The line physically cannot build it wrong, and every unit ships with proof. Re-run from scratch and re-earned against the strongest alternatives.
The June pack proposed Verified Assembly. A mid-June review sharpened the banner to Enforced Assembly. This V4 round then started again from zero: fresh research, a new method, and the prior answers treated as candidates to beat, never as anchors. The banner survived. Everything around it got stronger. The June pack stays linked above as the working archive.
ODIN's demonstrations win. The failure sits earlier: getting the benefit across before anyone reaches a demo stand. Today that gap is filled with eleven different self-descriptions across the sales materials, and a quarter of the planned raise is earmarked for physical demo centres. A category exists to close that gap: the right frame does the demo's persuasive work in words and numbers, so the demo confirms a decision the buyer has already made.
This round was built on Obviously Awesome and Sales Pitch, April Dunford's two books on positioning, both read in full and codified into our method library. Her discipline is strict about order: you never start with the product or the name. You start with what buyers would use if you did not exist, and the category comes last, chosen to make the strengths obvious.
Dunford's rule decided the round: position yourself in a frame that makes your strengths obvious to the people who care most. On MES criteria ODIN loses. On guidance criteria ODIN is invisible. On enforcement criteria ODIN is alone.
| Axis (5 = best for ODIN) | Enforced assembly | Unit-governed assembly | First-time-right assembly | Software-defined assembly | Productised line controller |
|---|---|---|---|---|---|
| Buyer-pain matchHow directly the name speaks to a pain the buyer feels every shift. | 5 | 4 | 5 | 3 | 4 |
| Engineering credibilityWhether a controls engineer accepts the claim against first principles. | 5 | 5 | 4 | 4 | 5 |
| Ownability in 24 monthsWhether ODIN can own the term before someone else plants a flag. | 4 | 5 | 2 | 3 | 3 |
| Language whitespaceHow uncontested the territory is in mid-2026, against live vendor marketing. | 4 | 5 | 2 | 3 | 3 |
| Channel transmissionHow cleanly the name travels through integrators, OEMs and partners. | 4 | 3 | 5 | 4 | 2 |
| Incumbent defence costHow expensive the claim is for Siemens, Rockwell, SAP or Tulip to copy. | 5 | 4 | 2 | 2 | 4 |
| Adjacent possibleWhether the frame carries the AI era without over-reaching today. | 4 | 4 | 3 | 5 | 3 |
| No-demo transmissibilityBuilt for this engagement: does the name alone convey the benefit with no product shown. | 5 | 4 | 5 | 3 | 3 |
| Total | 36 | 34 | 28 | 27 | 27 |
Discrete manufacturing where the correct build is physically enforced at every step, and every unit ships with proof. One data model holds the line's sequence, rules and settings. People, tools, machines and cameras can only execute what the model allows. Because every action is checked as it happens, the per-unit record is born complete, as a by-product of running the line.
A process engineer changes products, sequences and rules in minutes, in no-code, with a full audit trail. No PLC programmer in the loop.
The tool will not fire out of position or sequence. The station locks without its conditions. Advice is not authority.
Every result is born with its product, position, target, tool and operator. The unit's genealogy writes itself.
When control is lost, the line locks. It never runs blind.
These three criteria are the category's teaching payload. A buyer can qualify any vendor before a single meeting: can your line stop a wrong build, can your engineers change it without code, can you produce any unit's full record in minutes. Guidance platforms fail the first. MES fails the second and third at unit grain. The stitched stack fails all three at the seams.
ISA-95 gives every manufacturing buyer the same five-layer map: devices, PLCs, SCADA, MES, ERP. It works for process manufacturing. It breaks for discrete assembly, where every unit can differ and the floor needs unit-level control. Nothing standard exists in that gap, so VW, Mercedes-Benz, ZF and BMW each built their own software between MES and the PLCs. ZF saw the concept in South Africa, went home, and built its own version. That is the strongest validation a gap can have. Enforced assembly names what that layer does when it is done properly, and ODIN is the productised version, priced for every manufacturer that cannot staff an OEM software division.
Yanesh's own, from his book. Dashboards, analytics and now AI bolted onto a fragmented line that cannot carry them. Costed as the Fragmentation Tax: every seam between systems paid for at build, and again at every change. Around 70 per cent of Industry 4.0 pilots never scale, and the foundation is the reason.
