✅
Trusted Scholarly Publisher
Serving Researchers Since 2012

Optimizing Inventory Management in Agro-Processing: Moving from Static Safety Buffers to Seasonal Dynamic Lead-Time Optimization

DOI : 10.17577/IJERTCONV14IS120045
Download Full-Text PDF Cite this Publication

Text Only Version

Optimizing Inventory Management in Agro-Processing: Moving from Static Safety Buffers to Seasonal Dynamic Lead-Time Optimization

Kaustubh Kamlakar

Student, MBA- Operations and Supply Chain Management,

School of Business, Indira University, Pune, India 41103

Dr. Sanchit Pansare

Assistant Professor, School of Business, Indira University,

Pune, India-411033 sanchit.pansare@indirauniversity.edu.in

Dr. Hetal Bhinde

Associate Dean Ph.D. Programs, School of Business,

Indira University, Pune, India-411033 hetalb19@gmail.com

ORCID: https://orcid.org/0000-0002- 6435-511

Abstract – Agricultural inbound logistics operates in high- friction environments characterized by seasonal supply gluts, fragmented village aggregation points, and extreme stochas- tic lead-time variance. Conventional agro-processing enter- prises frequently rely on rigid, transactional spot purchasing and static inventory policiessuch as maintaining a flat 3- day buffer stock year-roundwhich lead to severe opera- tional inefficiencies, capital lock-up, and factory stockouts. This study investigates the theoretical and empirical transi- tion from static safety inventory models to an adaptive Dy- namic Seasonal Safety Buffer Model.

By systematically analyzing empirical inbound data gathered across 180 truck shipments (3,600 metric tons total volume) during a 60-day agricultural harvesting cycle at Sarovar Nu- trifert & Bio-Organics Pvt. Ltd., this research demonstrates how dynamically scaling reorder points and safety stocks across early, peak, and late harvest phases resolves supply chain friction. The findings indicate that transitioning to a dynamic safety buffering framework completely eliminates factory stockout downtime, reduces the average inventory holding footprint by 38.3% (lowering weighted average inven- tory from 300 MT to 185 MT), releases 34.50 Lakhs in blocked liquid working capital, and cuts warehouse biological spoilage by 71.1%.

Keywords – Agribusiness Supply Chain, Stochastic Lead Time, Dynamic Safety Buffers, Inventory Optimization, Agro-Processing Inbound Logistics.

  1. INTRODUCTION AND BACKGROUND

    1. Macro-Environmental Context of Indian Agribusiness

      The Indian agribusiness and food processing sector serves as the vital economic conduit connecting decentralized rural farm- ing communities to industrialized consumer markets. Accord- ing to industry estimates, agriculture and allied sectors contrib- ute approximately 18% to India's Gross Domestic Product (GDP) and employ over 45% of the national workforce. De- spite its macroeconomic prominence, the upstream supply chain remains structurally fragmented. Raw agricultural pro-

      duce typically transitions through multiple intermediariesin- cluding village-level aggregators, local commission agents, wholesale Agricultural Produce Market Committee (APMC) mandis, and regional stockistsbefore finally arriving at cen- tral processing facilities. This multi-tiered channel structure in- troduces significant operational friction, information asym- metry, biological quality degradation, and high transaction costs.

      For an agro-processing enterprise, this structure has a direct consequence: the factory does not buy from a single, predicta- ble supplier but from a shifting web of aggregation points whose reliability changes with the crop calendar. Procurement planners must therefore treat inbound lead time not as a con- tractual constant but as a managed variable. Reviews of plan- ning models in the agri-food supply chain stress that seasonal- ity, perishability and yield uncertainty together set agri-food lo- gistics apart from conventional manufacturing logistics (Ahu- mada & Villalobos, 2009), and that the performance of these chains has to be judged on efficiency, flexibility, responsive- ness and product quality at the same time (Aramyan et al., 2007).

      The processing plant, by contrast, is a fixed-capacity asset. Once commissioned, its conversion capacity, labour roster and utility costs are committed whether or not raw material is wait- ing on the dock. A line that stops for want of raw material con- tinues to incur labour and overhead cost while producing noth- ing, and a line that is fully fed but surrounded by excess stock incurs storage, handling and spoilage cost instead. Inventory policy is therefore the main lever through which the enterprise balances these two costs, and the quality of that policy depends on how faithfully it reflects the variability of the inbound flow.

    2. The Operational Dilemma: Static vs. Stochastic Inbound Flow

      In standard discrete manufacturing environments, production demand is relatively stable, and component lead times exhibit low variance, allowing planners to utilize static safety stock policies (e.g., maintaining a fixed 3-day or 5-day buffer) with high reliability. However, in agro-processing enterprises, the

      inbound supply chain is governed by biological, seasonal, and environmental realities. While daily factory processing demand

      (D) remains strictly fixed by plant conversion capacity, in- bound lead times (LT) are profoundly stochastic.

      Fig. 1. Inbound agricultural supply chain: from smallholder farms through village collection hubs and line-haul transit to the factory dock and ware- house.

      As harvesting seasons progress from early surplus to peak gluts and ultimately to late-season scarcity, rural aggregation net- works experience severe bottlenecks. Village collection centers face labor shortages, transport vehicles experience non-stand- ardized transit delays on unpaved rural roads, and processing plants suffer from weighbridge congestion. Maintaining a rigid, flat inventory buffer under these fluctuating conditions results in two major operational failures:

      1. Excessive Holding Costs During Gluts: During peak har- vesting phases, static buffers cause massive inventory ac- cumulation, leading to high storage costs, physical conges- tion, and increased biological spoilage.

      2. Plant Starvation During Scarcity: During late-season transitions, static buffers fail to absorb extended lead-time delays, resulting in chronic stockouts, unabsorbed factory labor, and costly production line downtime.

      Two features make the problem harder than it first appears. First, the variability is not constant through the season: the same corridor, vehicles and collection centers behave differ- ently in the first fortnight than in the last. A buffer calibrated to one phase is necessarily wrong for the others. Second, the two failure modes are not symmetric in their consequences. Excess stock is costly but recoverable, whereas a stockout stops pro- duction immediately and the lost hours cannot be recovered. Planners facing this asymmetry tend to inflate buffers across the whole season, which solves the stockout problem at the price of the holding-cost problem. The central question of this study is whether that trade-off can be broken by letting the buffer move with the risk.

    3. Problem Statement

      Many agro-processing firms inherit their inventory rules from general manufacturing practice or from the habits of the pro- curement team. A common rule is a flat buffer expressed in days of plant demand, such as three days of cover, applied from the first day of the harvest season to the last. Such a rule is easy to communicate, easy to audit, and easy to program into an en- terprise resource planning (ERP) system. Its weakness is that it implicitly assumes the risk it protects against is the same in every week of the season. The emprical evidence examined in this paper indicates that it is not.

      The problem addressed here can therefore be stated as follows: given a fixed plant consumption rate and an inbound lead time whose level and variability both change across harvest phases, how should the reorder point and safety buffer be set in each phase so that the plant is never starved, while the average stock held, the capital tied up in it, and the biological losses it suffers are kept as low as the evidence allows?

