Capacity & shift planning
Forecast workload shift by shift and match crews, overtime and line time to it.
Back to all solutions
OPTIMISE PROCESSES · MACHINE LEARNING & PREDICTIVE ANALYTICS
Get more from
your data.
Most operations collect far more data than they use. Machine learning can find the patterns in historical and real-time data that improve yield, energy use, planning and quoting — always inside the limits your engineers set.
01 / IN PRACTICE
Six examples,
problem to result.
Each example is a demonstration on illustrative data, laid out the way a real project runs: the business problem, the data, the AI, the result and the potential value.
01 / PROCESS SETTINGS
Could your existing line run at better settings?
- BUSINESS PROBLEM
- An extrusion line has more combinations of settings than anyone can try, so it runs at settings that are safe rather than best.
- DATA
- Months of process settings from the PLC, matched to the yield, scrap and energy records for each run.
- AI
- A model learns how the settings affect the outcome and suggests changes predicted to improve it, inside the limits your engineers set.
- RESULT
- Suggested settings with the predicted yield and energy use, approved by an engineer before a trial run.
POTENTIAL VALUEBetter output from the equipment you already have.
Process settingsDEMONSTRATION · ILLUSTRATIVE DATA
- Barrel temperature
212 °C→ 204 °C- Line speed
14.0 m/min→ 15.2 m/min- Cooling water
18 °C→ 16 °C
PREDICTED YIELD91.2% → 94.6%ENERGY PER TONNE −6%
- Within safe limits
ENGINEER APPROVES BEFORE ANY CHANGE trial run
02 / ROOT-CAUSE ANALYSIS
Why do good batches sometimes go bad?
- BUSINESS PROBLEM
- Scrap rises and falls with no obvious cause, and the data that might explain it is spread across different systems.
- DATA
- Batch records, material lots, lab results and environmental readings brought together for 18 months of production.
- AI
- A model ranks which factors — and which combinations of factors — best explain the variation in scrap.
- RESULT
- A short list of findings for engineers to test, such as an interaction between humidity and one resin lot.
POTENTIAL VALUEEvidence for where to look first.
Root-cause analysisDEMONSTRATION · ILLUSTRATIVE DATA
- FACTORSHARE OF VARIATION
- Resin lot supplier41%
- Ambient humidity27%
- Mould temperature18%
- Shift / operator5%
FINDING TO TEST
Scrap roughly doubles when humidity is above 68% and resin comes from lot B.
FINDINGS FOR YOUR ENGINEERS TO CONFIRM 18 months of data
03 / SOFT SENSORS
Could you measure quality continuously, not every four hours?
- BUSINESS PROBLEM
- Product moisture is tested in the lab every few hours, so a drift is found long after it started.
- DATA
- Readings from the sensors already on the dryer, matched to past lab results.
- AI
- A soft-sensor model estimates moisture every few seconds from the existing readings.
- RESULT
- A live estimate that tracks the lab samples closely, with an alert when it drifts.
POTENTIAL VALUEProblems spotted in minutes, without new hardware.
Soft sensorsDEMONSTRATION · ILLUSTRATIVE DATA
LAB SAMPLE EVERY4 hESTIMATE EVERY10 sTYPICAL ERROR±0.3%
Model estimateLab sample
ESTIMATED FROM 9 EXISTING SENSORS no new hardware
04 / ENERGY ANOMALIES
Could your energy data tell you what was left running?
- BUSINESS PROBLEM
- Energy bills are high, but nobody can see which equipment is using more than it should, or when.
- DATA
- Hourly meter and sub-meter readings, with the site’s shift pattern.
- AI
- A model learns the normal pattern for each hour and day and flags the periods that don’t fit.
- RESULT
- Unusual periods flagged within hours, with the likely equipment identified.
POTENTIAL VALUEWaste found in hours rather than on the next bill. Savings vary by site, so we measure them rather than assume them.
