Investor Q&A

AI is a pre-tagging layer only — it is not the final source of truth for any commercial element. Visual analysis (Gemini 2.5 Flash), song fingerprinting (ACRCloud), and speech transcription (ElevenLabs Scribe v2) identify candidates, which then enter a human review queue before going live. No commercial tag (product, brand, location) is published without a confidence threshold and a human approval step.

More importantly: the self-funding curation model means that the humans most motivated to correct errors are the ones who profit from accuracy. A retailer who has placed their jacket at the wrong timestamp will see zero clicks and fix it immediately. The AI gets the graph to 60–70% accuracy; commercial incentives get it to 95%+. This is the same mechanism Wikipedia uses — motivated contributors correct errors because they care about the outcome.

For elements with no commercial partner (background props, obscure brands), AI tags are labelled as "community-verified" rather than "confirmed" — managing user expectation at the UX layer.

Timeline scrubbing does not trigger live graph queries. Scene-level metadata is serialised to a static JSON payload at indexing time and cached in Redis — the timeline UI consumes this pre-built payload, which loads once and renders locally. The live Neo4j query only fires when a user clicks on a specific element (e.g., tapping the jacket icon to open a product panel). This keeps the graph query count per session low and latency imperceptible.

For popular films with high concurrent traffic, the JSON payload is served from CDN edge nodes. Neo4j's read performance at the element-level (a single node and its direct relationships) is sub-10ms at the dataset sizes we operate in Phase 1–3.

All third-party integrations are wrapped in a service adapter layer with circuit breakers and graceful degradation. If Spotify's affiliate API goes down, the music panel shows a "Listen" link without an affiliate deep link — the discovery value remains, the revenue temporarily pauses. If a hotel booking affiliate is unavailable, the location CTA falls back to a direct Google Maps link.

At Phase 1, only 5 integrations are required: Gemini 2.5 Flash (visual analysis), ACRCloud (song fingerprinting), ElevenLabs Scribe v2 (transcription), one affiliate network (Awin or CJ), and CineCore. The integration surface grows with phases, and a dedicated integration reliability engineer is budgeted from Month 6. No single integration failure disables the core product.

Scenery does not host, reproduce, or stream film footage. It indexes metadata — structured descriptions of elements that appear in films, associated with timestamps. This is legally equivalent to what IMDb, Wikipedia, and every published film guide has done for decades without legal challenge. Creating a database entry that says "a scorpion jacket appears at 00:14:32 in Drive (2011)" is no more a derivative work than a critic writing about it.

The legal exposure increases if Scenery were to display film frames, clips, or stills without licensing — which it does not do. Product imagery comes from retailer catalogues (licensed by the retailer). Location photography is sourced independently.

For Phase 4 (US/Hollywood entry), we will engage IP counsel to structure the studio approach proactively. At that point, the platform will have demonstrated significant audience traffic and affiliate revenue — creating a data-backed negotiating position. This is the same playbook Spotify used with major labels: build the audience first, negotiate the licence second.

This is a real complexity at Phase 4+ (Hollywood catalogue). The indie-first strategy is a deliberate mitigation: indie filmmakers typically hold consolidated rights and are contractually incentivised to grant Scene Graph indexing in exchange for revenue share through CineCore. A single filmmaker can sign off on the entire commercial layer for their film.

For commerce, the model is opt-in by layer — each commercial element is independent. A brand's product can be tagged in a scene without requiring the studio's involvement, if the product metadata is provided by the retailer. Music affiliates run on existing platform agreements (Spotify's Open Access programme). Incompleteness is a valid state — a scene without brand clearance simply shows no product CTA. The discovery experience is intact; only the commercial layer is absent.

As traffic grows, the multi-party permission problem inverts: studios, labels, and brands approach Scenery to be included rather than Scenery seeking their approval.

Scenery operates as a referral and affiliate platform across all three real estate sub-streams — not as a landlord, property manager, or estate agent. Short-term rental CTAs link to licensed operators (Airbnb, Booking.com, Vrbo) who carry all guest liability and insurance obligations. Sale referrals pass a referral identifier to licensed estate agents who manage the transaction. Filming location hire is facilitated through approved location agencies who hold their own professional indemnity insurance.

Scenery's contractual relationship is with the referring platform, not with the tenant, buyer, or production company. This is the same model used by every hotel metasearch and comparison site. The regulatory and insurance exposure sits with the licensed operator, not the referrer.