    4. Research Objectives and Questions

      The study pursues four objectives:

      1. to characterize how the level and variability of inbound lead time change across the early, peak and late phases of a harvesting season;

      2. to formulate a dynamic safety buffer and reorder point that scale with the phase-specific lead-time delay;

      3. to benchmark the dynamic policy against a static three-day policy in terms of stockouts, inventory footprint, working capital and spoilage; and

      4. to derive practical guidance for ERP configuration and pro- curement governance.

      These objectives translate into three guiding questions. RQ1: Does inbound lead-time variance differ materially between har- vest phases? RQ2: Can a simple phase-wise rule built on the maximum observed lead-time delay remove stockouts? RQ3: What is the effect of such a rule on average stock, working cap- ital and spoilage relative to the static benchmark?

    5. Purpose and Scope of the Study

      This study addresses these operational vulnerabilities by eval- uating the inbound supply chain of Sarovar Nutrifert & Bio- Organics Pvt. Ltd. The primary focus of this paper is to formu- late, implement, and empirically validate a Dynamic Seasonal Safety Buffer Model that scales safety stock requirements in direct proportion to stochastic lead-time variance across dis- tinct harvesting phases. The operational scope encompasses primary rural consolidation points (farm-gates and Village Col- lection Centers) through line-haul transit to the factory weigh- bridge, quality assurance laboratory, and raw material storage warehouse.

    6. Significance and Structure of the Paper

      The contribution of the paper is practical as much as theoretical. For practitioners, it shows that a three-parameter tableone safety buffer and one reorder point for each harvest phasecan replace a flat rule without any new technology, and it quantifies what the change is worth in stockouts avoided, storage released and capital freed. For researchers, it offers a documented case in which a classical worst-case cover rule is applied to a non- stationary, aggregation-driven inbound chain, together with a sensitivity analysis showing how robust the conclusions are to the main inputs.

      The remainder of the paper is organized as follows. Section 2 reviews the relevant literature on inventory theory, lead-time variability and agribusiness supply chain risk. Section 3 de- scribes the research design, the lead-time decomposition and the phase segmentation. Section 4 develops the mathematical model, applies it to each phase and tests its robustness. Section 5 reports the comparative results. Section 6 discusses strategic implications and an implementation roadmap, Section 7 sets

      out limitations and future research directions, and Section 8 concludes.

  2. COMPREHENSIVE LITERATURE REVIEW

    1. Inventory Management and Stochastic Lead-Time The- ory Under Uncertainty

      The theoretical foundation of inventory management is rooted in balancing holding costs against stockout risks. Classical in- ventory models, extensively detailed by Chopra and Meindl (2016), establish that safety inventory acts as a mathematical cushion against demand and supply fluctuations. In their foun- dational work, Silver, Pyke, and Peterson (1998) formalize the calculation of reorder points (ROP) under conditions of uncer- tainty, demonstrating that safety stock must account for lead- time variability (L) alongside demand variance (D).

      In the standard continuous-review treatment, the reorder point equals expected demand during lead time plus safety stock, and safety stock is set as a multiple of the standard deviation of de- mand during lead time, where the multiplier reflects the target cycle service level (Silver et al., 1998; Chopra & Meindl, 2016). When both demand and lead time are random, the vari- ance of demand during lead time combines the two sources of uncertainty. Eppen and Martin (1988) show that procedures commonly used to combine them can give results far from those intended, and they present a correct alternative; their work underlines that lead-time variability deserves careful treatment in its own right rather than being folded into a de- mand-based rule of thumb. When demand is constant, as it is for a plant running at fixed capacity, the combined expression collapses so that all residual uncertainty is attributable to lead time.

      However, standard economic order quantity (EOQ) and reorder point formulations assume that lead time is an independent, identically distributed random variable with low variance. When applied to agricultural raw materials, these assumptions break down. Agro-processing supply chains exhibit non-sta- tionary lead-time distributions that shift dynamically across harvesting windows. Consequently, static safety stock policies inevitably produce system imbalances, manifesting as chronic stockouts during high-variance periods or excessive working capital lock-up during stable periods. The important point for this study is that the usual statement of the model treats the lead-time parameters as fixed numbers estimated once and ap- plied throughout the planning horizon. If those numbers move with the harvest calendar, the policy must move with them.

    2. Lead-Time Reduction and Supply Chain Responsiveness

      Expanding on the mechanics of supply chain responsiveness, De Treville, Shapiro, and Hameri (2004) argue that managing and actively buffering against lead-time variability is signifi- cantly more critical for operational performance than merely inflating inventory buffers. In the context of agricultural supply chains, lead time is not a static parameter managed by a single vendor; it is a composite duration shaped by multiple sequential handoffs: order confirmation, farm-gate aggregation, primary rural transport, and factory dock offloading.

      Failing to isolate these component lead times prevents manage- ment from identifying specific bottlenecks. De Treville et al. emphasize that operational agility requires adaptive inventory

      mechanisms that react in real-time or seasonally to lead-time stretching, thereby minimizing the Total Cost of Ownership (TCO).

      Hopp and Spearman (2008) express a closely related idea in their treatment of variability in flow systems: variability that is not reduced at its source must be absorbed by one of three buff- ersinventory, capacity or timeand every system pays for its variability in one of these currencies. For the agro-processor, plant capacity is fixed and waiting time for raw material is simply idle labor, so in practice inventory remains the buffer of choice. The policy question is then not whether to hold a buffer but how much to hold, and when.

    3. Agribusiness Supply Chain Risks and Quantifiable Agil- ity

      Agribusiness supply chains possess unique physiological and economic characteristics that distinguish them from industrial manufacturing. Shukla and Jharkharia (2013) provide a com- prehensive review of agri-fresh and semi-processed produce supply chains, highlighting the structural friction involved in managing biological raw materials where moisture content, for- eign matter contamination, and post-harvestdecay introduce severe quality risks. They note that traditional supply chain frameworks fail to account for the physical degradation of ag- ricultural inventory over time.

      Furthermore, Behzadi et al. (2018) review quantitative models in agribusiness supply chain risk management, emphasizing that stochastic modeling and scenario planning are mandatory for handling agricultural supply uncertainties, seasonal crop availability, and upstream transport volatility. Their findings corroborate the need for dynamic inventory policies that protect processing plants against supply shocks while mitigating ware- house holding losses.

      The degradation of stock while it is held is the subject of the perishable-inventory literature. Nahmias (1982) reviews mod- els in which items lose value or expire with age, so that the cost of holding stock includes a component that grows with the time spent in storage. For semi-perishable commodities such as grains and pulses the rate of loss is far slower than for fresh produce, but in humid warehouses it is not zero, and it accumu- lates in direct proportion to the time inventory is held. This is the mechanism by which an oversized buffer imposes costs that go beyond interest on capital.

    4. Seasonality, Variability and the Case for Time-Varying Buffers

      Fisher (1997) argues that a supply chain should be matched to the nature of the product it serves, with physically efficient chains for predictable products and responsive chains for prod- ucts whose demand is uncertain. Although framed around de- mand, the logic extends naturally to supply: when the uncer- tainty sits on the inbound side, the buffer should sit there too, and it should move as the uncertainty moves.

      The literature on the bullwhip effect points in the same direc- tion. Lee, Padmanabhan, and Whang (1997) identify order batching, price fluctuation and rationing or shortage gaming among the causes of variability amplification upstream in a supply chain. Late-season competition among regional proces- sors for a dwindling crop resembles the rationing behavior they

      describe: buyers who expect shortage inflate their require- ments, which in turn makes the supplier's environment more volatile. A buffer policy that anticipates this phase, rather than reacting after the first stockout, is therefore well founded.

