The bullwhip effect (BWE) — the amplification of demand and order variability as one moves upstream in a supply chain — is theoretically well characterized for dyadic and serial downstream distribution chains, and information sharing is the most advocated mitigation. Giard and Sali's 2013 review of 53 bullwhip studies concluded that this literature is contingent and incomplete precisely for the productive, upstream part of the supply chain: for production systems serving highly diversified products, no model characterized the joint causal mechanisms by which upstream variability is generated — bill-of-material explosion, production scheduling, lead-time inertia, component commonality, and demand changes induced by pricing or limited-series policies — nor how shared information should be structured so that module- or component-level forecasts are actually usable for planning and control, and existing models were not valid for complex upstream networks as opposed to dyadic or serial chains. The core remains open: convergent/assembly structures still account for only about 4% of information-sharing studies (2015–2024), and reviews through 2025 continue to list supply-chain structure, product type, and information-structuring design as unaddressed research directions. The problem is to build and validate models of variability propagation and its mitigation for complex upstream production networks, and to specify the structure of shared information at the module/component level.

Background

The bullwhip effect (BWE) is the amplification of demand and order variability as one moves upstream — from final customers through distributors, assemblers, and component suppliers — in a supply chain, the network of organizations and activities that delivers a product. Lee, Padmanabhan and Whang (1997) gave the canonical analysis, identifying four operational causes (demand-signal processing, the rationing or shortage game, order batching, and price fluctuations) in largely single-supplier, single-buyer settings. Subsequent theory — variance-amplification bounds, order-up-to inventory policies (replenishing to a fixed target level each period), and the value of sharing point-of-sale demand data — was developed almost exclusively for dyadic (one buyer facing one supplier) or serial (a single line of tiers, each tier being one echelon of the chain) distribution chains.

Giard and Sali (2013) reviewed 53 bullwhip studies along 13 classification coordinates (supply-chain structure, demand model, replenishment policy, causes, information-sharing design, performance measures, and others) and concluded that the literature is contingent — its conclusions hold only under narrow assumptions — and structurally incomplete for the productive (upstream) part of the supply chain, where demand is not exogenous retail demand but is generated internally by production planning itself. In upstream production chains serving highly diversified products, several mechanisms act jointly and remained uncharacterized. Bill-of-material (BOM) explosion: the bill of materials lists the components and subassemblies of each product, and production planning unfolds ("explodes") end-product demand through the BOM into module- and component-level requirements. Production scheduling and lot-sizing: planned orders are time-phased and often batched, which itself creates variability. Lead-time inertia: multi-period procurement and production lead times, together with frozen horizons — the near-term portion of the master production schedule that planners deliberately keep fixed — delay and propagate disturbances. Component commonality: the same component is shared across many product variants, so component demand aggregates correlated variant-specific streams, producing both risk pooling (partial cancellation of variability when several demand streams are aggregated) and correlated surges. Marketing-driven demand changes: price modifications and limited-series policies (temporary or short-lived product variants) deliberately shift demand mix and volume. The review also found that 22 of the 53 studies assumed serial structures, that information sharing was the most-advocated mitigation yet was frequently too abstract or too distant from the physical flow to be usable — for instance, an aggregated end-product forecast is of little help to a component planner sizing orders — and that models valid for complex upstream networks were missing.

Later work confirms this diagnosis from adjacent directions. Dominguez, Cannella and Framinan (2015) showed by simulation that structural factors (the number of nodes and the divergence of the network) drive bullwhip in distribution networks beyond the well-known effect of echelon count, while explicitly restricting their analysis to divergent networks and identifying convergent (assembly) networks as an open extension. The reviews of Wang and Disney (2016), Brauch, Mohaghegh and Größler (2024), and Al-Sukhni and Migdalas (2025) continue to list supply-chain structure, product type, price-related causes, and the design of information sharing as open research directions; in the 2015–2024 information-sharing literature, serial (51%) and dyadic (19%) structures dominate while convergent structures account for only about 4%. What is still missing is a coherent account of how variability is generated and controlled in upstream production chains for diversified products, and of how shared information must be structured — in content, aggregation level (end product, module, or component), direction, and timing — to be usable for planning and control there.