No. And this is a feature, not a limitation. Scenery is a post-film platform. The primary use case is: you watched Drive last week, you keep thinking about that jacket, and today you want to find it. Right now, that search happens on Google across dozens of fragmented, low-quality results. Scenery centralises the answer.

The proof that this demand exists is already in the market. Sites like Movie-locations.com and boutiques like Cité des Nuages in Waterloo (Belgium) survive entirely on post-film discovery behaviour — visitors who want to book a filming location or buy a prop months after watching a film. They each solve one fragment of the demand. Scenery centralises all of it — products, locations, music, recipes — in a single explorable graph.

This also aligns with Scenery's cultural positioning. The platform doesn't interrupt the viewing experience. It extends it — and in doing so, it responds to a genuine felt need rather than creating an artificial one.

Yes — and we've restructured accordingly. Phase 1 is deliberately two streams: Shop Scene Placement and Streaming Referral via CineCore. That is sufficient to validate commercial intent. Eight streams in a pitch dilutes the message and raises legitimate questions about focus and IP risk. We've documented the full model as a growth roadmap, not a Day 1 plan.

The two active streams are complementary and self-reinforcing. Placement validates that users click through from cultural discovery to commerce. Streaming referral validates that discovery drives watch intent. If both perform, the entire commercial thesis is proven and the path to activating subsequent streams becomes a sequencing question, not a product question.

The other streams — IP licensing, music affiliate, brand analytics, hospitality, tours, and the data API — are each documented, modelled, and ready to activate as the graph reaches the density they require. They represent upside, not promises.

Year 3 profitability rests on two structural advantages. First, marginal cost is near zero for each additional film indexed — once the AI pipeline and curation infrastructure is built, adding films costs primarily compute and editorial time, not headcount. Second, the self-funding curation model means the primary data enrichment activity (tagging scene elements) is performed by commercial partners at their own cost.

The Year 3 cost structure (€1.49M) reflects 7 FTE, AI pipeline costs, and marketing — not a large operations team. Scenery is fundamentally a data and marketplace platform, not a content business. It does not produce films, host video, or manage inventory. The platform scales revenue with very limited operational scaling required.

We model Year 3 conservatively: 650,000 MAU on a platform that IMDb reaches 200M+ users. The market is large; our assumption is that we need to capture a fraction of it to hit these numbers.

Data is the strategic core of Scenery, and ownership is the critical distinction. Scenery owns and controls the Scene Graph. Shops, brands, and venues enrich it — but they do not own it. This is exactly the Google Maps model: businesses can appear on the map without owning the map. That structural distinction is what makes the data asset defensible and, eventually, licensable.

The platform generates three distinct layers of data:

• The Scene Graph itself — structured metadata of every cultural element in every indexed film. This is the primary, proprietary asset.
• User behaviour — what users look at, click on, share, and return to. This reveals which cultural elements drive genuine interest versus passive browsing.
• Commercial performance — what converts, in which scene context, for which cultural reason. This is data no brand can get anywhere else.

Monetisation of this data follows a three-phase sequence. First, indirect: the graph powers placement revenue (shops pay for access to it). Second, direct: a brand analytics dashboard lets companies see how their products perform culturally — €499–€1,999/month per brand. Third, at scale: a paid API for streaming platforms, studios, and researchers who want structured access to the graph. Institutional licence fees in this category typically run €50K–€500K/year per buyer.

Data licensing is excluded from seed-round projections to keep the financial model conservative. It represents the highest-margin upside for Series A investors — and the strongest long-term argument for why the graph, once built to sufficient density, becomes difficult to replicate at any price.

The long-term vision is to become the visual culture search engine for cinema. That is a fundamentally different and much larger ambition than a shopping site.

No search engine today can answer the queries that serious cinephiles actually have: "Films with romantic locations in southern Europe." "Films shot in Bruges." "Films with an 80s neon aesthetic." "Films where the protagonist wears Comme des Garçons." These are real, high-intent queries. They get fragmented, low-quality results today. With a sufficiently rich Scene Graph, Scenery becomes the only platform that can answer them precisely.

This framing is also a stronger commercial thesis. Cultural queries naturally generate lists, rankings, and editorial content — exactly the kind of content that drives organic SEO and social sharing. A query answered well becomes content. Content drives traffic. Traffic deepens the graph. The graph answers more queries. This is the flywheel.

Commerce — the jacket, the location booking, the film rental — is the natural consequence of that discovery. It is not the primary proposition. You're not selling jackets. You're building the search engine for visual culture. The commerce follows because users who have genuinely discovered something want to act on it.

