# EdgeBits — SWOT > Honest snapshot of the product, codebase, and go-to-market position as of 2026-04-21. Internal review document. Companion to `issues/`, [internal/market.md](internal/market.md), [internal/master-plan.md](internal/master-plan.md). The pattern across this document: **the work has been architecture-first instead of customer-first.** The platform is thoughtfully designed but missing the connectors, tests, and use-case docs that turn architecture into a sellable product. The single most useful discipline going forward — every engineering decision should pass the filter: *"does this help us ship Modbus + EtherNet/IP + SAP-ingest to a real plant?"* If no, defer it. --- ## 💪 Strengths (what's been built right) The middle of the platform is genuinely strong. These are the parts that took real architectural thinking and it shows. | Strength | Where | Why it matters | |---|---|---| | **Pipeline middle layer** — 15 presets, 6 rule types, 8 actions | [`blocks/presets/`](blocks/presets/), [`blocks/event-processors/`](blocks/event-processors/) | The hard part of an industrial-data platform (composable transforms + rules + actions) is done. | | **Pipeline Engine + Event Engine** | [`edge/pipeline-engine/`](edge/pipeline-engine/), [`edge/event-engine/`](edge/event-engine/) | Clean separation; rule = condition, action = effect, atomic event-processor pattern. Sound primitive. | | **UNS / Envelope / `derive()` / lineage** | [`edge/core/models/envelope.py`](edge/core/models/envelope.py), [`docs/architecture/data-envelope.md`](docs/architecture/data-envelope.md) | Frozen wrapper + immutable lineage chain is the right primitive. Will scale across services without re-derivation cost. | | **DMZ-native, outbound-only architecture** | [`edge/edge-sync/`](edge/edge-sync/), [`docs/architecture/deployment-topologies.md`](docs/architecture/deployment-topologies.md) | Real differentiator — many Indian plants have a buy/no-buy gate on this. Not just marketing. | | **ISA-95 topology + claim-code onboarding** | [`edge-manager/api/`](edge-manager/api/), [internal/master-plan.md:28-31](internal/master-plan.md#L28-L31) | Right shape for industrial fleet onboarding. End-to-end claim flow works (cloud generates → edge claims → heartbeat begins). | | **Edge Sync poller** | [`edge/edge-sync/`](edge/edge-sync/) | Load-bearing component. Without it, the rest of Edge Manager doesn't work. It works (3-step poller, SQLite offline queue, exponential backoff). | | **Catalog + manifest pattern** | [`blocks/`](blocks/) + [`scripts/publish-catalog.py`](scripts/publish-catalog.py) | Right primitive for "we'll keep adding blocks" growth model. JSON registry, manifest-driven UI, schema-rendered config forms. | | **Founding-team industrial-edge experience (since 2017)** | Team history | Real engineering track record across protocols, sites, and verticals. Verifiable through references on demo. Worth more than any "Industry 4.0" slogan. | | **Two SDKs already shipped** | [`sdk/edge/`](sdk/edge/), [`sdk/edge-manager/`](sdk/edge-manager/) | Typed Go clients for cross-service calls. Heartbeat, deploy, pipeline import — all working. | | **Architecture documentation density** | [`docs/architecture/`](docs/architecture/), `architecture-visualizer.html` | Comprehensive system docs + interactive 5-tab visualizer. Onboarding cost for a new engineer is low. | --- ## 🩹 Weaknesses (what's been built poorly or in wrong order) ### Build-order mismatches with strategy | Weakness | Reality | Strategy says | |---|---|---| | **Modbus = C++ stub** | [master-plan.md:38](internal/master-plan.md#L38) marks it 🟡 multi-lang stub | [market.md:55-60](internal/market.md#L55-L60) calls it the "cash cow" — *"70%+ of Indian factory floor equipment runs Modbus."* The single most-leveraged connector is the least mature. | | **EtherNet/IP doesn't exist** | No code in `blocks/connectors/` | [market.md:135-141](internal/market.md#L135-L141): *"~200 Tier-1 companies in Gurugram-Manesar... currently zero data extraction."* The biggest unmet need in the target geography has zero code. | | **SAP integration zero code** | `connectors/sap-ingest` and `egress/sap-rfc` directories don't exist | [market.md:91-97](internal/market.md#L91-L97): *"~40% of ₹500Cr+ Indian manufacturers run SAP. Highest revenue per deal: ₹20-50L/year."