Problem Statement

For an upstream production chain — a convergent (assembly) network producing highly diversified end products from shared modules and components under material-requirements-planning-type control — characterize how demand and order variability is generated and propagated by the joint operation of (i) bill-of-material explosion of end-product demand into module- and component-level requirements, (ii) production scheduling and lot-sizing rules, (iii) lead-time inertia (multi-period procurement and production lead times and frozen horizons), (iv) component commonality across product variants, and (v) marketing-driven demand changes such as pricing modifications and limited-series policies; and, on that basis, determine which information should be shared between which actors, at which aggregation level (end product, module, or component), in which direction and at which timing, so that shared forecasts are usable for planning and control and measurably reduce variability amplification. The single overarching objective is a validated, mechanism-level understanding of upstream bullwhip generation and its control through appropriately structured information: the mechanism model, the network-valid inventory and production control models, and the information-structuring principles are interlocking deliverables of that one objective, since each mitigation claim is only meaningful relative to an explicit propagation model, and an information design is only testable inside a control model that represents how planners actually use it. The target is validity for complex upstream production networks, not merely dyadic or serial downstream chains, and mitigation methods specific to these realistic production environments. The problem is resolved when the proposed models and information-structuring principles, applied to an explicit convergent (assembly) production network, both reproduce the known dyadic and serial variance-amplification results as limiting cases and yield a reduction in variability amplification — measured as the ratio of the variance of orders to the variance of demand at the module and component levels — relative to a no-information-sharing baseline.

References

  1. Lee, Hau L., Padmanabhan, V. and Whang, Seungjin (1997), 'Information Distortion in a Supply Chain: The Bullwhip Effect', Management Science 43(4):546–558. DOI 10.1287/mnsc.43.4.546. https://doi.org/10.1287/mnsc.43.4.546
  2. Giard, Vincent and Sali, Mustapha (2013), 'The bullwhip effect in supply chains: a study of contingent and incomplete literature', International Journal of Production Research 51(13):3880–3893. DOI 10.1080/00207543.2012.754552. https://doi.org/10.1080/00207543.2012.754552
  3. Sali, Mustapha and Giard, Vincent (2015), 'Monitoring the production of a supply chain with a revisited MRP approach', Production Planning & Control 26(10):769–785. DOI 10.1080/09537287.2014.983579. https://doi.org/10.1080/09537287.2014.983579
  4. Sali, Mustapha and Giard, Vincent (2015), 'Optimal stock-out risk when demand is driven by several mixed-model assembly lines in the presence of emergency supply', International Journal of Production Research 53(11):3448–3461. DOI 10.1080/00207543.2014.997398. https://doi.org/10.1080/00207543.2014.997398
  5. Chatras, Clément, Giard, Vincent and Sali, Mustapha (2015), 'High variety impacts on Bill of Materials Structure: Carmakers case study', IFAC-PapersOnLine 48(3):1067–1072. DOI 10.1016/j.ifacol.2015.06.225. https://doi.org/10.1016/j.ifacol.2015.06.225
  6. Dominguez, Roberto, Cannella, Salvatore and Framinan, Jose M. (2015), 'The impact of the supply chain structure on bullwhip effect', Applied Mathematical Modelling 39(23–24):7309–7325. DOI 10.1016/j.apm.2015.03.012. https://doi.org/10.1016/j.apm.2015.03.012
  7. Wang, Xun and Disney, Stephen M. (2016), 'The bullwhip effect: Progress, trends and directions', European Journal of Operational Research 250(3):691–701. DOI 10.1016/j.ejor.2015.07.022. https://doi.org/10.1016/j.ejor.2015.07.022
  8. Brauch, Manuel, Mohaghegh, Matin and Größler, Andreas (2024), 'Causes of the bullwhip effect: a systematic review and categorization of its causes', Management Research Review 47(7):1127–1149. DOI 10.1108/MRR-05-2023-0392. https://doi.org/10.1108/mrr-05-2023-0392
  9. Al-Sukhni, Muthana and Migdalas, Athanasios (2025), 'The impact of information sharing factors on the bullwhip effect mitigation: a systematic literature review', Operational Research 25(2), article 36. DOI 10.1007/s12351-025-00915-3. https://doi.org/10.1007/s12351-025-00915-3