A streaming platform can only build this for its own catalogue. Netflix cannot make a Drive scene shoppable if Drive is on Amazon Prime. Apple TV+ cannot index a Letterboxd favourite that streams on MUBI. The value of the Scene Graph is precisely its cross-platform, cross-studio, cross-distributor nature — it is the only place where a user can explore any film regardless of where it streams.

There is also a structural conflict of interest: Netflix building a shoppable layer for competitor films would accelerate users migrating to those other platforms. The incentive to build a neutral cross-platform graph does not exist inside any streaming service. It can only exist at an independent layer.

Finally: any streaming service attempting to build this would face the same multi-year data accumulation problem that makes the graph a moat. They would start from zero while Scenery has years of commercially-enriched curation already in place.

Which leads to the more interesting question: streaming platforms don't become competitors — they become customers. See Q14.

Yes — and this is one of the most important long-term opportunities. In Phase 3, Scenery plans to offer the Scene Graph as a licensed B2B API that streaming platforms embed directly into their interfaces. A "What's in this scene?" button inside Netflix, powered by Scenery's graph, surfacing the jacket, the location, the song — without the user ever leaving Netflix.

The logic for the platform is compelling: they enhance their product with a feature users clearly want, without spending 3–5 years building the underlying data infrastructure. The logic for Scenery is equally compelling: we get distribution inside the world's largest viewing interfaces while earning annual licence fees and per-query revenue.

This model has a clear analogue: streaming platforms don't build their own payment infrastructure — they use Stripe. They don't build their own CDN — they use Cloudflare or Akamai. Scenery becomes the scene intelligence layer, the same way other platforms are the payments layer or the infrastructure layer. The cross-platform nature of the graph is precisely what makes it valuable to each individual platform — they get intelligence about films they don't own or exclusively host.

The natural first-mover targets are regional streamers — Canal+, Viaplay, MUBI, RTL+ — which have the commercial incentive and organisational agility to move faster than the global giants. Once the API is proven at that tier, Tier 1 (Netflix, Prime, Disney+) becomes a realistic conversation.

The moat has three compounding layers:

• Data accumulation: Every film indexed, every element tagged, every commercial relationship established makes the dataset progressively harder and more expensive to replicate. A competitor starting today faces a multi-year catch-up against a graph that is growing faster as more partners join.
• Network effects: Retailers who have built their placement strategy around Scenery's timestamp system do not switch platforms. Tour operators who have built tours linked to the location graph do not rebuild them elsewhere. The commercial relationships compound the data lock-in.
• Self-funding curation: No well-funded competitor can replicate the mechanism by which Scenery's data is enriched for free by motivated commercial actors. A competitor would need to fund their own curation manually — expensive and slow — while Scenery's curation cost trends toward zero.

Three things converged recently that make this buildable now and not five years ago:

• Multimodal AI at viable cost: Gemini 2.5 Flash handles visual analysis at a fraction of what purpose-built CV APIs cost two years ago. ACRCloud identifies commercial music tracks. ElevenLabs Scribe v2 transcribes dialogue with speaker attribution. Running the full pipeline across a film catalogue is now economically viable without institutional funding — see the AI Pipeline cost calculator for exact figures.
• Streaming fragmentation creating discovery demand: With content distributed across 10+ platforms, users actively need a neutral layer that tells them what's worth watching and where. The discovery gap is larger than it has ever been.
• The commerce infrastructure is mature: Affiliate networks, deep-link APIs, IP licensing frameworks, and real-time inventory APIs all exist and are accessible. The "shop the scene" concept required all of these to be simultaneously available at scale.

Shazam created a new user behaviour (identify a song in real time) when mobile hardware made it possible. Scenery creates a new user behaviour (explore a film as a cultural map) at the moment when streaming density, AI capability, and commerce infrastructure have all matured simultaneously.

Four strategic reasons. First, rights simplicity: European indie filmmakers hold consolidated rights and are actively seeking new revenue streams — getting a film onto Scenery is a single conversation with one person. In the US, even a mid-budget indie can have rights fragmented across a studio, a distributor, a music licensor, and a foreign sales agent.

Second, uncontested beachhead: IMDb, Letterboxd, and JustWatch were all built in English for English-speaking audiences. European cinephile communities — which are among the most engaged film audiences in the world — are underserved by discovery platforms in their languages and cultural contexts.

Third, regulatory alignment: GDPR compliance and data-handling expectations are highest in Europe. Building for European compliance first means the platform is compliance-ready for every other market by default.

