LEAN IQ · THE RC BRAN EXECUTIVE PLATFORM

LEAN IQ The RC Bran Executive Platform

LEAN IQ is the executive platform for a value stream: the map, the work instruction, and the review documents, from one account. Describe the stream in your own words, or start from a worked example. LNSE, the Lean Neuro-Symbolic Engine, draws the current state, names what it had to assume, and answers the eight questions of Learning to See. Each future state gets its own DMAIC. Each kaizen item gets an A3. Film a process and LEAN IQ writes the standard work from what the video shows, then carries the cycle time, changeover, and setup back onto the map. A figure the stream does not state is left blank.

1,201 Worked example streams to start from, thirty in every industry
5 Future states per stream, each with its own DMAIC
A3 One A3 for each kaizen item on the future state
Video Standard work read from each process, then imported onto the map
LEAN IQ See how it works
ON THE PLATFORM

The Map, the Instruction, and the Review.

LEAN IQ runs on LNSE. Every number on a document is a number the stream already stated, or it is marked not observed.

account_tree LIVE

Value Stream Map

Speak or type one product family, or start from one of 1,201 worked streams. LEAN IQ draws the current state, lists what it assumed, answers the eight questions, and ranks five future states with the kaizen each one needs.

  • record_voice_overPlain words, spoken or typed. Or pick an industry and generate a complete stream, then edit it into your own plant.
  • help_outlineIt asks when it needs a number, and marks an assumption when it supplied one.
  • alt_routeWalk the future states. Each one keeps its own kaizen schedule and its own levelled year.
  • printPrint the maps and the analysis. Open an executive summary or a performance view for the review.
  • saveSave the stream and come back to it. One conversation per value stream.
LEAN IQ arrow_forward
videocam LIVE

Standard Work from Video

A value stream is a title and an ordered list of processes. Each process has a sequence number, a name, and a video. LEAN IQ reads the video and writes that process's work instruction. A time the video does not show is left blank.

  • format_list_numberedAdd or delete a process. The sequence numbers stay in order.
  • table_chartA work-element table for the value stream, with the time of each element.
  • bar_chartAn operator balance chart with the cycle time under each process.
  • content_cutA paper kaizen for each process: the non-value-added work named, and the future bar that remains after those cuts.
  • upload_fileA video over 200 MB is compressed before it is read. The largest file accepted is 4 GB.
LEAN IQ arrow_forward
sync_alt LIVE

From the Instruction onto the Map

The cycle time, changeover, and setup read from each process can be imported into the process table of the map. Setup arrives as internal setup. A clock the sheet did not record stays blank.

  • downloadImport from standard work, from the “What I have so far” table.
  • timerProcess name, cycle time, changeover, and setup travel together.
  • blockOne person's time is not written as the process clock when the people disagree.
LEAN IQ arrow_forward
slideshow LIVE

DMAIC and A3

Each future state downloads as its own DMAIC presentation. Each kaizen item downloads as its own A3. Several future states come back together, as separate files.

  • differenceImprove is a gap analysis: the current figure, the future figure, and the size of the gap.
  • routeThe path is the kaizen already on the map — the action, the target, and the undertaking. A step the map does not name is not added.
  • descriptionThe A3 carries the seven parts and the target of that kaizen item, with the related figures from both maps.
  • format_list_bulletedThe decks use the LEAN IQ name, the Sattvam mark, and pointers on the body of each slide.
LEAN IQ arrow_forward
START FROM YOUR INDUSTRY

1,201 Streams Already Written. Pick Yours and Press Generate.

The hardest map to draw is the first one, so there is one waiting for almost every plant and office we could think of.

factory

Thirty a trade, 38 trades, 972 departments

Foundry and forge, machining, stamping, pharmaceutical, food, print, footwear, shipbuilding, jewellery — and the transactional side too: claims, lending, service desks, outpatient clinics, admissions, distribution centres, restaurant kitchens. Every trade has at least thirty streams, so the picker offers a brake disc, a manifold and a pump casing rather than one generic casting.

linear_scale

Whole streams, not fragments

Each one runs raw material to customer, with the demand, the working calendar, every operation's cycle time, changeover, uptime and scrap, and the stock sitting between them — everything a map needs, so nothing is asked of you. The arithmetic is the trade's own: 120,000 blanks a month on a two-second takt in one, a third of a boat a day in another.

bolt

One click, then yours to change

Two dropdowns and Generate. Or copy the narrative into the chat and edit it into your own plant — change the demand, rename the stations, move the stock, and redraw.

verified

Every one of them commits

All 1,201 pass the verifier without a single clarifying question, and are checked on every build. An example that made the engine ask you something would be an example that failed at the one job it has.

