Aggregate production planning

Aggregate production planning: objectives, strategies (level/chase/hybrid), cost components, constraints, balance equations, and basic LP/heuristic formulations with numericals.

Drafted with Aria, reviewed by the AiCanCode.org team · Updated . How we write and check our notes. Spotted an error? Use Give Feedback at the bottom of the page.

Why it matters

Aggregate production planning (APP) aligns medium-term capacity (workforce, machine hours, overtime, subcontracting) with forecasted demand so factories meet market needs at minimum total cost. It converts forecasts into actionable, period-wise production, workforce, inventory, and backlog targets, stabilizing operations and upstream procurement before detailed scheduling and MRP explode the plan to parts.

Key ideas

  • Aggregate production planning (APP): A medium-term, period-by-period plan (typical horizon 6–18 months; we will use weeks for units here) that sets total output, workforce levels, inventory, backorders, and subcontracting for product families, to minimize total relevant cost while meeting demand and policy constraints. APP is the aggregate plan produced via the Sales & Operations Planning (S&OP) process and it precedes/guides the Master Production Schedule (MPS).
  • Planning horizon and period: Horizon long enough to span capacity changes (hiring/layoffs, overtime policy, subcontracting contracts) and seasonality. The plan is a rolling horizon: each period, we re-plan with updated forecasts and current state (inventory, workforce).
  • Aggregate unit (product family): APP uses a single “aggregate” unit to represent multiple products. Common aggregation bases: average units per family, or standard hours. If using hours, 1 aggregate unit may be 1 standard hour; product i consumes a conversion factor (standard hours/unit), enabling summation across items.
  • Decision variables (by period t):
    • Regular-time production Pᵗ (aggregate units): from in-house workforce within regular hours.
    • Overtime production Oᵗ (units): limited per worker.
    • Subcontracting Sᵗ (units): optional, capacity-limited and higher per-unit cost.
    • Workforce Wᵗ (workers): integer in practice; continuous in simplified models.
    • Hires Hᵗ and Layoffs Lᵗ (workers): adjust workforce across periods.
    • Inventory Iᵗ (units): end-of-period on-hand.
    • Backlog Bᵗ (units): end-of-period unmet demand (if policy allows backorders). If backorders are not allowed, Bᵗ = 0 and inventory must be nonnegative.
  • Constraints:
    • Capacity linkage: Regular-time output is limited by workforce and hours; overtime limited per worker; subcontracting bounded by supplier capacity.
    • Inventory conservation (balance) equation: End inventory minus backlog equals prior net inventory plus total supply minus demand.
    • Workforce balance: This period’s workforce equals last period’s plus hires minus layoffs.
    • Nonnegativity and policy constraints: E.g., no backorders, minimum inventory, service level constraints, or bounds on workforce changes per period.
  • Objective (total relevant cost): Sum of costs over all periods: regular-time labor, hiring, layoff, overtime, subcontracting, inventory holding (per unit per period), and backlog/penalty (per unit per period). If regular-time wages are fixed regardless of production (e.g., salaried), model includes the effect only when workforce changes.
  • Strategies:
    • Level strategy: Keep workforce (and hence regular output) constant; use inventory and backlog to absorb demand fluctuations. Good when hiring/layoff costs are high and holding or backlog costs are moderate.
