Monitoring the Future: How Smart Systems Enhance Dairy Quality and Operations

Engaging introduction

If you walked through a modern dairy plant today, you would see far more than stainless steel tanks and the familiar hum of pumps. You would find calibrated sensors capturing temperatures to a tenth of a degree, vision systems watching fill levels in real time, dashboards flagging anomalies before they become risks, and maintenance plans triggered by data, not guesswork. Technology is reshaping dairy production from farm gate to finished product. For plant leaders and aspiring operators, this transformation is not just about shiny equipment. It is about safer products, better yields, smarter energy use, and stronger careers.

At ELEC, we partner with dairy employers across Europe and the Middle East to recruit and develop talent that can run and improve these smart systems. In Romania, demand is rising in Bucharest, Cluj-Napoca, Timisoara, and Iasi for operators who can read a control screen as confidently as they can swap a gasket. Across the Gulf, mega-dairies invest in automation to maintain consistent quality in tough climates. No matter the region, the future of dairy operations is digital, data-driven, and deeply practical.

This guide explains how today’s processing equipment and monitoring systems enhance dairy quality and operations. You will learn where to start, which technologies matter, how to avoid common pitfalls, and how to upskill for the best roles. Whether you are evaluating a new pasteurizer, planning a SCADA upgrade, or mapping your next career move, the insights below are designed to be actionable.

The digital backbone of a smart dairy plant

What "smart" means in dairy processing

A smart dairy operation connects equipment, people, and decisions through reliable data. The stack typically includes:

• Field devices: Temperature probes, pressure transmitters, flowmeters, level sensors, pH and conductivity probes, inline spectrometers (NIR/FTIR), turbidity sensors, vibration sensors, and power meters.
• Control layer: PLCs and PACs managing interlocks, sequences, and safety; variable frequency drives for pumps and mixers; and HMIs for operator interaction.
• Supervisory and execution systems: SCADA for visualization and alarms; MES for scheduling, recipe enforcement, quality checks, and performance tracking; batch engines compliant with ISA-88 for consistent production steps.
• Business systems and labs: ERP for materials and costing; LIMS for sample plans and results; QMS for nonconformances and CAPA; CMMS for maintenance planning; and DMS for SOPs and training records.
• Analytics and integration: OPC UA and MQTT for connectivity, data historians to store time-series data, dashboards for OEE and SPC, machine learning models for predictive maintenance and process optimization, and APIs to connect it all.

When these components are aligned, operations shift from reactive firefighting to proactive control. Deviations are spotted early, root causes are found faster, and improvements are sustained because they are measured.

Where technology meets the milk: key processes and controls

Milk reception and raw milk assurance

• Tanker scheduling and identity: RFID tags or QR codes link each load to farm, route, and time of collection, preventing mix-ups and accelerating intake.
• Temperature and time integrity: Inline temperature probes verify delivery within chilling limits. Data loggers on tankers feed directly to reception software, flagging any warm loads.
• Rapid quality tests: Inline or at-reception analyzers run antibiotics screening, somatic cell counts, fat/protein/lactose by FTIR, freezing point depression to detect adulteration, and basic acidity.
• Representative sampling: Automated samplers collect proportional samples during unloading while mixing. The sample ID lives in LIMS and links to the batch record.
• Traceability: A unique lot is created right at intake. That lot ID follows the milk through silos, standardization, and packaging, enabling rapid backward and forward tracing.

Result: Early rejection of nonconforming milk, fewer at-risk batches, and better downstream control.

Standardization and blending

• Mass flow meters on skim and cream lines enable high-accuracy ratio control to hit target fat levels.
• Inline NIR spectrometers verify fat and protein in real time; the PLC adjusts valve positions within seconds.
• Recipe management in MES locks down targets by product code, with e-signature approvals for any change.

Result: Consistent composition, reduced over-spec give-away, and fewer lab hold-ups.

Pasteurization and heat treatment (HTST, ESL, UHT)

• PID temperature control with redundant RTDs maintains legal minima and product quality.
• Flow diversion valve interlocks ensure product only passes forward after confirmed time-temperature compliance.
• e-Recorders and data historians replace paper charts, storing complete thermal profiles with audit trails.
• Energy recovery via regenerative heat exchangers monitored for approach temperature; fouling is detected by pressure drop and approach drift, triggering CIP or maintenance work orders.

