We'd rather you know the boundary before you buy than discover it after. Four focused products, each doing one job well instead of one platform doing everything adequately. Every line below was checked against the product's own source before it was written here; where something is out of scope, it says so, and it says why.
A focused discrete-event simulator for serial and assembly production lines, engineered for accuracy with simplicity. Chasing every feature of a general-purpose platform would make it complex and slow — and still not out-power the specialists. So some things are in, and some are deliberately out of scope.
| Configuration / capability | Status | Why / our stance |
|---|---|---|
| What the Line Simulator models | ||
| Serial flow line — up to 49 stations + a line feeder | Included | Core. 49 is the app's own limit; the line feeder is configurable as a station, the warehouse is not. |
| Parallel identical machines (M/M/c) — PLC-style dispatch | Included | Each finished part goes to ONE of the parallel machines, chosen the way a real line PLC would — by committed load (buffer + parts already on their way): a starving machine still short of a full cycle is fed first (the one closest to completing a cycle), then a running machine still short of a cycle, then the lowest-load machine; stopped or idle machines only when there's nothing better. Full buffers are skipped; ties go to the shortest conveyor. Net effect: feeding rotates, completing one machine at a time instead of piling parts on one. |
| Synchronized assembly · multi-BOM feeders | Included | The assembly waits when a component line can't keep up — a real merge, not just costing. |
| Finite buffers + blocking & starvation | Included | The core Factory-Physics coupling most spreadsheets miss. |
| 7 statistical distributions (fit to data) | Included | Normal, lognormal, triangular, uniform, exponential, gamma, Weibull — with auto-fit. The curve can be set per product, not only per station: in a real line it is the product that governs variability. |
| Breakdowns — MTBF/MTTR, per cycle or per time | Included | Interval and duration are both distributional; failures accrue with usage, not idle time. |
| Scrap & at-station rework | Included | Rework re-runs the cycle where the defect happens — the common case. |
| Setups & multi-product campaigns | Included | Product and material setups; campaign sequencing. |
| Conveyors with transit time | Included | Length ÷ speed delays material and acts as a moving buffer. |
| Batch / container transport between stations | Included | A finite, shared fleet of returnable carts carries parts in batches (capacity, min-load, reorder point / kanban overlap); the cart is the buffer, so transfer-batch size and fleet limits both affect flow — not just one-piece movement. |
| Branched conveyors (fork & merge) | Included | Shared trunk with exclusive branches, fork and merge, with documented modeling recipes. (Mid-trunk accumulation is approximated — see the beta notes below.) |
| Replications · 95% CI · CRN · warm-up | Included | Statistical discipline — a run is not an anecdote. |
| Full-factorial DOE & automatic improvement study | Included | The DOE takes up to 3 factors and reports main effects, the significance Pareto and a recommended optimum. Separately, the report runs an improvement study on three targets — your two biggest losses plus the constraint's own cycle time — at three levels each, and prints the response curve of every one: "cut this by 20% and you gain X%", with the knee visible. It runs at the line's configured horizon, not a shortened one, because a short horizon overstates the main lever by roughly 25%, and in a tool that backs an investment that error points at spending too much. The gains are not additive, and the report says so: fix one, then run it again. |
| Recorded run you can rewind | Included | The recording travels as events, not as fixed-interval snapshots, so the browser reconstructs the line at any instant: 1× to 300×, pause, and drag back to watch the same jam again — none of which touches the server. The app tells you, before recording, how many hours fit your connection. |
| Financials — payback · ROI · ROIC · NPV · IRR · EVA | Included | A line decision becomes a financial one, down to value creation vs. cost of capital. The app itself labels this an engineering estimate for screening and prioritisation — it carries no working capital, financing, sales taxes, inflation, FX, ramp-up or residual value. The full incremental case is the Investment Analyzer's job, and it is in the same licence. |
| Validated vs M/M/1 · Little's Law | Included | Checked against closed-form queueing theory. The engine reproduces the exact results for WIP, flow time and Little's Law within ~2%, in every run, with no special mode. The worst case in the family is 2.24%, at ρ = 0.9; Little's Law itself closes to within 0.02%. And because the engine advances event by event, how often the simulation is polled cannot change the answer: a test sweeps that call granularity by a factor of 2,400× and demands KPIs and state accounting that are bit-identical, not merely close. There is also a suite of model invariants: series adds, parallel is the maximum, the operator limits, variability never creates capacity. A number can be right by accident and a law cannot. Manual §22. |