    • Chase strategy: Adjust workforce to match demand each period; minimize inventory/backlog. Good when hiring/layoff costs are low and backorders/holding are expensive or infeasible.
    • Hybrid strategy: Combine bounded workforce changes with overtime/subcontracting and limited inventory/backlog. This is most realistic.
  • Assumptions commonly used in basic models (state your assumptions when solving):
    • Deterministic period demands (from Forecasting; uncertainty handled by safety stocks or scenario planning outside the core APP model).
    • Constant productivity per worker-hour (no learning or fatigue in regular time), linear costs per unit/worker.
    • No setup-time/detail scheduling effects (handled in Scheduling and sequencing); APP plans totals only.
    • Backorders, if allowed, are fully recoverable (no lost sales unless modeled explicitly).
  • Sign conventions:
    • Iᵗ ≥ 0 (on-hand), Bᵗ ≥ 0 (backlog). Use the net-inventory form Iᵗ − Bᵗ to write one conservation equation that allows either side to be zero.
  • Standard results and insights:
    • If backorders are forbidden and hiring/layoff are prohibitively costly (very high), a level plan is often favored (produce to average demand plus safety stock drift).
    • If holding cost is very high (or material is perishable) and hiring/layoff are inexpensive, a chase plan is favored.
    • Overtime vs backlog: Use overtime in a period if overtime marginal cost per unit is less than backlog penalty per unit per period times expected backlog duration.
    • Subcontracting is used when its marginal cost is below the alternative (overtime or cumulative backlog penalties), and within capacity limits.
  • Model formulations you should know for exams:
    • Tabular/heuristic method: Trial level/chase/hybrid plans, compute costs period-wise, choose the least-cost.
    • Transportation-table approach: Treat production in any period as a “source” that can satisfy demand in the same or future periods with carrying cost arcs; choose min-cost flows subject to capacity. Typically assumes no backorders (only forward-in-time arcs). If backorders are allowed, use explicit backlog variables or allow reverse-time arcs with penalties.
    • Linear programming (LP) or mixed-integer program (MIP): Variables Pᵗ, Oᵗ, Sᵗ, Iᵗ, Bᵗ, Wᵗ, Hᵗ, Lᵗ with linear objective and linear constraints. Integer restrictions may be relaxed for tractability; rounding and feasibility checks follow.
  • Derivation outlines:
    • From workforce to regular production: Assume constant productivity e (units per worker-hour) and H (regular hours per worker per period). Then Pᵗ = e·H·Wᵗ. Add Oᵗ for overtime (bounded by e·Hₒ·Wᵗ).
    • Inventory conservation: Start with stock at the start, add all supply in the period, subtract demand; the remainder is end stock if nonnegative, or backlog if negative. Writing Iᵗ − Bᵗ = Iᵗ⁻¹ − Bᵗ⁻¹ + Pᵗ + Oᵗ + Sᵗ − Dᵗ unifies both cases without cases.
    • Workforce balance: Wᵗ = Wᵗ⁻¹ + Hᵗ − Lᵗ by definition of hires and layoffs.
  • Why real factories deviate from ideal models:
    • Productivity is not strictly constant (learning, fatigue, skill mix), overtime is less efficient, and quality risks rise at high utilization.
    • Hiring has lead times; minimum crew levels, seniority rules, and labor laws bound changes per period.
    • Setup times, lot sizing, and finite capacity at specific workcentres can invalidate family-level linearity; these are handled in Material requirements planning and Scheduling and sequencing.
    • Demand is uncertain; safety stocks (Inventory Control) and service levels shift the APP target trajectory.