      Textbook treatments of non-stationary demand recompute base-stock or reorder parameters period by period rather than fixing them once (Zipkin, 2000). The present study applies the same logic to non-stationary lead time, using the harvest phase as the planning period. The choice of phase rather than day as the unit of adjustment is deliberate: it keeps the policy simple enough for planners to administer while still capturing the ma- jor shifts in operating conditions.

    5. Supply Chain Resilience and Risk Management

      Christopher and Peck (2004) argue that resilience depends on visibility, flexibility and collaboration across the chain, and that firms should build resilience into the design of their supply net- works rather than rely on emergency responses. Tang (2006) surveys approaches to supply chain risk management, includ- ing inventory-based, supply-base and flexibility strategies, and stresses the value of proactive measures against operational risks such as delay and disruption. In both accounts, a seasonal buffer schedule is a proactive measure: it is set in advance using known features of the crop calendar and observed lead-time be- havior, and it does not wait for a disruption to reveal itself.

    6. Summary of the Reviewed Literature

      Table 1. Summary of reviewed sources and their relevance to this study

      Source

      Focus

      Relevance to this study

      Silver et al. (1998); Chopra &

      Meindl (2016)

      Safety stock and reor- der-point theory

      Baseline model; assumes stationary lead time

      Eppen & Martin (1988)

      Safety stock under ran- dom lead time and de- mand

      Care needed when lead time varies

      De Treville et al. (2004)

      Lead-time reduction and variability

      Variability, not the mean, drives cost

      Hopp & Spear- man (2008)

      Buffering of variability

      Inventory as buffer when capacity is fixed

      Zipkin (2000)

      Foundations of inven- tory management

      Period-by-period parame- ters for non-stationary con- ditions

      Shukla & Jharkharia (2013)

      Agri-fresh supply chains

      Biological degradation of stock

      Behzadi et al. (2018)

      Agribusiness risk mod- els

      Need for stochastic, sea- sonal policies

      Ahumada & Vil- lalobos (2009)

      Agri-food planning models

      Seasonality and perishabil- ity

      Aramyan et al. (2007)

      Agri-food performance measurement

      Multi-dimensional perfor- mance view

      Nahmias (1982)

      Perishable inventory

      Holding cost grows with time in store

      Fisher (1997); Lee et al. (1997)

      Matching chain to prod- uct; bullwhip

      Buffer placement; shortage gaming

      Christopher & Peck (2004);

      Tang (2006)

      Resilience and risk man- agement

      Proactive, calendar-based buffers

      Summary prepared by the author from the sources cited in Sections 2.12.5.

      Table 1 shows that the strands of literature are complementary rather than competing. Inventory theory supplies the buffering logic, the lead-time and variability literature explains why the

      buffer must respond to the spread of delivery times, and the ag- ribusiness and resilience literatures explain why the response must be seasonal and proactive. What is missing is the combi- nation of the three in a compact operating rule, which is the gap addressed next.

    7. Research Gap

      This study bridges the gap between theoretical inventory buff- ering and empirical agribusiness operations by demonstrating how a seasonally adaptive, dynamic safety buffer model can eliminate plant downtime, optimize warehouse footprints, and release trapped working capital in agro-processing inbound supply chains.

      Three specific gaps motivate the work. First, classical safety- stock models treat lead-time parameters as stationary, which does not hold across a harvest season. Second, within the sources reviewed here, agribusiness risk studies concentrate on modeling frameworks and review-level guidance, and less on compact operating rules that a procurement planner can apply on the shop floor. Third, there is limited published evidence quantifying the stockout, capital and spoilage effects of sea- sonal buffer adjustment in aggregation-based procurement of the kind common in Indian agro-processing. The following sec- tions address each gap in turn.

  3. METHODOLOGY AND DATA FRAMEWORK

    1. Research Design and Operational Scope

      To evaluate the operational impact of lead-time variance and establish a robust empirical foundation for dynamic inventory buffering, this study adopted a quantitative observational re- search design. Empirical data was gathered over a rigorous 60- day operational window spanning the active agricultural har- vesting season at Sarovar Nutrifert & Bio-Organics Pvt. Ltd. The dataset tracks a sample size of 180 inbound truck ship- ments, representing a cumulative raw material volume of 3,600 metric tons (MT) sourced across 12 regional Vllage Collection Centers (VCCs) and partnered Farmer Producer Organizations (FPOs).

      The design is observational and comparative. It does not ma- nipulate the supply chain; it records inbound shipments over a full harvest cycle, estimates phase-specific lead-time statistics, and uses those statistics to parameterize the buffer model. Pol- icy outcomes under the static rule, based on the enterprise's his- torical performance, are then compared with outcomes under the dynamic rule. The unit of analysis is the inbound shipment; the 180 shipments and 3,600 MT in the sample imply an aver- age consignment of 20 MT.

    2. Lead-Time Decomposition Framework

      To isolate the exact operational bottlenecks driving supply chain friction, total inbound lead time (LTtotal) was decomposed into four sequential operational stages:

      1. Purchase Order Confirmation (torder): The duration from internal plant MRP generation to formal PO transmission and vendor acknowledgment.

      2. Farm Aggregation (tharvest): The time required for small- holders and Village-Level Aggregators (VLAs) to harvest, thresh, dry, bag, and consolidate produce at decentralized Village Collection Centers.

      3. Line-Haul Transit (ttransit): The physical transportation du- ration for 3-ton rural collection trucks and 10-to-20 ton line-haul carriers to travel from rural aggregation hubs to the central processing facility.

      4. Dock Quality Control & Weighbridge Processing (tdock): The time spent at the factory gate undergoing gross/tare weighbridge measurement, pneumatic core sampling, and laboratory quality assay testing.

      LTtotal = torder + tharvest + ttransit + tdock (1)

      The sequence therefore runs from PO generation, through farm aggregation and line-haul transit, to dock quality control and weighing, as illustrated in Fig. 1. The decomposition matters because each stage responds to a different kind of intervention. Order confirmation delays are addressed by process and ERP automation; aggregation delays by coordination with VCCs and FPOs; transit delays by vehicle scheduling and route plan- ning; and dock delays by weighbridge and laboratory capacity. This paper reports lead time at the aggregate level, so the stage- level split is presented as a framework for diagnosis and is left for more detailed future measurement.

    3. Seasonal Harvest Categorization

      To capture non-stationary operational dynamics, the 60-day sourcing window was segmented into three distinct harvesting phases:

      • Early Harvest (Days 115): Characterized by initial mar- ket arrivals, uncalibrated field collection workflows, and moderate supply availability.

      • Peak Harvest (Days 1645): Marked by maximum volume gluts, intense farming activity, and severe downstream lo- gistics congestion at factory receiving docks.

      • Late Harvest (Days 4660): Defined by rapid crop exhaus- tion, widespread supply scarcity, extended aggregation times, and heightened competition among regional proces- sors.

        The early and late phases each span 15 days, or one quarter of the window, while the peak phase spans 30 days, or one half. Segmenting by phase rather than by calendar week keeps the policy simple enough for planners to administer: three parame- ter sets instead of sixty daily values. It also matches how pro- curement teams already talk about the season, which eases adoption.