They are worked examples rather than templates to fill in: the numbers in each one are internally consistent, so the takt, the loading and the lead time you see are the arithmetic of that stream and not a placeholder. Find the one nearest your own shop, read what it says, and edit from there.

FROM A SENTENCE TO A MAP

You Describe It. It Draws It. Then It Argues Its Case.

Four steps, and the only one that needs you is the first.

01 mic

Say it the way you'd tell a colleague

“We stamp steel brackets for State Street Assembly, 18,400 a month, two shifts…” — typed or spoken. It recognises the companies, the operations, the cycle times, the stock and the schedule out of ordinary prose. Or pick your industry and start from a stream that is already written.

02 draw

The current state, drawn from your words

Standard icons, real data boxes, the stock triangles where you said the piles are, and a lead-time ladder that puts your processing time against your lead time. Every figure traces back to the sentence it came from.

03 quiz

It asks — and admits what it assumed

Only what it genuinely needs, one question at a time, in plain language. Anything it had to supply from the substrate is listed as an assumption, with the reason, so you check the right numbers first.

04 insights

Future states you can defend

Takt, flow, pull, the pacemaker, levelling and the required process improvements — the eight questions, answered in writing. Five strategies, each with its kaizen. From there, a DMAIC for that future state and an A3 for each kaizen item.

LNSE

The Lean Neuro-Symbolic Engine. Not an LLM.

Same architecture and the same mathematics as MNSE on the medical side — pointed at manufacturing.

menu_book

A substrate, not a corpus

1,035 versioned substrates in 14 families — processes, metrics, mechanisms, waste, rules — mined from Learning to See with 1,226 phrase aliases. If a concept is not in the substrate, the engine will not draw it.

rule

Rules that fire, not text that sounds right

Every finding on a map is a named rule with a category — convention, discipline, design, improvement. You can argue with a rule. You cannot argue with a paragraph.

fact_check

A verifier that refuses

A map is committed only when the verifier clears it against a calibrated threshold. When it will not, the tool says which finding blocked it rather than shipping something that reads well.

bolt

The map is drawn on CPU

No GPU and no token bill for the map itself. Drawing a stream costs electricity, not an API invoice. The video reading is the one step that calls a vision model, and only for the frames of that process.

0.920 Encoder macro-F1 on held-out substrate spans
98.97% Verifier precision at the commit threshold
0.988 Verifier AUC on a 3,563-map holdout
100.00% Of 29,018,255 facts stated in narratives, read correctly

The corpus is adversarial on purpose: 100,000 maps with deliberate faults — a pacemaker in the wrong place, pull without kanban discipline, a takt computed on gross time — so the engine is measured on catching what is wrong, not on agreeing with what is easy.

Reading is drilled the same way, and against the harder standard. A second corpus writes 400,000 narratives about the products in that example library, across all 38 industries, and says every fact in a different way each time — the demand monthly, weekly, daily or annually; the stations numbered, lettered, dashed or unnumbered; scrap as a reject rate, a loss, a percentage or “none at all” — and then checks the engine read back exactly what was said. It reads every one of the 29,018,255 facts those narratives state: nothing misread, nothing missed, and nothing invented that was never there. Three quarters of them are a holdout the engine was never tuned against, and it scores the same on those as on the rest.

ACCESS

Signed In, for the Review.

LEAN IQ is the RC Bran executive platform. You sign in, the streams you draw stay on your account, and the documents you download carry the name of the platform. The map itself is still drawn by LNSE, on CPU, from the substrate — a narrative is used to draw that map, and it is not training data for a model.

Open LEAN IQ.

The RC Bran executive platform. Bring one product family, or a video of the process, and the numbers you already know.

LEAN IQ License LNSE for your plant
✓

Message sent successfully!