Formulas

  • Pᵗ = e · H · Wᵗ

    • Pᵗ: regular-time production in period t (units, count)
    • e: productivity (units per worker·hour, units/(worker·h))
    • H: regular hours per worker per period (h)
    • Wᵗ: workforce (workers, count)
    • Applies when constant productivity and all regular-time is usable for the aggregate family.
  • Oᵗ ≤ e · Hₒ · Wᵗ

    • Oᵗ: overtime production in period t (units)
    • Hₒ: overtime hours per worker per period allowed (h)
    • e, Wᵗ as above
    • Applies with linear overtime productivity and a hard overtime limit.
  • Iᵗ − Bᵗ = Iᵗ⁻¹ − Bᵗ⁻¹ + Pᵗ + Oᵗ + Sᵗ − Dᵗ

    • Iᵗ: end-of-period inventory (units)
    • Bᵗ: end-of-period backlog (units)
    • Sᵗ: subcontracted receipts (units)
    • Dᵗ: demand in period t (units)
    • Applies as the inventory conservation equation each period, with sign convention I, B ≥ 0.
  • Wᵗ = Wᵗ⁻¹ + Hᵗ − Lᵗ

    • Hᵗ: hires (workers)
    • Lᵗ: layoffs (workers)
    • Applies each period to track workforce changes.
  • C_total = ∑ₜ ( c_r·Wᵗ + c_h·Hᵗ + c_l·Lᵗ + c_o·Oᵗ + c_s·Sᵗ + h·Iᵗ + p_b·Bᵗ )

    • c_r: regular-time labor cost per worker per period (INR/worker·period)
    • c_h, c_l: hiring/layoff cost per worker (INR/worker)
    • c_o: overtime cost per unit (INR/unit)
    • c_s: subcontracting cost per unit (INR/unit)
    • h: holding cost per unit per period (INR/(unit·period))
    • p_b: backlog penalty cost per unit per period (INR/(unit·period))
    • Applies when costs are linear and additive. Note: If using per-unit regular-time cost c_p, replace c_r·Wᵗ by c_p·Pᵗ and omit c_r.
  • Q_eq = a_i · q_i

    • Q_eq: equivalent aggregate units for product i (aggregate units)
    • q_i: quantity of product i (units of i)
    • a_i: aggregate units per unit of product i (e.g., standard hours per unit of i if 1 aggregate unit = 1 standard hour)
    • Applies for aggregation by standard hours or a weighted unit. If instead you define b_i = units of product i per aggregate unit, then Q_eq = q_i / b_i.
  • Use overtime in period t if c_o < p_b · k

    • k: expected number of periods the backlog would persist without the overtime (dimensionless)
    • Compares marginal overtime vs cumulative backlog penalties.
  • Produce early (carry inventory k periods) if h · k < (Δc_hl)

    • Δc_hl: per-unit equivalent hiring/layoff cost avoided by smoothing (INR/unit). Estimating Δc_hl requires dividing the total hire/layoff cost by the units enabled/avoided via workforce change.
    • Heuristic inequality: carry inventory k periods only if its cumulative holding cost is less than the avoided adjustment cost.
  • Capacity_t = e · H · Wᵗ + e · Hₒ · Wᵗ + S_maxᵗ

    • S_maxᵗ: maximum available subcontracting in period t (units)
    • Gives an upper bound on supply to meet Dᵗ plus inventory build.
  • Rolling update: I⁰ ← I¹, W⁰ ← W¹, Dᵗ ← Dᵗ(new forecast)

    • At each review, shift horizon forward one period and re-solve with updated state and forecasts.

All quantities of time use hours h (accepted with SI); counts are dimensionless; currency is INR.

Worked examples

Assumptions for both: deterministic weekly demand, constant productivity, linear costs, end-of-week inventory/backlog accounting.

Example 1 — Direct calculation under a level plan Data:

  • Periods: 4 weeks
  • Demand Dᵗ (units): [500, 600, 700, 500]
  • Initial inventory I⁰ = 100 units, initial backlog B⁰ = 0
  • Productivity e = 0.5 units/(worker·h)
  • Regular hours H = 40 h/worker·week
  • Choose a level plan Pᵗ = 600 units each week (no overtime/subcontracting)
  • Holding cost h = INR 12 per unit per week; no backorders allowed (Bᵗ = 0 enforced)
  1. Compute workforce required for Pᵗ = 600. Formula: Pᵗ = e · H · Wᵗ ⇒ Wᵗ = Pᵗ / (e · H). Numbers: Wᵗ = 600 / (0.5 units/(worker·h) · 40 h) = 600 / 20 = 30 workers.

  2. Compute week-by-week ending inventory. Formula (no backlog): Iᵗ = Iᵗ⁻¹ + Pᵗ − Dᵗ.

    • Week 1: I¹ = 100 + 600 − 500 = 200 units.
    • Week 2: I² = 200 + 600 − 600 = 200 units.
    • Week 3: I³ = 200 + 600 − 700 = 100 units.
    • Week 4: I⁴ = 100 + 600 − 500 = 200 units. All Iᵗ ≥ 0, so the plan is feasible without backorders.
  3. Compute total holding cost. Formula: Cost_hold = h · ∑ Iᵗ. Numbers: ∑ Iᵗ = 200 + 200 + 100 + 200 = 700 units·week. Cost_hold = 12 INR/(unit·week) · 700 = 8400 INR.

