Files
OKF_Vault/knowledge/product/foundation-pitch.md
T
Noel Sabia 1dc33853fb
release-please / release-please (push) Failing after 3s
feat: added pictures and stuff
2026-11-15 14:15:24 +01:00

13 KiB
Raw Blame History

type, title, description, tags, status, generated, sources
type title description tags status generated sources
Product Strategy PulseCast as a foundation: integration-first use cases Platform thesis, pre-jury feedback, and integration-first use-case candidates for the Schwarz Digits Pulse of the City challenge.
hackathon
future-city-hackathon-2026
schwarz-digits
pulsecast
product-strategy
integration
platform
retail-media
waste
city
draft
by at
opencode/deepseek-v4.1-flash 2026-11-14T14:30:00Z
id title author
user-brief User brief, 14 Nov 2026 team:pulsecast
id resource title
current-product ../../../COOL_CITY.md Cool City product brief

At a glance

Field Value
Thesis Build the foundation — data analysis pipeline + ML forecast + API — and spin up many decision layers on top.
Integration rule Combine things that already exist. Do not build another silo; plug into existing systems.
Current state The pipeline, forecast, Go API, Nuxt/Onyx frontend, SSE, SKE deployment, k6 evidence and the Cool City reference integration are live. New applications are cheap to deploy.
Pre-jury signal The use cases were received as “okay”; the biggest plus was that the solution is deployed on STACKIT. This is user-reported, not a transcript.
Next move Sell the engine + STACKIT deployment as the product; add a private-sector Schwarz anchor with a zero-friction public interface, and use PreZero only as the plug-and-play proof.

The pitch

We built the foundation; you spin up the decisions.

The foundation is a data-analysis pipeline and a weather-aware machine-learning forecast over Heilbronn’s eight inner-city PAX counters. It is already deployed on SKE with a Go API, a Nuxt/Onyx frontend, live SSE, cached fallback and load-test evidence. The forecast is a relative device index, not a count of people; every integration must carry that caveat.

The product is not one dashboard. It is the demand signal layer that existing systems can consume:

  • a retail media planner schedules a campaign;
  • a store production system plans bakery batches;
  • a waste route planner adds a stop;
  • a city dashboard opens a cooling station or reroutes an event;
  • a workforce system fills a shift;
  • a loyalty app sends a quiet-hour offer.

The same API, the same forecast, different decisions. The value is the foundation and the connectors, not another app.

New constraints from the user — 14 November 2026

  1. STACKIT deployment is the strongest signal. The pre-jury liked that the solution is already deployed on STACKIT. Keep the production deployment and the engine as the headline.
  2. The wow must be zero-friction. A person with nothing on them should be able to experience it plug-and-play. Workable interpretation: no phone, no app, no account, no login, no setup. The interface is a public screen or kiosk that is already there. If “nothing on them” means no device at all, the public screen is the only answer; a browser-only experience is the fallback.
  3. The third asset is private sector, for Schwarz — not another community app. The community reference can stay, but the third asset must be a Schwarz business use case.
  4. PreZero is allowed only as proof of plug-and-play. The engine is the main selling point. The PreZero dataset is a test that the same engine can feed an existing waste-routing system, not the hero use case.
  5. “Barely legal” is treated as audacious, not unlawful. No cameras, no MAC addresses, no personal data, no deception. Aggregate counts only, anonymous interaction, public screen, DPIA before production. The edge is the public, ambient, zero-friction interface — not surveillance.

The three use cases

The user narrowed the product to three concrete decisions, one per surface:

  1. Staffing — “How many employees do I need?” A private-sector workforce decision. The current /plan is the reference connector.
  2. Cool route — “Which path in summer avoids sunlight and crowds, fastest from A to B?” A public/community decision. The current /cool is the live reference.
  3. PreZero — “Which containers or routes need service first?” A private-sector waste decision. This may be mocked from the real fill-level dataset; the engine is the product. The data has no coordinates, so the honest output is a service-priority proof, not a fake route.

The three are not three apps. They are three connectors on one engine.

The challenge insight we must make explicit

“The true value of this challenge isn't just crunching numbers; it's proving that teams can take raw, privacy-preserving sensor data, fuse it with real-world context, and turn it into an intuitive visual story. Adding a predictive model is the ultimate cherry on top that turns hindsight into foresight.”

The site and the pitch must show each clause:

Clause Where it appears
Raw, privacy-preserving sensor data The privacy ladder: sniff → hash → aggregate → privacy-first.
Fuse with real-world context Weather, holidays, events, and the live scenario controls.
Intuitive visual story The three short, sharp use-case previews, updated together over SSE.
Predictive model as the cherry on top The 7-day forecast and the “play with the engine” simulator.

Current gap

The pieces exist, but they do not feel fused. The frontend still reads as a set of separate demos rather than one official product. The next design pass must:

  • use the Onyx (Schwarz IT) light theme, not the current beige/black editorial look;
  • be mobile-first;
  • lead with “we provide the engine, you provide the system it works on”;
  • show the three use cases as one live system, controlled by SSE;
  • end with a frontend playground where a visitor can change place, weather, event and time and see a simulated decision;
  • keep the PreZero proof clearly labeled as mocked from the real fill-level dataset;
  • keep the privacy and deployment evidence visible but compact.

