Ask a plant operations head why their continuous reactor train scores 96% OEE while the bottling line next door barely clears 70%, and the honest answer is that the two numbers measure different things. The classic OEE formula was built for discrete units moving past a station — count good parts, divide by ideal cycle time, done. A distillation column doesn't make units; it makes a flow that either sits inside the spec envelope or drifts out of it. Borrow the discrete formula unchanged and you get a flattering number that hides the losses that actually cost a chemical plant money: rate throttling, short runs between grade changes, and yield bleeding off as off-spec product. iFactory's Process OEE Suite redefines availability, performance, and quality for continuous and hybrid chemical operations — measuring against throughput envelopes, run length, and yield rather than unit counts, on a turnkey on-premise NVIDIA stack inside your firewall.
iFactory Process OEE Suite · Chemical
OEE That Actually Fits a Continuous Plant.
Availability, performance, and quality re-derived for flow, not units — measured against rate envelopes, run length, and yield. Real Six-Loss visibility from PLC and DCS signals, on a turnkey on-premise NVIDIA stack, no manual time-logging gap.
90%+
continuous-process world-class band
30-60%
downtime understated by manual logs
12-18 pts
avg OEE gain in 12 months
On-prem
PLC/DCS capture, no cloud
Why Discrete OEE Lies to a Chemical Plant
The standard score multiplies availability by performance by quality, and for a stamping press that works perfectly. A continuous train breaks all three assumptions at once: there is no part count, the "ideal cycle time" is a throughput rate that varies by grade, and quality is a continuous spec band rather than a pass/fail gate. Plug raw flow into the discrete formula and the number inflates, because continuous operation with rare changeovers naturally posts high availability — chemical processing sits closer to the 85% benchmark than almost any other vertical precisely because it runs uninterrupted once started. That high number is real, but it conceals where the money leaks. The Process OEE Suite keeps the familiar three-factor structure so it still benchmarks against the rest of industry, but redefines each factor for flow.
Factor
Discrete definition
Continuous redefinition
Availability
Run time ÷ planned production time, counting full stops
On-stream time ÷ scheduled time — includes rate trips, partial-load running, and grade-change transitions, not just hard shutdowns
Performance
Ideal cycle time × count ÷ run time
Actual throughput ÷ nameplate throughput at the running grade — captures rate throttling the discrete formula never sees
Quality
Good count ÷ total count
In-spec quantity ÷ total quantity produced — off-spec, transition product, and reprocessed material all count against yield
Running a mixed continuous and hybrid plant? Get a turnkey AI quote and we'll map your rate envelopes and spec bands to the right OEE model in the pilot.
The Continuous OEE Waterfall
Every percentage point between scheduled time and truly effective production falls into a named loss. The waterfall makes the three factors legible in sequence — scheduled time at the top, each loss category subtracting from it, effective production at the bottom. This is where a 96% headline number resolves into the rate trips and yield gaps that a plant operations head can actually assign and close.
Availability losstrips, turnarounds, grade-change downtime · −8
Performance lossrate throttling vs nameplate · −9
Quality lossoff-spec + transition material · −5
Effective Production · OEE78%
Illustrative chain: 0.92 × 0.90 × 0.95 ≈ 0.78. The headline "on-stream factor" of 92% looked healthy — the real story was the 9 points of rate throttling and 5 points of yield, both invisible to a discrete unit count.
The Metrics a Continuous Plant Actually Runs On
Beyond the three headline factors, a chemical operations team manages the plant through a handful of continuous-specific measures. The Suite computes each from live process signals rather than shift-end spreadsheets, so they update in real time instead of arriving a day late.
On-stream factor
Availability, continuous form
share of scheduled hours the unit ran on-spec; the chemical industry's native availability measure
Rate vs nameplate
Performance, continuous form
actual throughput against design capacity at the running grade, throttling made visible
Yield gap
Quality, continuous form
in-spec mass ÷ total mass; off-spec, reprocessed, and transition product all subtract
Run length
Campaign stability
mean hours between grade changes or trips; longer runs dilute startup and transition losses
MTBF
Reliability
mean time between failures on critical rotating and static equipment driving availability
MTTR
Recovery
mean time to restore after a trip; the other half of the availability equation
The Measurement Gap Is the Real Enemy
The biggest threat to a useful OEE number isn't a low score — it's a false high one. Manual and semi-automated tracking systematically understates downtime by 30 to 60 percent because operators rarely log stoppages under five minutes, and micro-trips on a continuous unit add up fast. Manual OEE typically reads 8 to 12 percentage points above reality. The Suite reads availability, rate, and spec status directly from PLC, DCS, and historian signals, so the number reflects what the plant actually did rather than what got written down.
Manual / spreadsheet
Best-case
Micro-stops under 5 minutes go unlogged. Reason coding is inconsistent. Rate throttling is invisible. The score flatters — and hides the losses worth closing.
iFactory automatic capture
Actual
Availability, throughput, and spec status pulled straight from PLC/DCS and historian. Every trip and rate dip counted. The number is the one you can act on.
Suspect your current OEE is flattering? Start a 6-week pilot and we'll run automatic capture in shadow against your existing tracking and show the gap.