Fourth, festival infrastructure: Cannes, Berlinale, IFFR, and Ghent provide direct access to the indie filmmaker community for onboarding. This GTM channel doesn't exist in the same form in the US.

Three concrete mechanisms, each independently valuable:

• Timestamp sharing: Every scene has a unique URL. "Watch the exact second the scorpion jacket first appears in Drive." That is native TikTok and Reels content — the "did you know" film format performs extremely well organically without a paid budget. The link goes directly to Scenery. Paid ads come later, to amplify what's already working — not to create demand that isn't there.
• Existing cinephile communities: Letterboxd, Reddit film communities, and YouTube channels built on "every outfit in..." or "all filming locations in..." formats are natural seeding partners. They already produce the demand; Scenery gives them a better destination to point it at.
• Filmmakers as distribution: When an indie filmmaker shares their own Scenery page at release — "my film is now fully explorable on Scenery" — they're marketing to their existing audience for free. That is earned media Scenery doesn't pay for, reaching the exact audience most likely to engage deeply.

There is also a fourth mechanism specific to the visual culture search angle: cultural queries generate lists, rankings, and editorial content. "Best films shot in Brussels." "Most iconic fashion moments in cinema." These lists are highly shareable, highly SEO-indexed, and produced at zero cost from the graph. The content layer is not something Scenery needs to manufacture — it emerges from the data.

Phase 5 (2030) is Series B funded and assumes dedicated regional teams in both markets. The Scene Graph architecture is language-agnostic — node metadata is stored with locale variants. The product localisation (UI, search, editorial) is a known engineering problem with known costs. The cultural nuance challenge is primarily commercial: identifying the right retail and hospitality partners in each market.

For Korea, the Hallyu fanbase is already globally distributed and English-comfortable — the initial Korean Wave audience can be served in English with Korean-language support following. For Japan, the existing film location tourism infrastructure (Japan Film Commission, established location-travel routes) provides a commercial partner network to tap from day one.

Yes — and it is a Phase 3 play, not Phase 1. The concept is a QR code displayed in the cinema immediately after a trailer. The viewer has just been shown a film, in a darkened room, with full attention. That is the highest-intent moment in the entire discovery funnel. The QR links directly to the film's Scene Graph: merchandise pre-orders, the filming location, the soundtrack on Spotify — all accessible before they stand up.

This is not a new behaviour to teach. The cinema industry already uses QR codes in lobbies and on popcorn cups. The innovation is connecting that moment — peak enthusiasm, phone already in hand — to a structured commercial layer that didn't previously exist.

The model works for all three parties. Studios pay for placement (it is effectively a performance marketing channel with measurable conversion). Cinema chains (Kinepolis, Pathé, Vue) offer a richer, more valuable post-trailer experience to their audiences. Scenery acquires a high-intent user who has just self-selected as interested in that specific film — the most qualified possible audience.

The Kate Bush / Stranger Things data point is the clearest precedent: a streaming moment created a 9,600% surge in streams overnight. The behaviour — "I just saw something, I need it now" — is already proven. Scenery is the infrastructure that catches that moment in a physical cinema context, at the exact second it happens.

This channel requires catalogue depth and a working commerce layer to be valuable — which is why it is a Phase 3 initiative, not a launch feature.

The seed buys 24 months of runway for a 6-person core team (5 engineers + 1 designer), AI pipeline infrastructure, curation for an initial film catalogue, festival-based filmmaker outreach, paid community growth, and expanded legal groundwork for retailer agreements and IP. Full allocation is documented in the Business Plan.

The core principle: 50 perfectly documented films are worth more than 5,000 empty pages. A Drive page that is flawless — every product identified, every location GPS-linked, every scene shareable — creates more trust than superficial coverage of thousands of titles. Month 12 is about depth and quality, not catalogue size.

Month 12 success looks like:

• 50 films with exceptional Scene Graph coverage — every element tagged, verified, and commercially linked
• At least one formal partnership with a recognised film festival (IFFR Rotterdam or Ghent Film Festival) — positioning Scenery as cultural infrastructure, not e-commerce
• 30+ retail shops actively placing products at timestamps, with documented conversion data
• 10,000+ monthly active users returning at >3 sessions per month
• €46K ARR across two active revenue streams (placement + streaming referral)
• One high-profile "scene commerce moment" — a product sells out or a location gets booked out directly traceable to Scenery discovery

Month 12 is not about revenue. It is about proving the behaviour: that users explore a film as a cultural territory, and that commercial intent follows naturally — without the platform ever feeling like a shop.