Result: Verified lethality for safety, tight energy performance, and clear audit evidence.

Homogenization

• High-pressure homogenizer pressure, temperature, and power draw are monitored; trends identify seal wear or cavitation before failure.
• Inline particle size or viscosity proxies verify textural consistency on products like flavored milk and cream.

Result: Stable mouthfeel and extended shelf life with fewer breakdowns.

Fermentation and cultured products

• pH and temperature cascades control culture growth; slope of pH drop is monitored as a process fingerprint.
• Agitator speeds are optimized to prevent shear while ensuring homogeneity.
• End-of-fermentation detection uses pH endpoints, viscosity, or CO2 off-gas rate, feeding automated cooling and downstream scheduling.

Result: Fewer over-acid or underdeveloped batches, higher right-first-time rate, and predictable throughput.

Cheese making specifics

• Curd cutting guided by time, firmness sensors, or vision systems ensures consistent curd size distribution.
• Salting and brining controlled by salinity sensors and temperature; brine quality dashboards track microbial and mineral balance.
• Whey solids recovery is optimized with turbidity meters, reducing losses and improving by-product yields.

Result: Tighter yield, less variability, and stronger flavor profile consistency.

Aseptic processing and filling

• Sterile boundary monitoring with HEPA differential pressure sensors and particle counters.
• H2O2 dosage, UV intensity, or PAA parameters logged during pre-fill sterilization; validation tied to batch records.
• Automated container inspection using vision systems for fill level, cap application, and date code legibility, with feedback loops to adjust fillers.

Result: Secure shelf-life performance and compliant packaging quality.

CIP and hygiene verification

• CIP skids automated on time, temperature, flow, and conductivity; recipe variants for product families.
• Turbidity or fluorescence sensors confirm rinse water clarity and chemical carryover endpoints.
• Chemical dosing tanks monitored for concentration; alarms prevent under- or overdosing.
• ATP rapid tests or biofilm sensors support environmental verification on high-risk surfaces.

Result: Repeatable cleaning, chemical savings, and documented hygiene performance.

Cold chain, storage, and dispatch

• Smart refrigeration controls prioritize energy during off-peak tariffs while keeping product within spec.
• Wireless temperature loggers in cold rooms and trucks feed an integrated cold-chain dashboard.
• WMS with FEFO logic ensures oldest compliant inventory ships first; pallet IDs tie to product and conditions.

Result: Reduced spoilage, fewer temperature abuse incidents, and faster recall response.

How monitoring systems lift quality and safety

HACCP and CCP control

• Digital CCPs: Time-temperature at pasteurization, metal detection on packaging lines, and allergen changeover verifications are all captured digitally with operator confirmations.
• Alarms and interlocks: Threshold breaches auto-stop product flow or trigger line holds to protect consumers.
• Deviations and CAPA: Integrated QMS creates deviations with root-cause workflows and assigns owners and due dates.

Statistical process control (SPC)

• Control charts on fat, fill volume, pH, and viscosity distinguish common cause from special cause variation.
• Real-time alerts prevent drift and reduce overfill give-away by tuning setpoints with data.

Traceability and recall readiness

• End-to-end genealogy from raw milk lot to retail case is available within minutes.
• Mock recall drills use the system to extract affected lots, production windows, and customer shipments.

Data integrity and audit trails

• Electronic signatures, role-based access, and time-stamped logs meet retailer and certification body expectations under GFSI schemes like FSSC 22000, BRCGS, and IFS.

Efficiency, sustainability, and cost control

OEE and losses

• Availability: Monitor unplanned downtime by reason code from line sensors and PLC events.
• Performance: Compare actual to theoretical run speeds by SKU; find micro-stops and tune changeovers.
• Quality: Quantify start-up scraps, off-spec product, and rework; reduce losses with targeted Kaizen events.

Predictive maintenance

• Vibration, ultrasound, and motor current signature analysis predict bearing, pump, or homogenizer issues weeks in advance.
• Condition-based work orders in CMMS align with production windows to avoid peak-hour stoppages.

Utilities and environmental stewardship

• Steam: Boiler efficiency monitoring and condensate return tracking; heat recovery from pasteurizers and compressors.
• Refrigeration: Screw compressor loading optimization, floating head pressure control, and defrost scheduling.
• Water and CIP: Reuse final rinse water for first rinse; optimize caustic concentration and temperature for cleaning efficacy at lowest energy.
• Waste: Inline fat capture, DAF systems for effluent, and whey valorization.