Benchmark Honestly, Not Against the Wrong Number
The 85% world-class figure was defined for discrete, repetitive manufacturing — roughly 90% availability, 95% performance, 99% quality multiplied together. Capital-intensive continuous processes operate on a different curve by design: world-class for continuous chemical sits in the 90%+ band, while regulated steps with mandatory cleaning and validation legitimately run lower. The Suite benchmarks each unit against its own continuous-process peer band and, more usefully, against its own trailing baseline — because a sustained 5 to 10 point annual gain matters more than chasing a number borrowed from a different kind of plant.
Below 60%
Significant structural losses — major availability or yield problems worth a focused program.
60-75%
Typical operating range once measured accurately; rate throttling and short runs usually dominate.
75-90%
Advanced continuous operation; remaining gap is mostly availability and off-spec at transitions.
90%+
World-class for continuous chemical — sustained on-stream factor with minimal rate and yield loss.
Ask the Copilot
The Suite answers plain language too — for the operations head chasing a number down to its root cause without opening four dashboards.
Plant Copilot — Process OEE Suite
Operations Head
Train 2 OEE dropped from 81 to 73 this week. Where did it go?
Copilot
Almost all of it is performance, not availability. On-stream factor held at 93%, but average throughput fell to 84% of nameplate after Tuesday's feed-quality change forced a rate cut on the C-301 column. Quality stayed at 96%. So the 8-point drop is roughly 7 points rate throttling and 1 point yield — the unit kept running, it just ran slow. I've flagged the C-301 rate limit as the single largest recoverable loss this week.
Operations Head
How does that compare to our own baseline, not the 85% number?
Copilot
Train 2's trailing 90-day baseline is 80.5%, so this week sits about 7 points under its own norm — the dip is real, not noise against the continuous-process band where 78-82% is your steady state. Against discrete "world-class" 85% you'd call it a failure; against your actual peer band and your own trend, it's one recoverable rate constraint. Restoring C-301 to nameplate returns roughly 6 of the 8 points.
Turnkey: Hardware, Software, Live in 6-12 Weeks
iFactory ships a pre-configured NVIDIA AI server — racked, software pre-loaded. Rack it, plug in power and Ethernet, and the Suite is live inside your firewall. The engagement covers cabling, network, PLC/DCS and historian integration, OEE model configuration for your units and grades, operator training, and 24×7 remote monitoring. Your existing control systems are first-class inputs, not migration targets.
Phase 1 · Weeks 1-4
Ship & Connect
Edge server on-prem; PLC/DCS and historian connected. Rate envelopes and spec bands mapped per unit and grade.
Phase 2 · Weeks 5-8
Model & Pilot
Continuous OEE model runs in shadow on a pilot train; loss categories tuned against your operators' ground truth.
Phase 3 · Weeks 9-12
Roll Out & Go Live
Live dashboards across trains, alerting on loss thresholds, operator training, and 24×7 monitoring at 99.9% uptime.
1000+
clients running iFactory
6-12 wks
to live operation
On-prem
inside your firewall
What the Operations Head Gets
One OEE model built for flow means a number that's honest about continuous losses, real-time visibility into rate and yield, and benchmarking against the right band instead of a discrete-manufacturing target that was never meant for a reactor train.
Honest
OEE number
flow-based factors, no flattering unit count
Real-time
Loss visibility
rate and yield losses surfaced as they happen
Right-band
Benchmarking
continuous peer band plus your own trailing baseline
Air-gapped
On-prem deployment
process data never leaves your firewall
Frequently Asked Questions
How is continuous OEE different from the standard formula?
It keeps availability × performance × quality, but redefines each for flow: availability becomes on-stream factor, performance becomes actual throughput against nameplate rate, and quality becomes in-spec mass over total mass. The structure stays comparable to industry; the inputs fit a process that makes flow, not units.
What is the on-stream factor and how does it relate to availability?
On-stream factor is the chemical industry's native availability measure — the share of scheduled time a unit actually ran on-spec, capturing trips, turnarounds, and grade-change downtime. The Suite computes it from live PLC/DCS signals, so partial-load and short-trip losses that manual logs miss are included.
Why is our manual OEE higher than what iFactory reports?
Manual tracking systematically understates downtime by 30 to 60 percent because micro-stops under five minutes and rate throttling go unlogged, so manual scores typically read 8 to 12 points above reality. Automatic capture from control signals counts every trip and rate dip, which is why the honest number is lower — and actionable.
Is 85% the right target for our chemical plant?
Not directly — 85% was defined for discrete manufacturing. Continuous chemical processes run on a different curve, with world-class in the 90%+ band and regulated steps legitimately lower. The Suite benchmarks each unit against its continuous-process peer band and its own trailing baseline, which is the comparison that drives real improvement.
Where does our process data live?
Entirely on-premise inside your firewall on the pre-configured NVIDIA server — read-only and inbound-only to your control systems. OEE models and historical data never leave the plant, with 24×7 remote monitoring and 99.9% uptime. The deployment can be fully air-gapped where required.
Flow-Based OEE. Real Six-Loss Visibility. On-Prem.
See Your Real Continuous OEE
Bring one train and a week of historian data. We'll run the continuous OEE model in shadow, show on-stream factor, rate-vs-nameplate, and yield gap against your own baseline, expose the manual-tracking gap — then scope the 6-to-12-week turnkey deployment, on-prem, inside your firewall.
3 factors
re-derived for flow
90%+
continuous world-class
1000+
clients · 99.9% uptime