The first Stitch preview confirmed the gap: it was still a dashboard. It packed proof chips, a pipeline breadcrumb, floating chart annotations, a live telemetry bar, three metric tables, three mini-dashboards, a privacy trade-off panel and a compliance chip row into one page. The correction is Apple’s methodology 1:1: clarity, deference, depth; one idea per screen; one primary action; large type; generous whitespace; progressive disclosure. If a detail does not help the current idea, remove it. The site should feel almost empty compared with the preview.

Why the previous use cases were “okay, not great”

The private shift plan and the community heat app are both plausible, but each was pitched as a standalone product. Hypotheses, to validate with the challenge provider:

  1. They looked like new apps, not integrations. The judge has to imagine adopting another dashboard.
  2. The buyer was unclear. A store manager, a city department and a public app user are different customers with different budgets.
  3. The benefit was plausible but not urgent. “Better shifts” and “cooler routes” are good, but not a must-have next quarter.
  4. The platform story was hidden. The same forecast can feed many systems; the pitch led with one use case.
  5. The data caveat dominated. The index is not people, so absolute claims are hard; the product needs a decision that works with a relative signal.

The fix is not a new model. It is a clearer buyer, an existing system to plug into, and a demo that shows one foundation feeding several decisions.

Integration-first candidates

# Integration Existing system it plugs into What the foundation adds Buyer Demo feasibility
1 Schwarz Retail Media Pulse Screen Existing DOOH screen, store-window display or kiosk Live city pulse + best time/route; the screen adapts its message to the predicted audience; the person touches the screen, no phone needed Schwarz retail media / Kaufland or Lidl store operations High — a tablet/laptop in kiosk mode can demo the experience
2 Retail media / DOOH scheduling Retail media or digital-signage planner Predicted audience by place and hour; best time/screen to run a campaign Schwarz Digits / retail media High — one planner screen, forecast-driven slots
3 Fresh food / bakery production Store production or replenishment system Predicted footfall by hour; recommended bake batches; waste vs availability Kaufland / Lidl store operations Medium — needs transparent production assumptions
4 PreZero dynamic waste routing Existing route planner and fill-level data Predicted crowd hotspots and event peaks; add or move a collection PreZero Medium–high — combine footfall with the available fill-level dataset
5 City operations / heat & events City dashboard, heat map, event control room Predicted crowd/heat pressure; where to open a cooling station or add staff City of Heilbronn High — Cool City already exists as the reference
6 Loyalty / quiet-hour offers Lidl Plus / Kaufland Card or a city app Predicted quiet windows; send an opt-in offer when the store is calm Retail CRM / loyalty team Medium — privacy and spam risk; mock only
7 Workforce management connector Existing WFM system Predicted demand signal; the WFM system fills shifts Retail operations High — current /plan is the reference connector
8 Store location / real estate GIS or site-selection tool Predicted footfall for vacant sites and trade areas Schwarz Real Estate / city Medium — map and score, less live
9 Delivery / logistics Last-mile routing or cargo-bike dispatch Predicted pedestrian density; choose a quieter delivery window/route Logistics operator Medium — mock route, less direct

Recommended shape

The engine + STACKIT is the product. The connectors prove it. The public screen is the wow.

  1. The engine on STACKIT — the foundation and the main selling point. Data pipeline, ML forecast, Go API, Nuxt/Onyx, SSE, cached fallback, k6 evidence, deployed. The buyer is not adopting an app; the buyer is adopting a demand-signal layer they can deploy on their own STACKIT infrastructure.
  2. Schwarz Retail Media Pulse Screen — the private-sector anchor and the zero-friction wow. An existing DOOH screen, store-window display or kiosk shows the live city pulse and a best-time/route recommendation. A passer-by with no phone, no app and no account touches the screen and gets an answer. The screen is an existing retail-media asset; the engine makes it context-aware. This is the “barely legal” edge without surveillance: aggregate counts only, anonymous interaction, public screen.
  3. PreZero dynamic waste — the plug-and-play proof, not the hero. The same engine plus the PreZero fill-level dataset feeds an existing route planner and adds or moves a collection after a high-footfall event. It proves the engine can plug into another Schwarz Group system.
  4. Cool City — the public-good reference. Already live. It shows the same engine serving the city and the public without becoming the whole product.
  5. Workforce / bakery — optional commercial connectors if time allows. Do not lead with the shift plan again.

The demo should show the same forecast API response driving the Pulse Screen and the PreZero proof, with Cool City as the live public reference. That is the platform proof: one engine, several existing systems, no new silo.

Answered and remaining questions

Answered by the user on 14 Nov 2026:

  • The pre-jury’s biggest plus was the STACKIT deployment; the use cases were “okay.”
  • The third asset must be private sector, for Schwarz, not another community app.
  • The wow must be zero-friction: a person with nothing on them can experience it plug-and-play.
  • PreZero is allowed only as a plug-and-play proof; the engine is the main selling point.

Still open:

  • Which Schwarz business unit owns the Pulse Screen: retail media, Kaufland/Lidl store operations, or Schwarz Digits? This changes the pitch language and the buyer.
  • Which physical screen or kiosk can we use at the venue? A tablet or laptop in kiosk mode is enough; a real DOOH screen would be stronger but is not required.
  • Does “nothing on them” mean no phone, or no device at all? The current interpretation is no phone/app/account, with the public screen provided. If it means no device at all, the public screen is the only answer.
  • What is the smallest credible PreZero proof? A single high-footfall zone + one fill-level hotspot + one recommended route change is enough; a full router is not.
  • Do we keep Cool City as the live public reference, or hide it from the 5-minute preselection and use it only in Q&A?

Links