| Runs in the browser — no specialist, no install | Included | The whole point: anyone who knows the line can run it. The interface is in your browser; the simulation itself runs on a secure server — see "What your connection needs" below. |
| WIP cap / CONWIP (pull) — "Max WIP" | Included | A line-level WIP cap: a new part is released only while WIP is below the cap (pull). Per-stage kanban stays out of scope. |
| Shared / mobile operators with walk time | Included | One operator serves several stations and walks between them at a set speed — the machine waits for them to arrive, and walking is counted as waste, so tightening the layout shows a real gain. |
| Shift calendar & break stops | Included | The whole line pauses between shifts and during breaks (real availability), separate from the labor-cost side. |
| Sequence-dependent setup (from→to) | Included | An optional sparse from→to changeover matrix, consulted when a station switches product in a campaign. |
| Accumulating conveyor | Included | Set a product size and each belt gets a real capacity (length ÷ size); parts accumulate when blocked — toggle synchronous vs accumulating. |
| Advanced stochastic reliability (optional) | Included | An optional engine adds Weibull time-to-failure, micro-stops, imperfect repair and wear/learning curves — off by default, labeled experimental. Limit, stated up front: the micro-stop and medium-stop tiers are defined per completed cycle, so on a line whose shortest cycle is under 5 s they switch themselves off and the app says so; the learning curve and the ageing effects on breakdowns, repair time and quality keep running. |
| Flow view over a plant photo (digital-twin layout) | Included | Lay your stations over a plant photo or drawing, draw the conveyor paths, and watch per-product markers, buffer queues and the live bottleneck. |
| Automated diagnosis + executive report | Included | Ranked findings with cause, $/year value and Lean/TOC/6σ/TPM actions; a print-ready one-page executive summary; and a Check-line configuration validator. |
| Deliberately out of scope | ||
| Queue priority (non-FIFO · SPT / EDD) | FIFO | It doesn't move throughput, OEE or $/year — any work-conserving discipline yields the same throughput and mean WIP (Little's Law). It only changes per-order lateness, a scheduling concern outside our money-first purpose. |
| Rework routed to another station / repair area | At-station | At-station rework (the common case) is fully modeled and hits the bottleneck correctly. Routing defects back upstream is a rarer topology; leaving it out keeps the model simple without changing the throughput logic where rework actually bites. |
| Cumulative gain of elevating two constraints in a row | Out of scope | Removed on purpose in August 2026 — and we would rather say so than leave it implied. The cascade study cost up to 186 runs and only fitted inside a 48-hour horizon, and a short horizon overstates the main lever by ~25%, in the direction that recommends overspending. The study that replaced it runs at the line's real horizon in 8 runs and states plainly that the gains are not additive: attack one, then run it again. |
| Re-entrant flows (loops) | Out of scope | Semiconductor, PCB and plating lines are a specialist vertical; supporting a part that revisits a station would complicate the router for every other user. |
| Job-shop / arbitrary routing | Out of scope | A different tool class. This is a flow-line simulator; arbitrary routing belongs to general-purpose platforms and would trade away the simplicity that lets anyone run it. |
| Disassembly / split (1 → N) | Out of scope | Not our market (meat, primary cutting, recycling). It inverts the part model and adds user-facing complexity for a vertical we don't target. |
| Probabilistic / conditional routing | Out of scope | Rare in the serial and assembly SME lines we serve; branch logic would complicate modeling for the majority who don't need it. |
| AGV fleets / detailed material handling | Out of scope | We model the material handling that actually drives flow — conveyor transit time and parts moving between stations in batches (containers). What we leave out is full AGV-fleet dispatch and routing: a heavy modeling layer that's uncommon in the SME lines we serve. |
| Secondary constrained resources (tools / fixtures / molds) | Out of scope | Out of scope to preserve simplicity; multi-cavity capacity covers the common case. |
| Times below 1 second | Out of scope | Minimum time resolution is 1 second, and the reason is cost, not accuracy. The engine advances event by event, so what a run costs comes from how many pieces it makes and how many stations each piece crosses. Halving a cycle doubles the pieces per simulated hour, and doubles the run with them: a faster line costs more, not less. A second is far below what line balancing, buffer sizing or a business case can resolve, so the floor costs nothing real. Enter 0 for "not applicable". A fast machine that works a batch at a time is fully supported — model the cycle of the whole batch and set pieces/cycle. What is out of scope is piece-by-piece behaviour at sub-second rates (high-speed packaging, bottling, tablet presses). |
The question that comes before all the others. A simulation fed by a cycle time nobody verified, or a capability index computed on a gauge that cannot resolve the tolerance, produces a confident wrong answer — the most expensive kind. 30 analysis tools on one numerical core, with no external dependencies at all.