Result: Lower total cost per liter, reduced carbon footprint, and better alignment with corporate sustainability goals.

The human factor: roles, skills, and career paths

Smart systems do not replace people. They amplify skilled operators, technicians, and leaders. Here is how roles are evolving and what employers seek, with a focus on Romania’s key hubs and regional opportunities.

High-demand roles in Romania and the region

• Dairy process operator: Runs lines, monitors HMIs, executes changeovers, and performs in-process checks.
• CIP technician: Configures and verifies cleaning cycles, maintains chemical dosing, and documents outcomes.
• Quality control analyst: Manages raw milk, in-process, and finished product tests; interfaces with LIMS and SPC.
• Automation technician: Troubleshoots PLC, sensors, VFDs, and networks; supports SCADA, changeovers, and recipe downloads.
• Maintenance engineer: Oversees CMMS, predictive maintenance programs, and utility systems.
• MES or SCADA specialist: Configures dashboards, alarms, electronic batch records, and interfaces to ERP.
• Production supervisor or shift lead: Coordinates labor, enforces SOPs and safety, drives OEE and KPI improvements.

Salary ranges in EUR and RON (gross, typical ranges in 2026)

Salaries vary by plant size, shift patterns, and certifications. The following approximate monthly gross ranges reflect typical offers we see at ELEC in Romania. For a quick conversion, 1 EUR is roughly 5 RON. City differences reflect market demand and cost of living.

• Dairy process operator
  • Bucharest and Cluj-Napoca: 1,000 to 1,600 EUR gross (5,000 to 8,000 RON)
  • Timisoara and Iasi: 900 to 1,400 EUR gross (4,500 to 7,000 RON)
• Quality control analyst
  • Bucharest and Cluj-Napoca: 1,100 to 1,700 EUR gross (5,500 to 8,500 RON)
  • Timisoara and Iasi: 1,000 to 1,500 EUR gross (5,000 to 7,500 RON)
• CIP technician
  • Bucharest and Cluj-Napoca: 1,000 to 1,500 EUR gross (5,000 to 7,500 RON)
  • Timisoara and Iasi: 900 to 1,300 EUR gross (4,500 to 6,500 RON)
• Automation technician
  • Bucharest and Cluj-Napoca: 1,500 to 2,300 EUR gross (7,500 to 11,500 RON)
  • Timisoara and Iasi: 1,300 to 2,000 EUR gross (6,500 to 10,000 RON)
• Maintenance engineer
  • Bucharest and Cluj-Napoca: 1,900 to 3,200 EUR gross (9,500 to 16,000 RON)
  • Timisoara and Iasi: 1,700 to 2,800 EUR gross (8,500 to 14,000 RON)
• MES or SCADA specialist
  • Bucharest and Cluj-Napoca: 2,100 to 3,500 EUR gross (10,500 to 17,500 RON)
  • Timisoara and Iasi: 1,800 to 3,000 EUR gross (9,000 to 15,000 RON)
• Production supervisor or shift lead
  • Bucharest and Cluj-Napoca: 1,600 to 2,800 EUR gross (8,000 to 14,000 RON)
  • Timisoara and Iasi: 1,400 to 2,400 EUR gross (7,000 to 12,000 RON)
• Plant manager (for context)
  • Large plants in Romania: 3,500 to 6,000 EUR gross (17,500 to 30,000 RON)

In the Middle East, where dairy operations are often at greater scale, gross monthly salaries are generally higher, for example:

• Automation technician: 2,500 to 4,000 EUR equivalent
• MES or SCADA specialist: 3,500 to 5,500 EUR equivalent
• Production supervisor: 2,800 to 4,500 EUR equivalent
• Plant manager: 6,000 to 10,000 EUR equivalent

Packages can include housing, transport, and schooling allowances depending on the employer and location.