Final answer: Workforce 30 workers; ending inventories [200, 200, 100, 200] units; total holding cost = 8400 INR.

Example 2 — GATE-style plan with workforce, overtime/subcontracting, holding, and backlog Data (weekly periods, 3 weeks):

  • Dᵗ (units): [220, 320, 180]
  • I⁰ = 40 units; B⁰ = 0
  • e = 0.25 units/(worker·h); H = 40 h/worker·week ⇒ regular output per worker-week = e·H = 10 units
  • Overtime limit Hₒ = 8 h/worker·week ⇒ overtime output per worker-week = e·Hₒ = 2 units
  • Initial workforce W⁰ = 25 workers
  • Costs: c_r = 4000 INR/(worker·week), c_h = 3000 INR/worker, c_l = 5000 INR/worker, c_o = 50 INR/unit, c_s = 90 INR/unit (Sᵗ ≤ 100 units/week), h = 8 INR/(unit·week), p_b = 25 INR/(unit·week)
  • Backorders allowed; no lost sales.

We compare realistic alternatives and choose the least-cost.

Alternative A: Keep workforce at 25, no overtime/subcontracting.

  1. Regular production each week: Pᵗ = 10 · Wᵗ = 10 · 25 = 250 units/week.
  2. Inventory/backlog per week using Iᵗ − Bᵗ = Iᵗ⁻¹ − Bᵗ⁻¹ + Pᵗ − Dᵗ.
    • Week 1: I¹ − B¹ = 40 − 0 + 250 − 220 = 70 ⇒ I¹ = 70, B¹ = 0.
    • Week 2: I² − B² = 70 − 0 + 250 − 320 = 0 ⇒ I² = 0, B² = 0.
    • Week 3: I³ − B³ = 0 − 0 + 250 − 180 = 70 ⇒ I³ = 70, B³ = 0.
  3. Costs:
    • Regular labor: c_r · ∑ Wᵗ = 4000 · (25 + 25 + 25) = 4000 · 75 = 300000 INR.
    • Holding: h · (I¹ + I² + I³) = 8 · (70 + 0 + 70) = 8 · 140 = 1120 INR.
    • Hiring/layoff, overtime, subcontracting, backlog: 0. Total cost A = 300000 + 1120 = 301120 INR.

Alternative B: Lay off to 23 at start; hold 23 constant; no overtime/subcontracting.

  1. Workforce: W¹ = W² = W³ = 23. Layoffs L¹ = 2; Hires = 0.
  2. Regular production: Pᵗ = 10 · 23 = 230 units/week.
  3. I/B trajectory:
    • Week 1: I¹ − B¹ = 40 + 230 − 220 = 50 ⇒ I¹ = 50, B¹ = 0.
    • Week 2: I² − B² = 50 + 230 − 320 = −40 ⇒ I² = 0, B² = 40.
    • Week 3: I³ − B³ = 0 − 40 + 230 − 180 = 10 ⇒ I³ = 10, B³ = 0.
  4. Costs:
    • Regular labor: 4000 · (23 + 23 + 23) = 4000 · 69 = 276000 INR.
    • Layoffs: c_l · L¹ = 5000 · 2 = 10000 INR.
    • Holding: 8 · (50 + 0 + 10) = 8 · 60 = 480 INR.
    • Backlog: 25 · (0 + 40 + 0) = 1000 INR.
    • Overtime/subcontracting: 0. Total cost B = 276000 + 10000 + 480 + 1000 = 287480 INR.

Alternative C: Lay off to 24; hold 24 constant; no overtime/subcontracting.

  1. Workforce Wᵗ = 24; L¹ = 1.
  2. Pᵗ = 10 · 24 = 240 units/week.
  3. I/B:
    • Week 1: I¹ − B¹ = 40 + 240 − 220 = 60 ⇒ I¹ = 60.
    • Week 2: I² − B² = 60 + 240 − 320 = −20 ⇒ B² = 20.
    • Week 3: I³ − B³ = 0 − 20 + 240 − 180 = 40 ⇒ I³ = 40.
  4. Costs:
    • Regular labor: 4000 · 72 = 288000 INR.
    • Layoffs: 5000 · 1 = 5000 INR.
    • Holding: 8 · (60 + 0 + 40) = 800 INR.
    • Backlog: 25 · 20 = 500 INR. Total cost C = 288000 + 5000 + 800 + 500 = 294300 INR.