| Capability | Status | Why / our stance |
|---|---|---|
| What 6Sigma Studio does | ||
| Process capability — Cp, Cpk, Pp, Ppk | Included | With the distinction kept visible: within-subgroup capability is not overall performance, and reporting a Ppk under the name Cpk is the most common way a capability study flatters a process. |
| Control charts with the Nelson rules | Included | Stability judged by rule, with the verdict written out — a chart with limits and no rules is a drawing. |
| MSA / Gage R&R | Included | Repeatability and reproducibility as % of study variation — the number that decides whether the measurement system can even see the variation you are trying to control. |
| DOE — factorial designs, RSM, mixture, desirability | Included | From screening to optimisation, with the run-count and resolution trade-off made explicit before you run the experiment. |
| Distribution fitting, normality tests, Box-Cox | Included | Fit the distribution to your data instead of assuming normal — which is also exactly what the Line Simulator wants for its variability inputs. |
| Reliability & life data — Weibull, MTBF, censored records | Included | Maintenance history always contains units that had not failed yet. Dropping them biases the estimate; these are handled properly. |
| Acceptance sampling — ISO 2859-1 (attributes) & ISO 3951 / ANSI Z1.9 (variables) | Included | Plans are derived from the OC-curve mathematics, not looked up in a table. The variables plan reproduces the published ANSI Z1.9 entry for AQL 1% (n = 42, k = 1.905285), checked against an independent worked example. |
| Tolerance stack-up — worst case, RSS, Monte Carlo | Included | Worst case grows linearly and RSS with the square root; at 10 parts they are 3.16× apart, and which one you used decides whether the assembly is over-toleranced. |
| Regression, correlation, hypothesis tests, variance components | Included | The everyday statistics, each with an interpretation card next to the number instead of a bare p-value. |
| FMEA · Ishikawa · process map · control plan · project documents | Included | The quality-engineering paperwork that has to exist anyway, kept next to the analysis that justifies it. |
| DMAIC workbench as the entry screen | Included | Each step links straight to the tool it needs, so the method drives the software rather than the other way round. |
| 27-chapter manual, 11 chapters with a live demo | Included | The demos run the real engine — move a slider and the same function that produces your report produces the illustration. A test fails the build if a tool ships without its chapter. |
| Limits, stated | ||
| Variables sampling against two limits at once | Single limit | Only single-specification-limit plans are computed. A lot judged against an upper and a lower limit needs the combined-risk "method M", whose table is licensed and not reproduced here. Run each limit as its own plan and take the tighter of the two — conservative, and the app says so on screen. |
| Spreadsheet round-trip | Reads .xlsx, writes CSV | Reading .xlsx uses native browser APIs; writing it natively would mean pulling in a ZIP library, and this product has zero dependencies on purpose — it has to keep working on a plant PC that will never see an install. |
| Statistics as a service to the other products | Not yet | 6Sigma Studio is the suite's single owner of statistics, precisely so the others ask it instead of reimplementing the same maths. That call does not exist yet. |
A business case is not the new line's cash flow. It is with the investment minus without it — and most inflated cases come from comparing an optimistic proposal against a present that was assumed to stand still. This one builds two complete states and subtracts them, line by line.