Typical employers in Romania and the Middle East

• Romania

  • Albalact (part of Lactalis Group)
  • Danone Romania (Bucharest)
  • FrieslandCampina Romania - Napolact (Cluj-Napoca)
  • Covalact (part of Lactalis)
  • Olympus - Hellenic Dairies (Brasov region)
  • Hochland Romania (Sibiu area)
  • Dorna Lactate (Lactalis)

• Middle East

  • Almarai (KSA)
  • Nadec and Nada Dairy (KSA)
  • Al Safi Danone (KSA)
  • Baladna (Qatar)
  • Al Ain Dairy and Al Rawabi (UAE)

ELEC works with many of these employers and their suppliers, helping match skilled operators, technicians, and managers with roles that leverage advanced processing and monitoring systems.

Skills that set candidates apart

• Digital literacy: Confident with HMIs, SCADA screens, MES entries, and handheld devices; basic Excel or dashboard navigation.
• Process fundamentals: Pasteurization kinetics, homogenization effects, fermentation curves, and product-specific CCPs.
• Quality and food safety: HACCP Level 3, understanding of FSSC 22000 or BRCGS, allergen management, GMP, and sanitation.
• Automation basics: PLC I/O, analog scaling, sensor calibration, VFD tuning, and network awareness (Ethernet/IP, Profinet).
• Data mindset: SPC, OEE, Pareto analysis, and structured problem solving (5 Whys, fishbone, A3).
• Reliability: Preventive and predictive maintenance principles; interpreting vibration or thermography trends.
• Safety: Lockout-tagout, confined space, chemical handling, and hygiene zoning.
• Soft skills: Communication, shift handover discipline, and continuous improvement ownership.

Certifications and training to consider: Siemens TIA Portal basics, Rockwell Automation training, ISA courses (ISA-88, ISA-95), IFS/BRCGS internal auditor courses, and EIT Food modules. For supervisors, Lean Six Sigma Yellow or Green Belt can be compelling.

Implementation roadmap: from pilot to plant-wide scale

Step 1: Map goals and gaps

• Define metrics: Safety incidents, CCP compliance, OEE, water and energy per liter, yield, labor productivity, and right-first-time.
• Identify bottlenecks: Pasteurizer capacity losses, frequent filler stops, long CIP times, lab holds, and manual data entry.
• Audit your data: Where is data born, where does it go, and who needs it? Inventory all sensors, controllers, and systems.

Step 2: Design the architecture

• Connectivity: Standardize on OPC UA for PLC-to-SCADA and MES; use MQTT for IIoT devices and remote sensors.
• Data models: Align with ISA-95 to structure equipment, materials, and production contexts.
• Cybersecurity: Network segmentation between OT and IT, least-privilege access, patch management, and incident response aligned to IEC 62443.

Step 3: Instrument the process

• Reception: Temperature and flow sensors with historian links; automatic sampler and LIMS integration.
• Pasteurization: Dual RTDs, verified flow meters, and redundant recording to historian and e-recorder.
• Standardization: Inline NIR and mass flow meters tied to PLC ratio control.
• Filling: Checkweighers and vision systems; torque verification for caps.
• CIP: Conductivity, flow, temperature, and turbidity instrumentation; chemical concentration checks.

Step 4: Build the applications

• SCADA: Real-time screens with consistent symbols, alarm rationalization, and mobile views for supervisors.
• MES: Electronic batch records, recipe control, line performance dashboards, electronic checklists, and nonconformance capture.
• LIMS: Sampling plans by SKU and step; automatic import of instrument results; certificate of analysis generation.
• CMMS: Asset hierarchies, criticality ranking, preventive and condition-based tasks, spare-part strategies.

Step 5: Pilot and prove value

• Choose a line with visible pain, such as a UHT filler with frequent micro-stops.
• Baseline 6 weeks of OEE, waste, and energy; document failure modes and MTBF.
• Implement focused changes: vision inspection with closed-loop adjustments, refined changeover SOPs in MES, and vibration sensors on high-fail components.
• Measure again: target 5 to 10 percent OEE improvement, 20 percent reduction in unplanned downtime, and 5 percent energy savings on that line.

Step 6: Scale and sustain

• Standardize what worked: templates for dashboards, alarms, and SOPs.
• Train champions: cross-train operators, technicians, and quality staff; refresh quarterly.
• Governance: Weekly KPI reviews, monthly root-cause clinics, and quarterly audits against digital procedures.