    4. Statistical Variance Tracking

      Statistical deviation (L) was computed for each harvesting phase to measure lead-time uncertainty. Empirical tracking re- vealed that lead-time variance escalated significantly as the season progressed, shifting from ±11.4 hours in the early har- vest to ±24.2 hours in the late harvest. This escalating variance drove maximum observed lead times (LTmax) from an early baseline of 3 days up to 5 days during late-season shortages.

      Table 2. Observed lead-time profile by harvest phase

      Phase

      Days

      Avg. LT (days)

      Avg. LT (hours)

      Max. LT (days)

      Early

      115

      2.0

      48.9

      3.0

      Peak

      1645

      2.0

      47.1

      4.0

      Late

      4660

      2.7

      64.6

      5.0

      LT = inbound lead time. Values as reported by the study.

      Three patterns stand out in Table 2. First, the maximum lead time rises by two thirds between the early and late phases (3 to 5 days), and the reported deviation more than doubles (11.4 to 24.2 hours, an increase of about 112%). Second, the average lead time rises by about 32% over the same interval in hours (48.9 to 64.6). Third, and most instructive, the average does not rise in the peak phase: at 47.1 hours it is slightly below the early-harvest figure of 48.9 hours, yet the maximum is a full day longer. Congestion at peak therefore shows up in the tail of the distribution, not in its center. A policy that is calibrated to average lead time would miss precisely the risk that matters.

    5. Modeling Assumptions

      The model developed in Section 4 rests on the following as- sumptions, which are stated explicitly so that readers can judge where the results do and do not transfer:

      • Daily plant demand D is constant at 100 MT/day, fixed by processing capacity, so demand uncertainty is negligible relative to lead-time uncertainty.

      • Within a phase, the lead-time statistics (average, maximum, deviation) are stable, and the phase boundaries are known in advance from the crop calendar.

      • Replenishment is triggered when inventory reaches the re- order point for the current phase.

      • Capital is valued at a flat 30,000 per MT, the average raw material valuation used in the study.

      • The raw material is semi-perishable (grains and pulses), and spoilage is measured as a percentage of stored volume.

      • The static benchmark is the enterprise's historical policy of a flat 300 MT buffer (three days of plant demand) held year- round.

    6. Validity Considerations

      Because the comparison is between the enterprise's historical performance under the static policy and its performance under the dynamic policy, differences in crop conditions, market ar- rivals or weather between the two periods could influence the outcomes independently of the policy itself. The results should therefore be read as strong operational evidence from one en- terprise and one season, not as the output of a controlled exper- iment. Section 7 returns to this point and outlines how future work could strengthen causal attribution.

  4. MATHEMATICAL MODEL AND DYNAMIC SAFETY BUFFER COMPUTATION

    1. Theoretical Formulation of the Dynamic Buffer

      Classical inventory management models safety stock as protec- tion against demand uncertainty. However, in agro-processing operations, daily factory conversion demand (D) is held strictly constant by plant processing capacity (configured at 100 MT/day), whereas replenishment lead time (LT) is highly sto- chastic. When lead time fluctuates between average duration (LTavg) and maximum observed duration (LTmax) due to rural aggregation and transit bottlenecks, the required safety buffer must absorb the maximum lead-tme delay (LT = LTmax LTavg).

      The mathematical formulation for the dynamic safety buffer is expressed as:

      SB = D × (LTmax LTavg) (2)

      where SB is the safety buffer in MT. To maintain continuous factory feeding without holding excessive inventory during low-variance periods, the dynamic Reorder Point (ROP) is computed as:

      ROP = D·LTavg + SB (3)

      Substituting Eq. (2) into Eq. (3) yields a useful identity. The average-lead-time terms cancel, leaving:

      ROP = D·LTmax (4)

      The dynamic reorder point is therefore the quantity of material the plant would consume during the longest lead time observed in the phase. The model is a worst-case cover rule: it guarantees that an order placed at the reorder point will not run the plant dry even if it arrives as late as the slowest delivery seen in that phase. The advantage of this formulation is its transparency. Every input is directly observable, and a planner can reproduce the calculation with a spreadsheet. Its main limitation, dis- cussed in Section 7, is that the maximum is a single extreme observation and is therefore sensitive to outliers.

      For comparison, the classical service-level formulation sets safety stock as a multiple of the standard deviation of lead-time demand. With constant demand, that deviation equals D times the lead-time deviation, giving:

      SS = z·D·L (5)

      where z is the standard normal deviate for the chosen service level. Section 4.4 uses Eq. (5) as an independent cross-check on the buffers produced by Eq. (2).

    2. Phase-Wise Empirical Computations

      Applying the dynamic safety buffer formulas to the empirical lead-yield data yields distinct operational inventory parameters across the three harvesting phases.

      Early Harvest Phase (Days 115). Average lead time is 2.0 days (48.9 hours) and maximum lead time is 3.0 days (72.0 hours), so the lead-time delay is LT = 1.0 day. The required safety buffer is 100 MT/day × 1.0 day = 100 MT, and the dy- namic reorder point is (100 × 2.0) + 100 = 300 MT.

      Peak Harvest Phase (Days 1645). Average lead time is 2.0 days (47.1 hours) and maximum lead time is 4.0 days (96.0 hours), so LT = 2.0 days. The required safety buffer is 100 MT/day × 2.0 days = 200 MT, and the dynamic reorder point is (100 × 2.0) + 200 = 400 MT.

      Late Harvest Phase (Days 4660). Average lead time is 2.7 days (64.6 hours) and maximum lead time is 5.0 days (120.0 hours), so LT = 2.3 days. The required safety buffer is 100 MT/day × 2.3 days = 230 MT, and the dynamic reorder point is (100 × 2.7) + 230 = 500 MT.

      Fig. 2. Dynamics of inventory reorder point (ROP) and safety buffer by harvest phase. Quantities in MT.

      Table 3. Dynamic safety buffer and reorder point by harvest phase

      Phase

      LT

      (days)

      Safety buffer (MT)

      ROP (MT)

      Capital in buffer ( Lakhs)

      Early

      1.0

      100

      300

      30.00

      Peak

      2.0

      200

      400

      60.00

      Late

      2.3

      230

      500

      69.00

      Capital in buffer = safety buffer × 30,000 per MT (the valuation used in the study).

      Table 3 adds a financial reading of the buffers. At 30,000 per MT, each day of plant demand held as buffer (100 MT) ties up

      30.00 Lakhs. The early-harvest buffer therefore carries 30.00 Lakhs, the peak buffer 60.00 Lakhs and the late buffer 69.00 Lakhs. The dynamic policy spends this capital only when the lead-time risk justifies it. The static policy, by contrast, carries a flat three-day cover (300 MT, 90.00 Lakhs) in every phase, regardless of how much delay risk the phase actually presents.

    3. Reading the Model: Coverage Gaps of a Static Rule

      Because Eq. (4) expresses the reorder point as the demand dur- ing the longest observed lead time, it also gives a direct meas- ure of how well a static three-day trigger (300 MT) would have covered the worst case in each phase. The gap is simply the difference between the lead-time demand at LTmax and the static trigger.