Check of overtime/subcontract for Alternative B (to see if further improvement is possible):

  • Week 2 shortage before backlog is 40 units. Using overtime in week 2 to remove backlog would cost 40 · 50 = 2000 INR but would save backlog penalty 40 · 25 = 1000 INR ⇒ not worthwhile.
  • Producing the 40 units via overtime in week 1 and carrying them one week would cost 40 · 50 + 40 · 8 = 2320 INR, still greater than the 1000 INR backlog penalty ⇒ not worthwhile.
  • Subcontracting the 40 units (in week 2) would cost 40 · 90 = 3600 INR ⇒ also not worthwhile.

Conclusion: Among A, B, C, the least-cost is Alternative B: lay off 2 workers at start, keep W = 23, accept a 40-unit backlog in week 2 that is cleared by week 3. Final answer: Total cost = 287480 INR, with period-wise (W, P, I, B) as computed.

Answers:

  • Alternative B is optimal among tested plans with total cost 287480 INR.
  • Ending (I¹, I², I³) = (50, 0, 10) units; (B¹, B², B³) = (0, 40, 0) units.
  • Workforce plan: start 25 → lay off 2 → 23, then hold.

Common mistakes

  • Mixing signs in the inventory/backlog balance (forgetting that Iᵗ − Bᵗ carries forward). This creates phantom inventory.
  • Ignoring initial inventory or backlog when computing the first period’s needs, shifting the entire plan off by one period.
  • Double-counting labor: once via c_r·Wᵗ and again via a per-unit regular-time cost derived from e and H. Use one consistent costing approach.
  • Using average demand blindly for level production without checking feasibility with no-backorder policy and capacity limits.
  • Forgetting overtime bounds per worker, or using overtime to build inventory while a cheaper alternative (subcontracting or minor backlog) exists.
  • Treating workforce as continuous and then reporting fractional workers without rounding or feasibility adjustment.
  • Mixing time bases (e.g., monthly demand with weekly hours) without proper conversion.

For GATE ME

  • Short numericals on computing workforce given productivity and hours (P = e·H·W), or computing inventory/backlog trajectories under a proposed plan.
  • Cost-comparison of pure strategies (level vs chase vs basic hybrid) for a 3–6 period dataset; pick the least-cost plan.
  • Formulation of a small LP/MIP: define variables, objective, and constraints (inventory balance, capacity, workforce balance). Often you only need to write the model, not solve it.
  • Transportation-style min-cost flow setup for meeting multi-period demand from multi-period production with carrying costs.
  • Conceptual questions: role of APP in the planning hierarchy (S&OP → APP → MPS → MRP), meaning of aggregate unit, rolling horizon, and when to prefer chase vs level strategies.
  • Practise: set up the inventory/backlog balance correctly; carry units; be fluent with conversions between worker-hours and units; evaluate marginal trade-offs (overtime vs backlog, holding vs hire/layoff).

Quick check

  1. What is the inventory conservation equation with backlog allowed, using the sign convention I, B ≥ 0?
  2. When is a chase strategy more attractive than a level strategy?
  3. If e = 0.4 units/(worker·h), H = 35 h/week, and W = 18 workers, what is the regular-time production per week?

Answers: 1) Iᵗ − Bᵗ = Iᵗ⁻¹ − Bᵗ⁻¹ + Pᵗ + Oᵗ + Sᵗ − Dᵗ. 2) When hiring/layoff costs are low and holding/backlog is expensive or infeasible (e.g., perishables or strict service levels). 3) P = e·H·W = 0.4 · 35 · 18 = 252 units/week.

Finished this topic? Mark it so your progress, study plan and readiness keep up.

Stuck on something here?