* Highest-revenue wedge undelivered. | | **OPC-UA missing** | Not in `blocks/connectors/` | [market.md:62-67](internal/market.md#L62-L67): premium tier for Siemens S7-1500, pharma, auto OEMs. | ### Architectural debt - **Multi-language connector sprawl.** Modbus in C++, MQTT in Go, REST-ingest in Node, S7 in Python. Egress: REST in Node + Go + C++, Sparkplug in Node, AVEVA PI in Python. **No coherent principle** documented for which language goes with which workload. Result: 4× the build/test/deploy surface for 4 connectors. - **Three REST egress implementations** ([`rest-cpp`](blocks/egress/rest-cpp/), [`rest-go`](blocks/egress/rest-go/), [`rest-node`](blocks/egress/rest/)) — only `rest-node` has a manifest. The other two are scaffolds. Either kill duplicates or document the rationale. - **Gateway carries business logic that should live elsewhere.** [master-plan.md:58](internal/master-plan.md#L58) flags this: *"Functions/Actions in gateway. Gateway should be pure proxy. Functions → pipeline-engine, Actions → event-engine."* Known refactor coming. - **Edge Manager on in-memory store.** Anything called *"production-ready fleet management"* with in-memory state can't survive a restart. PostgreSQL migration is on the punch list (now tracked in `issues/`). - **TimescaleDB as the edge buffer.** Heavy SQL engine on a low-end edge node (Raspberry Pi, ARM gateway). Acceptable for industrial PCs; mismatch for the SMB factory segment. - ~~No OpenAPI spec~~ — **incorrect, retracted.** Both APIs have OpenAPI specs: Edge Gateway via FastAPI's auto `/openapi.json` + Swagger UI at `/docs`; Edge Manager via custom `docs.ServeSpec` at `/api/v1/openapi.json` + Swagger UI at `/docs` ([edge-manager/api/cmd/server/main.go](edge-manager/api/cmd/server/main.go), [edge-manager/api/internal/docs/swagger.go](edge-manager/api/internal/docs/swagger.go)). [master-plan.md:60](internal/master-plan.md#L60) is stale on this point. - **TUIs built before connector maturity.** [`edgebits-edge-tui`](edge/tui/) and [`edgebits-manager-tui`](edge-manager/tui/) shipped while Modbus is a stub. Craft project before commercial substance. - **Architecture visualizer (`architecture-visualizer.html`)** — interactive 5-tab HTML built before the product had a customer. Premature self-documentation. - **Brand kit migration** done in parallel with this review. Cosmetic refresh applied to a pre-PMF product. ### Discipline gaps (process / culture) - **No tests anywhere** ([master-plan.md:45, 61](internal/master-plan.md#L45)). Edge Manager API handlers, edge-sync poller, SDK clients all untested. **Largest single risk in the codebase.** Field bugs hit customer 1, not customer 10. - **No CI** that I can see in git history. issues/pending/chore-docs-site.md mentions a CI grep gate as future. Approaching pilots without CI = unsafe to ship. - **Edge UI ships with mock data in some views** ([master-plan.md:42](internal/master-plan.md#L42)). "Ship the demo, not the product." Real button → fake data → trust drops. - **No use-case markdown files** despite [master-plan.md:74-104](internal/master-plan.md#L74-L104) listing 13 of them. The most useful sales/onboarding asset, missing. Every demo reinvents the explanation. - **Documentation on niche MkDocs theme** (`mkdocs-shadcn`) hitting limits ([issues/pending/chore-docs-site.md:21](issues/pending/chore-docs-site.md#L21)). Migration cost looming. - **Recent commit messages are 1-2 words** (*"docs org," "updates," "cleanup"*) — either no code-review culture exists or it happens off-repo. Hard to onboard new engineers; hard to audit decisions later. ### Scope sprawl - **Three products in parallel** (Edge Node, Edge Manager, Analytics) before any one has paying customers. Post-PMF shape applied pre-PMF. - **Five active roadmap plans** (AI, SAP, Digital Twin, Alerting, Product Family). Each plan creates implicit promises and reading load. - **Six target verticals** in market.md (auto, pharma, steel, FMCG, textile, power) — each has different protocols, compliance, sales motions. PMF needs **one vertical narrowly won**, not five hedged. - **Edge Manager built with 12 UI pages.