      Table 4. Worst-case coverage of a static three-day (300 MT) trigger

      Phase

      Demand over LTmax (MT)

      Static trigger (MT)

      Worst-case gap (MT)

      Early

      300

      300

      0

      Peak

      400

      300

      100

      Late

      500

      300

      200

      Illustrative worst-case bounds computed from the study parameters; not observed outcomes.

      Table 4 shows that the static rule is exactly sufficient in the early phase, where the longest lead time is three days, but leaves a worst-case shortfall of 100 MT (one day of production) in the peak phase and 200 MT (two days of production) in the late phase. These are upper bounds that materialize only when the slowest deliveries coincide with a depleted stock, so they do not predict how often a stockout will occur. They do, how- ever, explain why the observed stockouts concentrated in the late harvest: that is where the static rule's exposure is largest. In the early phase the dynamic reorder point coincides with the static cover; the schedule departs from the flat rule only as the season's delay risk grows.

      A worked illustration makes the mechanics concrete. Suppose that in the late phase stock falls to 500 MT and an order is placed. If the delivery arrives after 2.7 days (the phase average), the plant has consumed 270 MT and still holds 230 MT, which is the safety buffer, when material arrives. If the delivery ar- rives after the maximum of 5.0 days, the plant has consumed 500 MT and stock reaches zero exactly as material arrives, with no production lost. Under a static 300 MT trigger, the same five-day delivery would leave the plant two days short, which corresponds to the unabsorbed labor and idle-line hours de- scribed in Section 5.2.

    4. Cross-Check Against the Classical Service-Level Ap- proach

      As an independent test, the buffers from Eq. (2) can be com- pared with those given by Eq. (5), treating the reported ± values in Section 3.4 as one standard deviation of lead time. The peak- phase deviation is not reported in the study, so the comparison is made for the early and late phases only. Two common service levels are used: z = 1.645 (95%) and z = 2.326 (99%).

      Table 5. Dynamic buffer versus classical service-level buffer

      Phase

      L (h)

      D·L

      (MT)

      z = 1.645 (MT)

      z = 2.326 (MT)

      Dynamic buffer

      (MT)

      Early

      11.4

      47.5

      78.1

      110.5

      100

      Late

      24.2

      100.8

      165.9

      234.5

      230

      D·_L is computed with D = 100 MT/day and _L converted to days. Illustrative cross-check.

      In the late phase the dynamic buffer of 230 MT sits almost ex- actly at the 99% service-level figure of 234.5 MT. In the early phase it lies between the 95% and 99% figures (78.1 and 110.5 MT). The two approaches are built from different statistics one from the maximum, the other from the standard devia- tionyet they point to similar buffer sizes. Tis agreement supports the reasonableness of the worst-case rule. Under the stated assumptions (deviation read as one standard deviation, approximately normal lead time), the dynamic buffer behaves like a high-service-level policy, giving roughly 9599% pro- tection in the two phases where both methods can be computed.

    5. Sensitivity Analysis

      The model has three inputs that a planner might doubt: the lead- time delay, the daily demand and the unit valuation of stock. Because the formulas are linear, their effects can be traced ex- actly.

      Lead-time delay. From Eq. (2), each additional half-day of de- lay adds 50 MT of safety buffer, which at 30,000 per MT ties up 15.00 Lakhs. If the late-phase delay had been 2.8 days in- stead of 2.3, the buffer would be 280 MT rather than 230 MT. The buffer therefore scales one-for-one with the quality of the lead-time estimate, which argues for refreshing it at every har- vest season.

      Daily demand. Safety buffers scale proportionally with plant throughput. A ±10% change in D (90 or 110 MT/day) changes the early, peak and late buffers as shown in Table 6.

      Table 6. Sensitivity of safety buffer (MT) to plant demand

      Phase

      D = 90

      D = 100

      D = 110

      Early

      90

      100

      110

      Peak

      180

      200

      220

      Late

      207

      230

      253

      Safety buffer = D × LT, with LT of 1.0, 2.0 and 2.3 days.

      Unit valuation. The capital released by the policy equals the reduction in average inventory (115 MT) times the price per MT. Table 7 shows how the capital figures change if the valu- ation is 10% lower or higher than 30,000.

      Table 7. Sensitivity of working-capital effect to raw-material price ( Lakhs)

      Price per MT

      Static (300 MT)

      Dynamic (185 MT)

      Released

      27,000

      81.00

      49.95

      31.05

      30,000

      90.00

      55.50

      34.50

      33,000

      99.00

      61.05

      37.95

      Computed from the study parameters; price levels are illustrative.

      Even with a 10% fall in price the released capital remains above

      31 Lakhs, so the conclusion about capital is robust to realistic price movements. The release also earns a recurring return. Ap- plying illustrative annual cost-of-capital rates of 8%, 10% and 12% to the 34.50 Lakhs released gives indicative annual car- rying-cost savings of about 2.76, 3.45 and 4.14 Lakhs re- spectively, before counting the avoided spoilage and stockout costs. The actual rate depends on the enterprise's financing mix and should be substituted by its finance team.

    6. Extension: A Percentile-Based Variant

      Because Eq. (4) anchors the reorder point to a single extreme observation, planners who prefer a less outlier-sensitive rule can replace the maximum with a high percentile of the phase's lead-time distribution. Writing LTp for the p-th percentile (for example the 95th), the variant is:

      SB*_{p} = D × (LT_{p} LT*avg) (6)

      ROP*_{p} = D·LT_{avg} + SB***p (7)

      When p = 100 the variant reduces to the model used in this pa- per. Lower values of p trade a small, explicit risk of stockout for a smaller buffer, which allows management to price the pro- tection it is buying. The variant has one practical requirement: percentile estimates are themselves noisy when a phase con- tains few shipments, so a minimum number of observations per phase should be agreed before the estimate is trusted, and the worst-case rule should remain the default until that threshold is met.

    7. Sensitivity to Phase Boundaries

      The phase boundaries in Section 3.3 are an input to the policy, so it is reasonable to ask how much the average inventory im- plied by the model depends on where they are drawn. Using the phase buffers of 100, 200 and 230 MT, the days-weighted av- erage safety buffer under the 15/30/15 segmentation is (15 × 100 + 30 × 200 + 15 × 230) ÷ 60 = 182.5 MT, within about

      1.4% of the 185 MT weighted average inventory reported in Section 5.3. Table 8 shows how the figure moves if the bound- aries shift.

      Table 8. Time-weighted average safety buffer under alternative phase lengths

      Early / Peak / Late (days)

      Average buffer (MT)

      Capital ( Lakhs)

      Change vs.

      182.5 MT

      15 / 30 / 15 (study)

      182.5

      54.75

      15 / 25 / 20

      185.0

      55.50

      +1.4%

      10 / 30 / 20

      193.3

      58.00

      +5.9%

      20 / 30 / 10

      171.7

      51.50

      5.9%

      Capital = average buffer × 30,000 per MT. Illustrative; phase lengths other than the study segmentation are hypothetical.

      Even a five-day shift of days from one phase to another changes the average buffer by under 6%, and every alternative in Table 8 remains far below the static 300 MT. The policy is therefore not fragile with respect to the exact placement of the phase boundaries, which is reassuring for a procurement team that cannot know the crop calendar to the day. What matters more is that the late phase, with its higher buffer, is recognized before it starts.

