partnership is opportunistic when warm channels open.
  • In-shop DM-led groups (first-party, smallest unit, highest leverage): the DMs running games at Dungeon Books and the equivalent at Phil’s. Each DM is a curator with a player group on a recurring schedule, in the shop’s space. Role = the substrate the platform actually owns. No cold outreach, no permission, existing loyalty.

The metro-density rubric should count layers 2 and 3 together, not just public Discord servers. JC currently has Victory Point’s server, Dungeon Books’ server, and TTRPG HQ on the external-curator side; JCS on the broadcast side; and the DMs running games inside DB and VP on the first-party side.

DMs as super nodes. Gaming-and-nerd-culture retail is the only third-space vertical where paid or volunteer curators run scheduled programming inside the shop. The DM layer is a structural property of the beachhead vertical and part of why the platform doesn’t generalize beyond it without losing the moat.

In network terms, a typical member is a low-degree node connecting to 1–2 others. A DM running a recurring table is a high-degree node connecting to 4–6 players directly, and those players connect to each other through the DM. Without the DM, the cluster doesn’t exist as a stable subgraph. That gives DMs four compounding effects:

  • Acquisition multiplier. One DM = 4–6 prospect slots already filtered for taste, system interest, and willingness to show up on a recurring schedule.
  • Retention coupling. A player who stops coming because the DM stopped running is structural churn. Healthy DM = healthy table.
  • Cross-merchant edge embodied. A DM running at DB on Tuesdays and VP on Thursdays is the cross-shop social graph in one person. Concrete super-node example for cross-merchant-identity before any schema work ships.
  • Information broker. DMs know who’s looking for players, who runs what, who just lost a campaign. Local TTRPG market-intelligence layer.

This connects directly to research-lab: super nodes are where leverage concentrates in network-coordination dynamics. Mechanism design around DMs is the difference between a working coalition and a churning marketplace.

Super-node mechanism design — anti-extractive. Don’t pay per session run or rank by session count; that attracts mercenaries and dilutes the taste filter. Design moves:

  • Recognition surface, not paycheck. DM badge on the character sheet, shoutout in activity feed, “DM’d by X” credit on quests their players completed.
  • Tools that reduce their cognitive load. Session sign-up, attendance ledger, table-level XP awards, player-at-a-glance.
  • Player-DM matching as a Quest type. “Looking for 4 for Mausritter, monthly at DB.” Platform routes interested members. Solves a real DM problem.
  • Cross-DM visibility within a shop. DMs see who else runs what. Peer group, not isolated freelancers.
  • Cross-shop DM mobility. Once domain routing and multi-org tier work, a DM running away sessions at a partner shop gets recognition at both. The cross-merchant edge compounds in practice.
  • Comped membership, not paid-per-head. DM-tier comp or venue priority as a relationship benefit, not piece-rate.

Risks.

  • Single-DM dependency. The shop’s economics shouldn’t ride on one DM. Spread the table portfolio. New-DM recruitment funnel visible in the platform.
  • Relationship custody. The shop owner has the DM relationship; the platform is infrastructure. Don’t insert the platform between the shop owner and their super nodes (the WithFriends/Kunal-as-bottleneck failure mode, see sales-lessons-from-withfriends).

The 25% non-Discord cohort is non-trivial. Email/phone-as-identity stays load-bearing and can’t be deprecated, even at the beachhead. Discord is the moat for the cohort that uses it; a fallback identity path remains required for the rest.

What this means for the platform:

  • The beachhead’s addressable market is larger than it looks. Asking “do your regulars use your Discord?” filters in the right shops by construction. Shops that say no aren’t in the beachhead anyway.
  • The cross-merchant identity thesis has a real substrate. Cross-shop social graphs already exist on Discord. The work in cross-merchant-identity is overlaying purchase, visit, and XP data on a network Discord maintains, not building one from zero.
  • Geography heuristic is operational. “Pilot shops should be geographically clustered and subculture-shared” is testable: are the candidate shops in each other’s servers, or in the same metro-area servers? Are those servers actively curated? VP overlap ≥50% would be a strong inclusion signal for any new pilot.