** Day-1 customer flow needs ~4 (claim node, see health, deploy pipeline, view events). 8 pages are scope ahead of need. - **Per-node-monthly website pricing** ($120/$240) doesn't match the team's actual commercial model in [market.md:59](internal/market.md#L59) (annual per-plant). Website built without reading strategy doc. --- ## 🚀 Opportunities (market hands these to us) | Opportunity | Why it's leverageable now | |---|---| | **BEE/PAT compliance wedge** | Mandatory for ₹500Cr+ plants. Hard deadline = budget. Modbus + REST + AVEVA PI cover the technical scope; team can ship on existing-block inventory. | | **North-India process plant geography** | Faridabad, Ludhiana, Manesar, Panipat — ~1000 candidate plants in 4 cities. Drivable from Gurugram for support. | | **Energy savings ROI** | Energy is 30-40% of cost in steel / textile / foundry. Hard ROI numbers close CFO meetings. | | **AVEVA PI install base in process plants** | EdgeBits already ships the egress. PI-owning plants (pharma, oil & gas, steel) get a coverage story today. | | **Ignition compatibility via Sparkplug B** | Free coverage. Already shipped. Just needs documentation, not a build. | | **Maruti supplier ecosystem (~200 Tier-1 plants)** | EtherNet/IP unlocks the entire wedge. Single connector build = one of the largest addressable cohorts in the country. | | **Make-in-India / NMP / MII government schemes** | Create budget windows, procurement preference, "Indian-vendor" credentials. Untouched in current marketing. | | **DMZ-native is a buy/no-buy criterion** | Most cloud-SaaS competitors can't satisfy this. Not a feature — a differentiator. | | **Hindi / regional-language UI** | International vendors don't ship localised operator UI. Genuine differentiation in tier-2 deployments. | | **SI / OEM channel** | L&T Tech Services, Cognizant, TCS Industrial, Schneider India, Yokogawa — partner-led scale path. None engaged today. | | **Founding-team industrial-edge experience** | Opens enterprise conversations. Verifiable through references on demo. Should be the heritage anchor in every commercial conversation. | | **SAP integration as enterprise wedge** | [market.md:97](internal/market.md#L97): *"₹20-50L/year. Currently solved by ₹1-3 Cr SI projects."* Disruptive economics if delivered. | | **AI/Twin/MCP narrative for thoughtful buyers** | Reasonable defer to 2027 differentiator. Don't build speculatively, but the roadmap is real and demonstrable. | | **OT cybersecurity tailwind** | IEC 62443, CERT-In, increasing data residency mandates — create budget for the kind of platform EdgeBits ships. | --- ## ⚠️ Threats (what could kill or compress us) ### Competitive - **Established historians + SCADA platforms have install base.** Wonderware, Ignition, AVEVA System Platform are entrenched. Head-on replacement is a losing fight; positioning as **augmentation** is the only viable wedge. - **AVEVA's acquisition of OSIsoft (PI)** consolidated the historian + SCADA market. Cross-protocol + historian integration play (which is EdgeBits's strength) faces a vertically-integrated competitor. - **Cloud-native edge platforms** (Azure Arc, AWS IoT Greengrass, Google Edge) have hyperscaler distribution and infinite engineering. The differentiation window for an independent industrial edge platform compresses every year. - **Indian-domestic competitors** ([Altizon](https://altizon.com/) — Pune, [Hiotron](https://www.hiotron.com/) — Pune *FactoryMetrics*) target the same segments: OEE, predictive maintenance, asset health, auto / pharma / heavy manufacturing. The founding team's industrial-edge experience gives EdgeBits an edge but doesn't make the field empty. (Earlier draft listed FlytBase — drone autonomy, wrong segment; Falkonry — California-based, acquired by IFS; OnGo IIoT — unverifiable. All removed after a verification pass.) ### Commercial / market - **Indian factory CFO preference for CapEx + AMC over SaaS subscription.** The website pricing model is misaligned. If the first 10 sales conversations stall on this, that's a real pipeline-conversion problem. - **Direct-sales scale wall at ~20 plants.