Practical, actionable advice

For plant leaders and engineering managers

• Start with the problem, not the platform. Define the top 3 losses or risks you want to eliminate this year and align technology to those goals.
• Keep humans in the loop. Build HMIs and dashboards around operator workflows. A system people ignore is a system you do not have.
• Insist on data integrity. Enforce user roles, e-signatures, and audit trails. Avoid spreadsheets for CCPs.
• Rationalize alarms. If everything alarms, nothing alarms. Classify severity, add deadbands, and set clear responses.
• Instrument for decisions, not decoration. Every sensor should have a use case, an owner, a calibration plan, and a response rule.
• Treat CIP as a profit center. Optimize cycle counts, reuse water, and check verification trends. Cleaning is often your biggest utility consumer.
• Future-proof connectivity. Standardize on OPC UA and MQTT. Avoid vendor lock-in where possible and plan for API access.
• Bake in reliability. Specify vibration sensors on critical rotating assets, oil analysis ports on gearboxes, and IoT power meters on major loads.
• Pilot quickly, then scale. Ninety-day sprints create momentum and evidence for investment councils.
• Partner smartly. Use system integrators with food experience and references in dairies. Ask for ISA-88 recipe management competence.

For aspiring operators and technicians in Romania

• Build a portfolio. Keep a simple log of problems you solved: a PLC input you diagnosed, an SPC chart you used to stop drift, or a CIP cycle you optimized.
• Get certified. HACCP Level 3 and an internal auditor course for FSSC 22000 or BRCGS help you stand out.
• Learn by doing. Volunteer for changeovers, instrument calibrations, and root-cause meetings. Exposure beats theory.
• Strengthen fundamentals. Know how to read a P&ID, scale a 4-20 mA signal, and verify a RTD calibration.
• Practice data stories. In interviews, explain how you used data to improve yield, reduce downtime, or pass audits.
• Target employers strategically in Bucharest, Cluj-Napoca, Timisoara, and Iasi. Larger sites are more likely to deploy MES, SCADA, and advanced instrumentation, giving you better learning and pay.
• Consider regional mobility. Gaining experience at a major Romanian dairy can be a springboard to roles in the Middle East with higher pay and international exposure.

What good looks like: a day in a smart dairy

• 06:00 - Shift start: Operators review digital shift notes and yesterday’s OEE on the line dashboard. Today’s priorities include ramping yogurt output by 5 percent.
• 07:00 - Reception: A tanker fails antibiotic screening. The system blocks unloading and creates an automatic deviation ticket with supplier details attached.
• 08:00 - Pasteurization: A minor temperature oscillation appears; the PID is autotuned during a planned pause, and the drift disappears.
• 09:30 - Filler: Vision systems catch a trend toward lower fill levels as a nozzle wears. Maintenance swaps the nozzle during a micro-stop, avoiding customer complaints and overfill.
• 11:00 - CIP: Conductivity traces show early plateau, saving 5 minutes per cycle; chemical usage down 8 percent week-over-week.
• 13:00 - Quality: SPC highlights pH variability in a yogurt run. Operators adjust fermentation temperature by 0.3 C and extend hold by 15 minutes, bringing pH back to target.
• 15:00 - Maintenance: Vibration data flags a bearing at risk in a homogenizer. A work order schedules replacement tonight, preventing a line crash tomorrow.
• 17:00 - Dispatch: FEFO logic selects pallets; route temperature records attach automatically to shipping documents.

Result: Targets hit, compliance assured, and no firefighting.

Quick wins to deliver in 90 days

• Reception: Implement automated sample ID and LIMS link. Benefit: eliminate mislabeling and speed up release decisions.
• Pasteurizer historian: Record every second of time-temperature-flow data with simple dashboards. Benefit: instant proof for audits and early drift visibility.
• Filler weight control: Add checkweighers with SPC alarms and feedback to filler setpoint. Benefit: cut overfill costs by 10 to 30 percent.
• CIP verification: Use turbidity or fluorescence sensors for rinse endpoints, plus digital checklists. Benefit: consistent cleaning and 5 to 15 percent chemical savings.
• Downtime tracking: Simple andon or SCADA reason codes per stop. Benefit: fact-based maintenance and OEE gains.
• Energy submetering: Power meters on compressors, pasteurizers, and CIP heaters. Benefit: targeted energy projects with measurable ROI.