Next pilot geography. JC first (Phil at VP), Brooklyn / NYC second. Same Discord-density logic applies. Metro density rubric (game-store count, healthy local TTRPG/SFF servers, indie bookstore density) belongs in a follow-on analysis.

Discord’s own Orbs and Quests launches are a design study, not a competitive threat. Discord can’t pay out in real-world value, can’t curate at the shop level, can’t network across independent businesses, and can’t put quests in physical space. The transferable insight is the design ethos — quests as discrete actions with explicit rewards, ambient dismissible UI, dual-currency pricing, account-bound to kill resale markets, curator-gated quest pipeline. Lift the patterns into Guild’s IRL + small-business context where Discord won’t go.

Tier/Brand Configuration — Decided (Hybrid)

Hybrid model. Tier names (bronze/silver/gold/mithril + iron non-member) are a platform-global hardcoded enum — part of the cassette-futurism + RPG voice and subculture moat, not negotiable per shop. Tier config (price, Stripe price ID, loyalty multiplier, benefit table) is per-shop in the Payload Tiers collection.

Dungeon Books is the reference implementation (current: Bronze 20, Gold 60). Reconciled in membership-platform Tiers section and guild-multitenancy Key Decisions.

Pync Decision — Reframed

Pync is salary and visa. Stop rationalizing it as strategically aligned with platform thesis. Pet tech hardware has zero subculture transfer to gaming/nerd-culture third-spaces, zero customer base overlap, zero distribution channel overlap.

Honest framing: Take Pync (or equivalent fractional role) for the three things that matter — visa sponsorship, capital to fund platform work, low cognitive load preserving evenings/weekends for distribution work and partner pilots. Stop trying to make it strategically aligned.

Hard constraints to negotiate into Pync agreement:

  • Written weekly hour cap
  • Pre-decided response policy for off-hours CEO requests (default: no, with explicit exception conditions)
  • IP arrangement already done (Script Wizards / Dungeon Books carved out via PIIA)

3–6 month revisit point must be a real decision, not procedural. Most fractional CTOs drift to full-time without forcing function. Set quantitative milestone for platform — see criterion section.

Criterion for Platform Go/No-Go

“Vibes” is acceptable for next 6 months only. Pick one concrete criterion, write it down, share with Carrie:

Option A — Revenue threshold: Platform B2B MRR (partner shop subscriptions, excluding Dungeon’s own retail) crosses $5k MRR sustained 3 months. ~3–5 partner shops at current pricing (see platform-pricing). Achievable 12–18 months if Phil + 2 JC leads convert.

Option B — Pilot conversion rate: Onboard 3 pilot shops. 2 of 3 still active and paying at 6 months with Dungeon-comparable engagement metrics = platform generalizes, commit. 0 or 1 of 3 = Dungeon-specific product, different path.

Option C — External capital signal: Someone who understands vertical SaaS or community-driven retail commits at terms that allow full-time. Sahil/Gumroad-aesthetic angel is the implicit model. If Sahil is plausible, the forcing question is: what would move him from “interesting” to writing a check?

Pick one. Doesn’t have to be perfect. Has to exist.

Items Requiring Vault Update (Backport Queue)

All resolved 2026-05-01 except where noted.

  • Withfriends migration completed 2026-05-01 → noted in 2026-05-01 journal “Also today” section.
  • Guild activity snapshot → guild-activity-2026-05-01.
  • Tier-naming reconciliation → hybrid, see membership-platform Tiers and guild-multitenancy Key Decisions.
  • Discord-as-identity rationale → “Why Discord (subculture filter, not bug)” section in discord-integration.
  • Platform vs Applications taxonomy → projects and index reorganized.
  • Facebook group of Square bookstore owners + Edelweiss VP relationship → already captured in dave-lucey: runs the Square booksellers Facebook group (10 years), drove Omnibus creation as Edelweiss Analytics “store 0,” standing up an Omnibus Advisory Council with Panat invited.

Open Questions Surfaced, Not Answered

  • Brand/tier configurability decision (per-shop vs global vs hybrid)
  • Sahil approach — real conversation or hypothetical reference point only
  • Pilot geography sequencing (JC density vs cross-region)
  • Brian buyout resolution and entity structure implication

Graph View

Backlinks