  5. RESULTS AND COMPARATIVE EMPIRICAL ANALYSIS

    1. Benchmarking Static vs. Dynamic Inventory Policies

      To measure the operational and financial efficacy of the pro- posed mathematical model, the historical performance of the enterprise under a traditional static inventory policy (maintain- ing a flat 300 MT buffer stock year-round) was directly bench- marked against the newly implemented dynamic seasonal safety buffer policy. Four outcome measures are examined: stockout incidents and associated downtime, average inventory footprint, blocked working capital, and in-storage spoilage.

    2. Elimination of Factory Stockouts and Production Down- time

      Under the legacy static policy, late-season lead-time spikes and aggregation bottlenecks stretched actual delivery times beyond the 3-day buffer threshold, resulting in 4 major stockout inci- dents during the late harvest phase. These stockouts forced pro- cessing lines to halt operations, generating 38 operating hours of unabsorbed labor and factory idle downtime. Transition- ing to the dynamic seasonal buffer policywhich escalated safety stock to 230 MT during the late harvestcompletely ab- sorbed the 2.3-day lead-time delay, achieving 0 stockout inci- dents and ensuring uninterrupted manufacturing continuity.

      The location of the failures is itself informative. All four inci- dents fell in the late phase, which is the phase with the largest worst-case gap in Table 4 and the highest lead-time deviation in Table 2. This is consistent with the model's central claim that exposure to stockout risk is concentrated where lead-time vari- ance is highest, and that a buffer allocated by phase protects the plant more effectively than one spread evenly across the sea- son.

    3. Inventory Footprint and Working Capital Optimization

      Holding a rigid, maximum-capacity buffer yer-round under the static policy locked up excessive warehouse space and fi-

      300 MT down to 185 MT, representing a massive 38.3% re- duction in average inventory storage.

      Priced at an average raw material valuation of 30,000 per MT, this reduction in holding volume successfully lowered blocked capital from 90.00 Lakhs down to 55.50 Lakhs. Conse- quently, the dynamic inventory policy permanently released

      34.50 Lakhs in liquid working capital, which was redirected toward core business expansion and farmer partnership pro- grams.

      Fig. 3. Capital impact of the static policy (flat 300 MT buffer, 90.00 Lakhs blocked) versus the dynamic policy (185 MT weighted average,

      55.50 Lakhs blocked), releasing 34.50 Lakhs.

      The released capital has a strategic as well as a financial value. In agro-processing, working capital is often the binding con- straint at exactly the moment procurement is most intense, be- cause farmers and aggregators expect prompt payment at the peak of the season. Freed capital can be used to pay partners faster, to secure supply in tight periods, or to finance capacity upgrades, all of which reinforce the reliability of the inbound chain that the buffer is designed to protect.

    4. Mitigation of Warehouse Biological Spoilage

      Excessive inventory holding periods in humid warehouse envi- ronments accelerate biological degradation and insect infesta- tion in agricultural grains and pulses. By maintaining a leaner, faster-moving inventory profile under the dynamic policy, the enterprise successfully reduced in-storage biological spoilage from 3.8% of stored volume down to 1.1%, marking a 71.1% reduction in warehouse holding losses.

      A simple flow relationship helps interpret this result. When throughput is fixed, average inventory equals throughput mul- tiplied by the average time material spends in stock (Little's law). At 100 MT/day, an average inventory of 300 MT corre- sponds to about 3.0 days of cover, while 185 MT corresponds to about 1.85 days. The dynamic policy therefore shortened the average residence time of stock by the same 38.3% by which it reduced the footprint. Because biological degradation grows with time in storage (Nahmias, 1982), a shorter average resi- dence time is consistent with the observed fall in spoilage, alt- hough this study does not isolate other contributing factors such as storage conditions or lot quality.

    5. Consolidated Performance Comparison

      Table 9. Consolidated comparison of static and dynamic policies

      Indicator

      Static

      Dynamic

      Change

      Stockout incidents (late harvest)

      4

      0

      100%

      nancial capital. By aligning inventory holding levels with ac- tual seasonal variance rather than maintaining a static cushion,

      the weighted average inventory holding footprint dropped from

      Indicator

      Static

      Dynamic

      Change

      Idle / unabsorbed la- bor hours

      38 h

      0 h

      Eliminated

      Weighted average in- ventory (MT)

      300

      185

      38.3%

      Average days of cover*

      3.00

      1.85

      38.3%

      Blocked capital ( Lakhs)

      90.00

      55.50

      34.50 L

      In-storage spoilage (% of stored volume)

      3.8%

      1.1%

      71.1%

      *Derived: average inventory ÷ 100 MT/day. Other values as reported by the study.

      Table 9 collects the principal results. The dynamic policy im- proves all four dimensions at once, which is notable because inventory policies usually trade one against another: a larger buffer reduces stockouts but raises holding cost and spoilage. The explanation lies in how the buffer is allocated. Instead of carrying a uniform cushion that is too thick in some phases and too thin in others, the dynamic policy redistributes the same protective logic across the season so that the stock is held when and where the delay risk is real.

    6. Cost-Benefit Considerations

      The benefits reported above must be set against the costs of adopting the policy. Those costs are modest and mostly organ- izational. They consist of the analyst time needed to estimate phase parameters, the effort of configuring the ERP calendar and parameter table, and the discipline of the review at each phase boundary. No capital equipment is required, and no change to physical storage is needed. The dynamic policy does carry a larger buffer in the late phase than in the early phase 130 MT more, equal to 39.00 Lakhs at the stated valuation but this increment is deliberate and is more than offset, over the season as a whole, by the lower weighted average.

      On the benefit side, the study documents four distinct items: 38 hours of idle production avoided, 34.50 Lakhs of capital re- leased, a 38.3% smaller storage footprint, and a spoilage rate cut from 3.8% to 1.1% of stored volume. Because the policy requires no capital outlay, the payback is effectively immedi- ate, and the main risk is not cost but complacency: if parameters are not refreshed each season, the policy will gradually revert to a stale flat rule.

  6. DISCUSSION AND STRATEGIC IMPLICATIONS

    1. Re-engineering Agribusiness Supply Chain Strategy

      The empirical findings of this study challenge the uncritical adoption of manufacturing inventory theories in agribusiness. Because agricultural supply chains are subject to biological cy- cles, weather shocks, and fragmented rural infrastructures, static inventory policies are inherently flawed. Supply chain managers in agro-processing enterprises must recognize that lead-time variance is not an exogenous nuisance to be absorbed by brute-force inventory inflation; rather, it is a quantifiable op- erational parameter that must be managed through dynamic, phase-aware mathematical modeling.

      This conclusion is consistent with the broader literature. De Treville et al. (2004) argue that variability in lead time, rather than its average, drives operational penalties, and the observa- tion in Section 3.4 that peak-phase congestion appears in the

      tail rather than the mean of lead time supports that view. Hopp and Spearman (2008) frame inventory as one of three possible buffers against variability. The results here indicate that, when plant capacity is fixed and idle time is expensive, inventory is the right buffer, but only if its size is tied to the variability it is meant to absorb.

    2. Why a Phase-Aware Buffer Outperforms a Flat One

      The flat buffer fails for a structural reason. It is a single number chosen to cover the worst phase, or a compromise that covers no phase well. If it is set high enough for the late harvest, it overstocks in the early phase by two days of demand; if it is set for the early phase, it leaves the plant exposed in the late phase. No single value can be right across phases whose delay risk ranges from one day to more than two. A phase-aware schedule removes this dilemma by allowing the cushion to rise and fall with the risk.