One model per line carrying the sequence, rules, variants, settings and per-unit record. The unit becomes the control object: each unit carries its route, allowed actions, tool settings, checks and proof through the line. Competitors can adopt the category. They cannot claim the model without rebuilding their architecture.
Audi: two lines of 160 stations rebuilt as one six-station line building all 60 variants, output more than doubled. Tenneco +80 per cent quality output. SJM -30 per cent downtime. 85+ lines, 5M+ units, 12,000+ devices. Behind it, machine-builder credibility: 3,000+ assembly machines delivered to 35 countries.
On the line, largely yes. And the research reframes the question: in the target mid-market there is usually no rival platform to displace. Only 8 per cent of plants worldwide run a commercial MES, and 54 per cent of small and mid-sized plants still run on pen, paper and spreadsheets. Half of US manufacturing leaders still work from manual logs; 65 per cent of supervisors lose up to four hours a shift reconciling disconnected data. For these plants ODIN is the first real execution system the line has ever had, and it retires the point tools and the reconciliation labour around it. For tier-one plants with an estate, ODIN replaces the fragile line-level stitching and feeds the systems above it. Nothing gets ripped out.
Work instructions and guidance, enforcement and interlocks, error-proofing, per-unit genealogy, rework management, line analytics, variant and BOM logic at line grain, tool settings, and the point tools and seams around them.
Quality capture and SPC, OEE and monitoring, maintenance tasks, station-level training. Claimed in the order of proof depth, never ahead of it.
Scheduling and the build queue (the ERP feeds orders), label systems, energy monitoring, enterprise MES where one exists.
ERP core (finance, procurement, orders, payroll, the enterprise BOM), PLM and CAD, regulated QMS document control, deep multi-plant planning, and the safety-critical PLCs.
The economics carry the claim: systems-integrator fees run 40 to 60 per cent of enterprise-project cost, and a typical eight-to-twelve-system integration bill lands between $100k and $400k. That is the Fragmentation Tax, quantified. The honest position by segment: for the no-MES mid-market line, ODIN is the line execution system of record. For the tier-one estate, it is the missing layer between MES and the PLCs. In both, the ERP, the product source of truth and the quality system of record stay, and saying so is what keeps the claim credible.
Each of these came up in the working sessions and the book. Each has a place in the design. None is dismissed.
Positioning is a hypothesis until the field confirms it. Dunford's own practice is to test the frame on real sales conversations and re-check every six months. Seven tests, each with a number and a pass line. The first three can start within two weeks and need no build.
Four email variants to matched prospect lists: enforced assembly framing, software-defined framing, the current authority framing, and a consolidation framing (one execution layer for the assembly operation, and what you get to turn off). No screenshots, no deck.
Sales runs the Dunford setup on the next ten first calls: the insight, the alternatives, the perfect world, closing on "that is what you would want, right?". Product shown only after agreement.
After each first call, ask the champion to describe ODIN to a colleague in their own words, captured on the follow-up call or in writing.
Put the three criteria and the enforcement claim in front of five controls engineers, Jendamark's own plus two friendly customers, and ask them to break it.
A landing-page variant led by the category definition and the three questions, run against the current page.
Yanesh presents the movement-to-category arc once: the software-defined assembly line as the shift, enforced assembly as the standard that proves you have one. A keynote, a customer session or one of his videos.
Tag every open opportunity with the frame the buyer uses in their own emails and calls: enforcement, OS, no-code, MES, traceability.
The honest exit: T7 is the decision rule. This category earns permanence in the pipeline data or it goes back on the bench. Nothing here is faith.
Every assembly line already runs on software. The problem is that the software is a dozen disconnected systems, and the logic that decides what happens is buried in PLCs where only a controls programmer can touch it. Every change is a project. And nothing on the line has authority: instructions advise, dashboards observe, MES records. At two in the morning, when the wrong part meets the right station, nothing stops the build.
There is another way to run a line. Call it enforced assembly.
The line's entire logic lives in one model a process engineer changes in minutes. Devices physically cannot execute outside it. And because every action is checked as it happens, every unit comes off the line with its proof already written.
Enforced assembly. Build it right first time. Prove every unit.
The eleven descriptions consolidate. Each word keeps one job, and the founder's language keeps the two jobs it does best.