** Without an SI/OEM channel built early, growth flatlines. No partner enablement story exists today. - **Pre-revenue + multi-product roadmap = burn-rate risk.** Engineering attention split across Edge Node, Edge Manager, Analytics, AI, Twin, Alerting plans = no single product reaches ship-grade fast. - **The first 3-5 customers are the only thing that matters now.** Anything not directly contributing to those wins is a threat to runway. ### Technical / operational - **No tests = field bugs at customer 1.** Single largest technical threat. A bad first deployment kills word-of-mouth in a tightly-clustered industrial geography. - **Modbus stub status if exposed in a sales conversation.** *"You support Modbus?"* → *"It's a C++ stub"* → conversation over. Risk grows the more the website implies Modbus is shipping. - **Marketing claims eroding credibility if challenged.** *"10+ protocols," "100+ nodes," fictional case studies* — first sales engineer who's grilled by a skeptical IT head loses the deal *and* the reputation. Cleanup pass underway in issues/pending/chore-website-copy-honesty.md. - **Multi-language connector sprawl burns engineering hours on duplication.** Three REST-egress implementations is a real waste of cycles that should be on Modbus production-grade or EtherNet/IP greenfield. ### Strategic temptations - **AI / Twin / MCP roadmap expansion creates promises with no engineering capacity to fulfill.** Every plan in `docs/roadmap/` is a soft commitment to prospects who read it. Five plans = five disappointment vectors. - **Open-source temptation could repeat an OSS-monetisation mistake the team has already lived through.** It is why we're building EdgeBits as a commercial product rather than another OSS giveaway. Don't relive it. - **Indian buyer skepticism of AI hype.** Marketing that leads with "AI-powered" loses OT credibility. Reasonable to keep AI on the roadmap; not reasonable to lead with it. - **Architecture-craft temptation.** TUIs, brand kits, design systems, interactive visualizers — these are post-PMF activities. Each hour spent here is an hour not spent on Modbus / EtherNet/IP / SAP. - **Government regulation changes** (data residency, IEC 62443 mandates) could require unplanned rework. Plan for this; don't be surprised by it. --- ## 🎯 What this SWOT implies (priorities) If only 3 things change in the next quarter: 1. **Ship Modbus production-grade.** The cash-cow connector being a C++ stub is the single biggest gap between strategy and reality. One language, one team, one deadline. 2. **Add tests + CI to Edge Manager + edge-sync.** Not coverage-driven; failure-mode-driven. The first 3 customers will rest on this. 3. **Pick one vertical and ship Day-1 wedge.** *Energy + PAT compliance for North-India process plants* (per the analysis in the prior review). Stop building Twin / MCP / Analytics speculatively until that wedge has 5 paying plants. If only 3 *not* to do: 1. **Don't build EtherNet/IP, OPC-UA, SAP-ingest in parallel.** Sequence them. Modbus first; one of the three next; the other two after first deal closed. 2. **Don't add a sixth roadmap plan.** Five is already too many for a pre-revenue product. 3. **Don't ship more cosmetic / craft work** (TUI extensions, brand-kit refinements, architecture visualizer features) until customer-facing reliability gaps are closed. If only 3 to invest in operationally: 1. **Use-case markdown files** (the 13 named in master-plan, currently 0 written). Sales asset, onboarding asset, SE-enablement asset. 2. **SI / OEM partner enablement** — even a one-page "How to deploy EdgeBits at a customer site as an SI" doc would unblock channel conversations. 3. **Hindi operator UI** (and one regional language). International vendors don't ship this; it's a $1L engineering investment that buys differentiation in tier-2 deployments forever. --- This document is honest by design. It's the read for someone deciding what the next quarter looks like — not a sales artifact and not a product brochure. It pairs with [internal/market.md](internal/market.md) (which says *"go here, sell this"*) and [internal/master-plan.md](internal/master-plan.md) (which says *"this is what's built"*). The gap between the two is what this SWOT is documenting.