| Capability | Status | Why / our stance |
|---|---|---|
| What the Analyzer does | ||
| Two complete states, subtracted on every line | Included | Not a delta model. Tax is non-linear — carried-forward losses and thresholds mean the tax on a difference is not the difference of the taxes — so both states are computed in full and then subtracted. The same mechanism gives the value tree and the post-investment review for free. |
| NPV · MIRR · payback · discounted payback · PI · EAA | Included | Payback is counted from the first outlay, so the engineering and installation months where the project only burns cash are inside the number. EAA appears when you compare alternatives with different lives, because ranking those by NPV is simply wrong. |
| IRR — and the refusal to show one | Included | Staged CAPEX plus working-capital recovery produces a cash flow that changes sign more than once, and such a flow has several IRRs or none. The engine counts the sign changes and, when the flow is non-conventional, withholds the IRR and shows MIRR with the reason. A tool that prints "IRR = 340%" loses the CFO in the first meeting. |
| Risk enters as dispersion — scenarios, tornado, Monte Carlo | Included | One discount rate per company, reflecting only the cost of capital. Risk never enters by inflating that rate: charging for risk twice — a padded rate and conservative volumes — rejects good projects and cannot be defended when challenged. The Monte Carlo correlates the inputs that move together, because sampling volume and price independently gives a distribution that is narrow in a way that merely looks like rigor. |
| Every input is an auditable assumption | Included | Value, unit, source, owner, date and confidence — high, medium or low. A number with no origin is a guess wearing a suit, and by the third meeting nobody remembers which was which. |
| "How was this calculated?" on every figure | Included | The trace is emitted by the engine itself, not rebuilt in the interface — so the explanation cannot drift away from the calculation it explains. |
| 14-step case builder that never blocks you | Included | Every step arrives pre-filled with a declared default marked low-confidence; results are visible from step one, labelled preliminary; and after a run the sensitivity ranks which remaining steps can still change the verdict. All 117 inputs carry help explaining what they change in the study. |
| Working capital, depreciation and tax shield, foreign-currency CAPEX | Included | A nominal model with end-of-period discounting, stated in the report rather than left implicit. Equipment quoted in another currency against revenue in yours is the normal case in capital goods, so FX is an input, not an afterthought. |
| Signed snapshot of the approved case | Included | The server signs the inputs, results, engine version and date — and keeps none of it. Tampering with an approved business case becomes detectable without us holding a copy of it. |
| Your financial data never reaches us | Included | The server computes and does not store: no database of cases, and a test that fails the build if a financial field can reach a log. CAPEX and margin are among the most sensitive numbers a factory has, and the strongest way to protect them is not to hold them. |
| Limits, stated | ||
| Multi-user approval workflow | Out of scope | A direct consequence of not storing your data: the case lives in your browser and in a file you control. An approver approves by receiving the signed file, not by clicking inside a shared system. |
| Sync between your devices | Out of scope | Same reason. The case opened on the desktop is not on the laptop unless the file travels with you. |
| Perpetuity terminal value | Out of scope | A machine has a finite life. Putting a perpetuity at the end of it inflates the NPV in a way that is easy to write and impossible to defend. Terminal value is the residual asset plus recovered working capital. |
Most improvement programmes end where the money is approved. This one is built around what comes after: the benefit is measured, not declared, and a project is allowed to close with the gain unconfirmed. A system in which every project ends green is a system that lies.
| Capability | Status | Why / our stance |
|---|---|---|
| What the Project Manager does | ||
| Five methodology templates — DMAIC · A3 / Kaizen · 8D · Safety · blank | Included | Templates are data, not code: exit criteria are composed from a fixed vocabulary of checks. A methodology that does not exist yet can be added without touching the engine — and a test proves it, by inventing one and running it end to end. |
| A phase is cleared by the data, not by a button | Included | There is no "advance anyway". If the exit criterion is not true, the phase is not won — otherwise the gate is decoration and everybody learns to click through it. |
| Measured benefit, with "not confirmed" as a legitimate outcome | Included | There is no path to green without a measured verification window. The success criterion is stamped on its first save and never rewritten — that stamp is the proof the criterion came before the measurement, and editing the text afterwards cannot erase it. |
| Improvement phases you control | Included | Charter · Baseline · Analysis · Implementation · Verification · Standardisation as the starting point — measure before you analyse, analyse before you change, verify afterwards. Add, remove and reorder them; the numbering is positional, so it stays correct when you do. |
| Critical path, with an honest warning | Included | A forward and backward pass giving total and free float per task. But dates here are typed, not derived from durations, so in a project with few links the "critical path" degenerates into "whatever finishes last" — below 50% linked tasks the screen says so, in amber, instead of pretending. Negative float is not a defect: it is the schedule contradicting itself, and it is the most actionable number on the board. |
| Gantt, predecessors (FS · SS · FF), printable status report | Included | Cascading only ever pushes forward — it never pulls a task earlier on its own, because you put that date there for a reason the software does not know. "As soon as possible" is a switch you turn on per task. Once the schedule is approved, every automatic move is logged with its reason. |
| Work stations, not machines | Included | An improvement project attaches to one or more stations, and a station can be entirely manual. |
| Signed project snapshot | Included | The same mechanism as the Analyzer: proof the record has not changed since it was signed. It is not proof the number is right, and the screen says exactly that. |
| Limits, stated | ||
| Working calendar in the schedule cascade | Calendar days | A pushed task moves in calendar days, not working days, and the lag between linked tasks is fixed rather than configurable per link. |
| Start-to-Finish (SF) links | Out of scope | It exists in the textbooks, almost never in practice, and it is the link type that confuses schedules most. It can be added without touching the engine if a real need appears. |
| Statistics inside this product | Out of scope | Deliberate. Capability, control charts and MSA belong to 6Sigma Studio; reimplementing them here would produce two answers that diverge exactly on the hard case. |
| Multi-user, shared server | Out of scope | Single user, data in your browser plus a file you export. The same posture as the rest of the suite. |
The honest version, because this is exactly the kind of claim that ages badly.