      There is also a behavioral advantage. A flat rule gives planners no signal when conditions change, so they respond to the first stockout by raising the buffer permanently, and it is rarely low- ered afterward. A phase table builds the expectation of changeinto the policy itself: planners know in advance that the buffer will rise at the start of the peak phase and again in the late phase, and can plan storage, financing and supplier engagement accordingly.

    3. Integration with Enterprise Resource Planning (ERP) Systems

      For these dynamic models to scale beyond single-season pilot projects, organizations must institutionalize automated ERP re- order alerts. By programming ERP systems to automatically adjust reorder points (ROP) and safety stock thresholds as har- vesting calendars transition from early to peak and late phases, companies can eliminate human error, reduce administrative overhead, and ensure seamless coordination between field pro- curement officers and central warehouse planners.

      Fig. 4. Proposed flow from field procurement and FPOs, through an auto- mated ERP reorder engine and dynamic ROP / seasonal buffer, to the fac- tory processing plan.

      In practical terms the ERP configuration is modest. It requires a calendar table that maps dates to harvest phases, a parameter table holding the safety buffer and reorder point for each phase, and a rule that applies the parameters of the current phase when the system evaluates stock against the reorder point. Because the phase table is the only moving part, it can be reviewed and signed off by procurement, finance and operations together be- fore each season, which provides a natural governance check- point.

    4. Implementation Roadmap

      The transition from a static to a dynamic policy can be staged to limit risk. Table 10 outlines a five-step roadmap that the en- terprise, or a similar processor, could follow.

      Table 10. Implementation roadmap for a dynamic seasonal buffer

      Step

      Activity

      Outcome

      1

      Define phases

      Agree harvest-phase boundaries from the crop calendar and prior-season ar- rivals.

      2

      Capture lead times

      Record shipment-level timestamps for each of the four lead-time stages.

      3

      Set parameters

      Compute LTavg, LTmax, safety buffer and ROP for each phase using Eqs. (2)(3).

      4

      Configure ERP

      Load phase calendar and parameter table; enable automatic reorder alerts.

      5

      Review and refine

      Compare outcomes with targets after each phase; update parameters before the next season.

      Roles: procurement, warehouse, finance and IT/ERP teams jointly own steps 15.

    5. Procurement and Supplier-Side Implications

      The buffer model takes lead time as given, but lead time is partly under the enterprise's influence, and procurement prac- tice can complement the inventory policy. Closer coordination with Village Collection Centers and Farmer Producer Organi- zations can shorten the aggregation stage, for example through advance commitments in the early phase when supply is plen- tiful and pickups can be scheduled in batches. Staggering truck arrivals reduces queues at the weighbridge during the peak phase, where dock congestion is most severe. Laboratory ca- pacity at the dock can be planned in line with the peak arrival schedule so that quality assays do not become the binding stage.

      These measures act on the numerator of the buffer formula. Every day removed from the maximum lead time reduces the late-phase safety buffer by 100 MT and releases 30.00 Lakhs of capital at the stated valuation. Inventory policy and supplier development are therefore complements rather than substitutes: the first protects the plant against delay that already exists, and the second reduces the delay that has to be protected against.

    6. Monitoring and Governance

      A dynamic policy is only as good as the feedback that maintains it. A small set of indicators, reviewed at each phase boundary, is sufficient to keep it calibrated:

      • Service indicators: stockout incidents and idle production hours, tracked against a target of zero.

      • Lead-time indicators: average, maximum and deviation of lead time by phase, compared with the values used to set the buffer.

      • Inventory indicators: average stock and days of cover against the phase reorder point.

      • Financial and quality indicators: capital blocked in raw material and spoilage as a percentage of stored volume.

        If observed lead-time maxima exceed the values built into the table, the buffer should be revised upward before the next phase; if they fall consistently below it, the buffer can be re- duced. Assigning one owner to this review, and recording each revision with its rationale, prevents the parameters from drift- ing back to an informal flat rule.

    7. Positioning Against Prior Work

      Relative to the classical approach in Silver et al. (1998), the contribution here is not a new formula but a re-parameteriza- tion: the same safety-stock logic, applied with lead-time statis- tics that are allowed to change by phase. Relative to the agri- business risk literature (Behzadi et al., 2018; Shukla & Jharkha- ria, 2013), the contribution is an operating rule simple enough to be implemented without optimization software, together with an empirical estimate of its effects. Relative to resilience frameworks (Christopher & Peck, 2004; Tang, 2006), it offers one concrete, measurable instance of building resilience in ad- vance rather than responding after a disruption.

    8. Change Management and Adoption

      Technical correctness is not sufficient for adoption. Procure- ment officers who have lived through stockouts tend to distrust any proposal that lowers a buffer, and warehouse managers who have managed overflowing stores tend to distrust any pro- posal that raises one. Three practices ease the transition. First, present the phase table as an agreed, jointly owned plan rather than as an analyst's output, with procurement, warehouse, fi- nance and operations all signing it off. Second, define in ad- vance a small set of conditions under which planners may over- ride the tablefor instance, a confirmed transport strike or a weather eventand require overrides to be logged. Third, share the results of each phase openly, including cases where the buffer proved too high or too low, so that the table is treated as a living instrument that improves with evidence rather than as a rule to be defended.

  7. LIMITATIONS AND FUTURE RESEARCH DIREC- TIONS

    1. Limitations

      Several limitations qualify the findings and should be kept in view when applying them elsewhere.

      • Single enterprise and single season. The data cover 180 shipments over one 60-day cycle at one company. Lead- time behavior in other regions, crops or seasons may differ, and the phase boundaries and parameters would need to be re-estimated.

      • Before-and-after comparison. The static and dynamic policies are compared across periods rather than through a controlled experiment, so differences in crop conditions or market arrivals cannot be fully excluded as contributing factors (Section 3.6).

      • Reliance on the maximum. Eq. (4) shows that the reorder point is anchored to the longest observed lead time in a phase. A single extreme delivery can inflate the buffer, and a phase without extreme deliveries can understate the risk of the next one. Percentile-based or distribution-based esti- mates would be more stable.

      • Aggregate lead time. Lead time is analyzed in aggregate, so the contribution of each of the four stages to the observed variance is not separately quantified.

    2. Future Scope of Work

      While this study successfully models dynamic inventory buff- ering across a 60-day agricultural harvesting cycle, several av- enues remain open for future academic and applied research:

      • Predictive Analytics and Machine Learning: Future op- timizations should focus on integrating machine learning forecasting algorithms and satellite Normalized Difference Vegetation Index (NDVI) crop health data to proactively forecast farm-gate arrival volumes and yield quality before harvests begin.

      • Cold-Chain Integration: The current mathematical mod- els assume semi-perishable grains and pulses; extending this framework to highly perishable horticultural crops (such as berries, fruits, and leafy greens) will require incor- porating thermal degradation variables and refrigerated transit tracking into the inventory matrix.

      • Sustainable Green Logistics: As regional sourcing net- works expand, future studies should evaluate the feasibility of deploying battery-electric commercial trucks (e-LCVs) for short-radius rural aggregation milk-runs to lower carbon emissions and align with corporate sustainability mandates.