Energy anomaliesDEMONSTRATION · ILLUSTRATIVE DATA
Sat 00:00–06:00 · +48%
MONTUEWEDTHUFRISATSUN
- Mon – Fri within expected range
- Unit 4 compressor left running overnight
1 UNUSUAL PERIOD FLAGGED ~£38 saved per week
05 / DEMAND & PRODUCTION PLANNING
Could you plan production around what’s coming?
- BUSINESS PROBLEM
- Customer call-offs are hard to predict, so materials are over-ordered or production is rushed.
- DATA
- Two or more years of order history and call-off schedules, plus known seasonal or promotional events.
- AI
- A forecasting model predicts weekly demand with a realistic range rather than a single guess.
- RESULT
- A 12-week forecast with its likely range, refreshed weekly and compared with the current method.
POTENTIAL VALUELess cash tied up in stock, and fewer rushed runs.
Demand & production planningDEMONSTRATION · ILLUSTRATIVE DATA
NEXT 4 WEEKS3,420LIKELY RANGE±6%ORDER STEEL BYTue 14 Oct
TODAY
FORECAST WITH A REALISTIC RANGE refreshed weekly
06 / QUOTE ESTIMATION
Could your job history price the next quote?
- BUSINESS PROBLEM
- Quoting machine hours relies on one estimator’s experience, and quotes vary from person to person.
- DATA
- Past jobs with their part details, materials, operations and actual hours.
- AI
- A model estimates the hours for a new enquiry and shows the most similar past jobs for comparison.
- RESULT
- An estimate with a likely range, ready for the estimator to check.
POTENTIAL VALUEQuicker, more consistent quotes.
Quote estimationDEMONSTRATION · ILLUSTRATIVE DATA
- Part
- Machined bracket
- Material
- Stainless 316
- Batch size
- 250 off
- Operations
- Mill · drill · deburr
ESTIMATED MACHINE HOURS38 h32 h45 h
- Bracket · 316 · 200 off35 h.93
- Bracket · 304 · 250 off41 h.89
- Plate · 316 · 300 off44 h.86
ESTIMATE READY TO CHECK 3 similar jobs
02 / AND MORE
Other ways to
optimise operations.
Late delivery risk
Rank open orders by their chance of missing the due date, with the reasons, early enough to act.
Simulation surrogates
Learn from CFD or FEA runs to screen thousands of design options in milliseconds.
Production scheduling
Sequence jobs to cut changeovers and keep bottleneck machines busy.
Scrap & yield prediction
Predict scrap or yield from material, settings and conditions before a run starts.
Inventory levels
Set reorder points for materials, spares and consumables from real usage.
03 / BEFORE WE BUILD
What we look at
before building anything.
History that covers the patterns
Records going back far enough to include the cycles that matter — often at least a year where demand is seasonal.
A decision it will feed
A prediction is only useful if it changes something: a setting, an order, a shift plan. We start from that decision.
A simple baseline to beat
Compare the model with what you do now, so any improvement is clear and honest.
Engineers set the limits
Suggestions stay inside agreed operating limits and are approved by a person before anything changes.
04 / HOW WE'D START
Idea → evidence
→ investment.
You never commit to a large project on a hunch. Each step answers whether the next one is worth taking — and “no” is a useful answer too.
- 01
Free conversation
30 minutes about a process, machine or business problem. No technical sales pitch.
- 02
AI Opportunity Review
A short paid review of the problem, your data and systems, ending in a roadmap and a recommendation.
- 03
Proof of concept
One problem tested on your real data, with a clear Go, Modify or Stop at the end.
- 04
Build & improve
Only if it works: a practical system, deployed, monitored and refined.
05 / OPTIMISE PROCESSES
Data, but not
enough value?
Tell me about the process or decision you would like to improve and the data you already keep. In a free 30-minute conversation we can talk through what the data could tell you — and whether something simpler would do the job.
Independent · North Yorkshire · working across the UK