Common pitfalls and how to avoid them

• Buying tech without a use case: Result is shelfware and frustration. Tie each feature to a KPI and owner.
• Poor alarm discipline: Leads to alarm floods and missed events. Apply an alarm philosophy from day one.
• Neglecting calibration: Drifting sensors ruin decisions. Create a calibration plan with traceability and exceptions management.
• Manual islands of data: Paper checklists break continuity. Digitize CCPs and integrate lab, production, and maintenance systems.
• Over-customization: Hard-to-maintain spaghetti. Prefer configuration over custom code and document everything.
• Skipping change management: People need training, practice, and feedback. Involve operators early and celebrate wins.

Budgeting and ROI: what to expect

• Sensors and instrumentation: 20,000 to 100,000 EUR for a mid-size plant to close critical gaps.
• SCADA upgrade: 30,000 to 120,000 EUR depending on scope, tags, and redundancy.
• MES starter pack: 70,000 to 250,000 EUR for electronic batch records, performance, and quality modules.
• Vision inspection on a filler: 20,000 to 60,000 EUR per line.
• Predictive maintenance starter kit: 10,000 to 40,000 EUR for sensors and analytics across key assets.

Typical payback windows range from 12 to 24 months, often faster when overfill and unplanned downtime are large contributors.

Compliance and customer expectations

• Certifications: FSSC 22000, BRCGS, and IFS Food are common customer requirements; smart systems simplify evidence.
• Regulatory: EU hygiene package and national authorities expect demonstrable control. In Romania, maintain clear records for inspections by relevant agencies, including traceability and CCP proof.
• Retail audits: Digital records reduce friction in second-party audits and speed up new product introductions.

Conclusion: build tomorrow’s dairy team and systems today

Technology is not a silver bullet, but the combination of well-instrumented processes, disciplined monitoring, and skilled people transforms dairy operations. You get safer products, tighter yields, faster problem solving, and a workplace where operators are empowered by data. In Romania’s hubs like Bucharest, Cluj-Napoca, Timisoara, and Iasi, and across the Middle East, employers are seeking talent that blends process know-how with digital fluency.

If you are a plant leader ready to modernize, or a professional looking to step into a smarter role, ELEC can help. We connect dairies with the technicians, analysts, and managers who can implement and run these systems, and we support candidates with career guidance and market insights. Talk to ELEC to build a hiring plan, benchmark salaries, or explore opportunities that align with your skills.

FAQ

What is the difference between SCADA and MES in a dairy plant?

• SCADA visualizes and supervises real-time equipment, showing states, setpoints, alarms, and trends. It is focused on the here and now of machines and processes.
• MES orchestrates production: scheduling, recipe enforcement, quality checks, electronic batch records, and performance analytics. It connects operations to business outcomes and ensures compliance.

Do small dairies really benefit from smart monitoring?

Yes. Start with the few measurements that matter: pasteurizer time-temperature-flow, filler weight control, and CIP verification. Even a small dairy can cut giveaway, avoid rework, and simplify audits with a historian and basic SPC. Scale features as the plant grows.

Is AI necessary to improve dairy operations?

Not at first. Strong gains come from basic instrumentation, SPC, and OEE discipline. AI adds value for predictive maintenance, anomaly detection, and optimizing complex fermentations once your data is clean and consistent. Crawl, walk, run.

How do we ensure data integrity for audits?

Use role-based access, electronic signatures, time-stamped logs, and immutable historians. Eliminate paper for CCPs. Calibrate sensors on a defined schedule and store certificates. Back up systems and test restorations. Train staff on good documentation practices.

What KPIs should we track weekly?

• OEE by line and SKU
• Right-first-time and quality holds
• Giveaway on fillers by product
• Utilities per liter (water, steam, electricity, refrigeration)
• Planned vs unplanned downtime and top 3 failure modes
• CIP success rates and chemical consumption

What skills help operators transition into automation or MES roles?

Build comfort with HMIs and SCADA navigation, learn basic PLC concepts and networks, practice reading P&IDs, and take an internal auditor course. Volunteer for data projects like downtime coding or SPC charting. A short course in Siemens TIA or Rockwell Studio 5000 boosts credibility.

How fast can we implement a first digital pilot?

In 60 to 90 days for a focused scope. Choose one line, define clear KPIs, instrument the gap, deploy simple dashboards and alerts, and train a cross-functional team. Document results to justify scaling.