      • Multi-Season and Multi-Site Validation: Repeating the study over several harvest seasons and at several plants would test whether the phase structure and the parameters are stable, and would allow formal statistical comparison between policies.

      • Probabilistic Service-Level Formulation: Replacing the worst-case rule with a percentile or service-level rule, pa- rameterized from the full lead-time distribution, would make the trade-off between protection and holding cost ex- plicit and tunable.

      • Stage-Level Lead-Time Analysis: Measuring each of the four lead-time stages separately would show which stage drives late-season variance and where supplier or dock in- vestments would yield the largest reduction in buffer re- quirements.

      • Joint Optimization with Price and Storage Costs: Ex- tending the model to include price volatility, storage capac- ity constraints and financing costs would allow buffer siz- ing to be optimized against total cost rather than service alone.

        enterprises can effectively absorb lead-time variance (L) with- out inflating inventory holding costs. Empirical validation across 180 shipments proves that aligning inventory reorder points (ROP) with seasonal lead-time behavior entirely elimi- nates factory stockout downtime, reduces the average inventory holding footprint by 38.3% (from 300 MT to 185 MT), releases

        34.50 Lakhs in blocked working capital, and cuts warehouse spoilage by 71.1%.

        Three lessons follow for practice. First, the average lead time is a poor guide to risk; the tail of the distribution, which widens as the season advances, is what stops production. Second, the policy that responds to this is simplea three-row table of phase-specific buffers and reorder pointsand can be admin- istered in a spreadsheet or an ERP parameter table with no new technology. Third, the benefits are not limited to service: the same schedule shortens the average residence time of stock, frees working capital and reduces biological losses, so that pro- tection of the plant and efficiency of the warehouse are pursued together instead of traded against each other.

        The sensitivity analysis shows that these conclusions are robust to moderate changes in demand and price, and the cross-check against a classical service-level formulation suggests that the dynamic buffer provides protection broadly in line with a high service level. At the same time, the evidence comes from one enterprise and one season. Replication across seasons, sites and crops, together with a probabilistic refinement of the worst-case rule, is the natural next step. Until then, the study offers agro- processors a practical, quantified and transparent alternative to the flat buffer: let the cushion rise and fall with the risk.

        APPENDIX A. NOTATION AND ABBREVIATIONS

        Symbol / term

        Meaning

        D

        Daily plant processing demand, fixed by capacity (100 MT/day)

        LT

        Inbound lead time from order to availability at the factory

        LTavg, LTmax

        Average and maximum observed lead time within a harvest phase

        LT

        Lead-time delay, LTmax LTavg

        L

        Standard deviation (reported deviation) of lead time

        SB

        Dynamic safety buffer in MT

        ROP

        Reorder point: stock level at which a replenishment order is triggered

        z

        Standard normal deviate for a chosen service level

        MT

        Metric ton

        Lakh

        Rupees one hundred thousand

        VCC / FPO / VLA

        Village Collection Center / Farmer Producer Organi- zation / Village-Level Aggregator

        APMC

        Agricultural Produce Market Committee

        ERP / MRP / PO

        Enterprise resource planning / material requirements planning / purchase order

        TCO

        Total Cost of Ownership

        Table 11. Symbols and abbreviations used in the paper

  8. CONCLUSION

Agricultural inbound logistics inherently operates in an envi- ronment of high stochastic friction, where lead times fluctuate due to rural aggregation bottlenecks, transit delays, and pro- cessing plant queues. This study demonstrates that traditional, rigid inventory policiessuch as maintaining a flat, static safety buffer year-roundare structurally inadequate for man- aging seasonal harvesting cycles.

By replacing static inventories with a Dynamic Seasonal Safety Buffer Model that scales safety stock from 100 MT in the early harvest to 230 MT in the late harvest, agro-processing

APPENDIX B. PLANNER'S WORKSHEET

The calculation in Section 4 can be reproduced in a spreadsheet with the layout in Table 12. Each row of the phase table holds the inputs; the safety buffer and reorder point are computed from them, and the final check confirms the identity in Eq. (4).

Table 12. Suggested worksheet layout for the phase parameter table

Field

Source or formula

Example (late phase)

Phase and dates

Crop calendar

Late, days 4660

D (MT/day)

Plant capacity

100

LTavg (days)

Mean of shipment lead times in the phase

2.7

LTmax (days)

Longest shipment lead time in the phase

5.0

LT (days)

LTmax LTavg

2.3

Safety buffer (MT)

D × LT

230

ROP (MT)

D × LTavg + buffer

500

heck

ROP = D × LTmax

100 × 5.0 = 500

Example values are those reported by the study for the late harvest phase.

REFERENCES

  1. Ahumada, O., & Villalobos, J. R. (2009). Application of planning models in the agri-food supply chain: A review. European Journal of Operational Research, 196(1), 120.

  2. Aramyan, L. H., Oude Lansink, A. G. J. M., van der Vorst, J. G. A. J., & van Kooten,

    O. (2007). Performance measurement in agri-food supply chains: A case study. Sup- ply Chain Management: An International Journal, 12(4), 304315.

  3. Behzadi, G., O'Sullivan, M. J., Olsen, T. L., & Zhang, A. (2018). Agribusiness sup- ply chain risk management: A review of quantitative models. Omega, 79, 2142. https://doi.org/10.1016/j.omega.2017.07.005

  4. Chopra, S., & Meindl, P. (2016). Supply Chain Management: Strategy, Planning, and Operation (6th ed.). Pearson Education.

  5. Christopher, M., & Peck, H. (2004). Building the resilient supply chain. The Inter- national Journal of Logistics Management, 15(2), 114.

  6. De Treville, S., Shapiro, R. D., & Hameri, A. P. (2004). From supply chain to de- mand chain: The role of lead time reduction in improving demand-chain perfor- mance. Journal of Operations Management, 21(6), 613627. https://doi.org/10.1016/j.jom.2003.10.001

  7. Eppen, G. D., & Martin, R. K. (1988). Determining safety stock in the presence of stochastic lead time and demand. Management Science, 34(11), 13801390. https://doi.org/10.1287/mnsc.34.11.1380

  8. Fisher, M. L. (1997). What is the right supply chain for your product? Harvard Business Review, 75(2), 105116.

  9. Hopp, W. J., & Spearman, M. L. (2008). Factory Physics (3rd ed.). McGraw-Hill/Ir- win.

  10. Lee, H. L., Padmanabhan, V., & Whang, S. (1997). The bullwhip effect in supply chains. Sloan Management Review, 38(3), 93102.

  11. Nahmias, S. (1982). Perishable inventory theory: A review. Operations Research, 30(4), 680708.

  12. Shukla, M., & Jharkharia, S. (2013). Agri-fresh produce supply chain management: A state-of-the-art literature review. International Journal of Operations & Produc- tion Management, 33(2), 114158. https://doi.org/10.1108/01443571311295608

  13. Silver, E. A., Pyke, D. F., & Peterson, R. (1998). Inventory Management and Pro- duction Planning and Scheduling (3rd ed.). John Wiley & Sons.

  14. Tang, C. S. (2006). Perspectives in supply chain risk management. International Journal of Production Economics, 103(2), 451488.

  15. Zipkin, P. H. (2000). Foundations of Inventory Management. McGraw-Hill.