Operations Management: Production Systems, Facility Location and Layout, ERP, Scheduling, Quality Management and Operations Research

Enterprise resource planning (ERP) is an integrated software system that runs the organisation's core processes — finance, human resources, procurement, production, inventory, sales and distribution, CRM — on a single shared database, so that a transaction entered once updates every function in real time. It evolved from inventory control packages through MRP (material requirements planning, 1970s: explode the master production schedule through the bill of materials against inventory records to time purchase and production orders) and MRP II (manufacturing resource planning, 1980s: adds capacity, finance and marketing) to ERP in the 1990s; the term ERP is credited to Gartner. Benefits are integration, a single version of the truth, process standardisation, reduced inventory and faster closing; the risks are cost and time overruns, customisation creep, resistance to change and the 'big bang' cut-over.

1. Aggregate planning — medium-term (3–18 months) match of capacity to demand by product family, using overtime, hiring, subcontracting, inventory and backorders; options are chase (vary output to demand), level (constant output, absorb with inventory) or a mix.
2. Master production schedule and MRP — what end items to make and when; MRP nets gross requirements against on-hand and scheduled receipts, offsets by lead time, and releases planned orders.
3. Loading — assigning jobs to work centres; the Gantt chart (Henry Gantt) shows load against time. Sequencing — the order of jobs at a centre by priority rules.
4. Dispatching — releasing orders and instructions to start work; monitoring (follow-up, expediting) — comparing progress with schedule and correcting; input-output control tracks planned versus actual flow at each centre.
5. Pull systems — JIT/Kanban (Toyota, Taiichi Ohno) produce only what the next stage withdraws, exposing waste (the seven wastes: overproduction, waiting, transport, over-processing, inventory, motion, defects); the theory of constraints (Goldratt, 'The Goal') schedules to the bottleneck by drum-buffer-rope.

Statistical quality control (SQC) distinguishes chance (common) causes of variation, inherent in the process, from assignable (special) causes that can be found and removed. Control charts plot a sample statistic against a centre line and control limits set at ±3 standard errors: for variables (measured data) the X̄ chart tracks the process mean and the R chart the range; for attributes the p chart tracks the fraction defective, the np chart the number defective, the c chart the number of defects per unit (Poisson) and the u chart defects per unit for variable sample size. A point outside the limits, or a run of seven on one side of the centre line, signals an assignable cause. Acceptance sampling judges a lot from a sample by a plan (n, c): accept if defectives ≤ c; the operating characteristic curve shows the probability of acceptance against lot quality, with the producer's risk α at the AQL (good lots rejected) and the consumer's risk β at the LTPD (bad lots accepted). Process capability compares the specification spread with the process spread: Cp = (USL − LSL)/6σ ignores centring; Cpk = min[(USL − μ), (μ − LSL)]/3σ penalises an off-centre process; Cp ≥ 1.33 is a common target and Six Sigma implies Cp = 2.

• Quality circles — small voluntary groups of 6–12 workers from one area who meet regularly to identify, analyse and solve work-related quality problems using the seven tools; began in Japan in 1962 (Ishikawa, JUSE); in India the first quality circle was formed at BHEL Hyderabad/Ramachandrapuram in 1981 and the Quality Circle Forum of India (1982, Hyderabad) promotes them.
• Total quality management (TQM) — a management philosophy of continuous improvement in every process, customer focus (internal and external), total employee involvement, fact-based decisions, supplier partnership and top-management leadership; its cost-of-quality logic (prevention, appraisal, internal failure, external failure) shows that spending on prevention cuts total cost; the Malcolm Baldrige National Quality Award (USA, 1987) and the EFQM model (Europe, 1991) are TQM frameworks; in India the Rajiv Gandhi National Quality Award (BIS, 1991) and CII-EXIM Bank Award.
• Kaizen — 'change for the better': continuous, incremental, low-cost improvement by everyone every day, as against innovation's large, costly leaps; tools are 5S (seiri, seiton, seiso, seiketsu, shitsuke — sort, set in order, shine, standardise, sustain), gemba walks, the PDCA cycle and suggestion systems.
• Benchmarking — measuring one's processes and performance against the best (Xerox, 1979, against Japanese competitors and later L.L. Bean's warehousing); types are internal, competitive, functional (same function, different industry) and generic; steps: plan, collect data, analyse the gap, adapt and improve.
• Six Sigma — developed at Motorola (Bill Smith, 1986) and made famous by GE under Jack Welch (1995); a process at six-sigma level, allowing the 1.5σ shift, produces 3.4 defects per million opportunities; the improvement roadmap is DMAIC (define, measure, analyse, improve, control) for existing processes and DMADV/DFSS for new ones; belts — champion, master black belt, black belt, green belt, yellow belt.
• ISO standards — the International Organization for Standardization (Geneva, 1947); ISO 9000 family for quality management systems (first 1987; current ISO 9001:2015, built on seven principles — customer focus, leadership, engagement of people, process approach, improvement, evidence-based decision making, relationship management — and a risk-based, high-level structure); ISO 14001 environmental management; ISO 45001 occupational health and safety; ISO 27001 information security; ISO 22000 food safety; ISO 26000 social responsibility (guidance, not certifiable); in India the BIS is the national standards body and accreditation is through NABCB under the Quality Council of India (1997).

Operations research applies mathematical models to decisions; it grew out of military planning in the Second World War (Blackett's team in Britain) and entered industry after 1945. The transportation problem ships from m sources to n destinations at minimum cost. An initial feasible solution comes from the north-west corner rule (start top-left, ignore cost), the least-cost method (fill cheapest cells first) or Vogel's approximation method (VAM: penalty = difference between the two lowest costs in each row and column, allocate in the row or column with the largest penalty — usually nearest to optimal). A non-degenerate basic solution has exactly m + n − 1 allocations; fewer means degeneracy, cured by placing an ε in an empty cell. Optimality is tested by MODI (u–v method): compute uᵢ + vⱼ = cᵢⱼ for allocated cells and the opportunity cost Δᵢⱼ = cᵢⱼ − (uᵢ + vⱼ) for empty cells; all Δ ≥ 0 means optimal, otherwise the most negative cell enters by a stepping-stone loop. An unbalanced problem (supply ≠ demand) gets a dummy row or column with zero cost.

The assignment problem — n jobs to n persons or machines, one each, at minimum cost or time — is a special transportation problem with all supplies and demands equal to 1, solved by the Hungarian method (Kuhn, 1955, after König and Egerváry): subtract row minima, then column minima, cover all zeros with the minimum number of lines; if the lines equal n assign on zeros, otherwise subtract the smallest uncovered element from all uncovered elements, add it at line intersections, and repeat. A maximisation problem is converted by subtracting every element from the largest; an unbalanced one gets a dummy row or column of zeros; a prohibited assignment gets cost M (infinity).

Decision theory classifies decisions by knowledge of outcomes. Under certainty, one state of nature. Under risk, probabilities are known: choose the act with the highest expected monetary value (EMV = Σ pᵢ × payoffᵢ) or lowest expected opportunity loss (EOL); the expected value of perfect information EVPI = expected payoff with perfect information − best EMV = minimum EOL, and is the most one should pay for a perfect forecast; decision trees (squares for decisions, circles for chance) roll back EMVs from right to left. Under uncertainty, no probabilities: maximax (optimist: best of the best), maximin (Wald, pessimist: best of the worst), minimax regret (Savage: minimise the maximum opportunity loss), Laplace (equal likelihood: highest average payoff) and the Hurwicz criterion (coefficient of optimism α: α × best + (1 − α) × worst).