The four products sit behind the same access, and the RASQI Hub at app.rasqintelligence.com is where you enter: it launches all four and lays out the improvement cycle. Signing in to one signs you in to the others — verified on the live hostnames, not assumed.
Capacity comes from the simulator, statistics from 6Sigma Studio, the financial verdict from the Analyzer, the measured gain from the Project Manager. No product recomputes another's number — which is why they cannot contradict each other in front of your board.
Not available yet. It is being built. Today you carry the number across yourself. When it ships it will be described here — and not one day earlier.
Not planned, and not by accident. A figure shown away from the conditions that produced it is a figure waiting to be misread: a throughput from an 8-hour run with 3 replications is not the same number as one from 500 hours, and a tile cannot carry that.
Lines, cases and projects live in your browser and in files you export. The servers compute and do not store — the Analyzer's has a test that fails the build if a financial field can reach a log.
Current limitations that aren't design decisions — the honest state of the beta, being worked on.
No accounts, cloud storage or multi-user yet — your lines, cases and projects live in your browser. Portability is via file export/import.
The Line Simulator interface and its 27-section manual are in English and Português; the other three products are in English. Español is coming.
Access is behind a one-time email code — nothing to create, nothing to lose. A running copy auto-detects a newer build and offers to update.
A shared conveyor trunk is modeled as the receiver's input buffer — exact for capacity and blocking; mid-trunk accumulation and junction re-routing aren't physically modeled.
The products run in a browser tab; there's no public API or headless batch runner yet.
These numbers are for the Line Simulator, which is the demanding one: its simulation runs on a secure server while the interface runs in your browser. They are measured, not estimated. The other three compute much smaller models and are undemanding by comparison.
A 2.5 MB transfer, repeated on every visit so you always get the current build. About 2 s on a 10 Mbps link, 20 s on 1 Mbps.
Almost no bandwidth — the line definition goes up, the results come back. A normal office link or 4G is plenty.
About 145 KB per simulated hour, measured against production (the response arrives uncompressed). An 8-hour run is about 1.2 MB. The app tells you, before recording, how many hours fit your connection with full playback controls.
A long study keeps running on the server while your browser checks on it. If the connection drops for more than ~25 s, the study in progress is lost — you start it again. This is the one requirement that changes the decision on an unstable link.
Your saved lines, products and backups live in the browser. Your financial configuration — prices, margins and demand — is never sent: the money side of your business case does not leave your computer.
The last honest thing on a page about capabilities.
Reference figures — not a decision.
Everything these products produce comes from a model, computed from the data you entered. Its accuracy is bounded by that data: a model fed a wrong number returns a confident wrong answer, and no software can detect that the number was wrong — only that it was possible.
Every result must be reviewed by the competent function before it is acted on: process or industrial engineering for capacity and flow, your finance function for a business case, the qualified professional who signs for anything engineered, and your safety and compliance function for anything normative. We give no warranty that an action taken from an output will deliver the modelled result — that depends on execution and on conditions in your plant, neither of which we can see.
This is not us stepping back from the quality of the software. The engines are validated against closed-form theory, every screen states the assumption behind its number, and this page tells you what each product does not do. It is the boundary between a model and a plant, and it does not move because a tool is good. The full wording is in the Terms.
Answer 8 short questions about how material flows through your line and get an honest verdict — the Line Simulator is a flow-line simulator, not a fit for every layout, and we'll tell you straight.
Check compatibility →Between the OEE spreadsheet and the $15k general-purpose simulators: the rigor of Factory Physics — finite buffers, real variability, blocking, 95% confidence intervals, validation against queueing theory — plus the statistics that check the inputs, the incremental business case that survives a CFO, and the project tool that goes back afterwards and measures whether the gain arrived. For serial and assembly SME lines, at a fraction of the price and with no specialist. It gives up job-shop routing, re-entrant flows and AGV-level material handling on purpose, so it stays simple enough for anyone who knows the line.
Buy a machine, add a shift, rebalance, expand — the expensive moves should be tested first.