Master planning (MRP) module in Dynamics 365 F&O and Infor LN

Dynamics 365 F&O and Infor LN size replenishment orders almost identically. The difference is one F&O setting. The coverage time fence costs about half a point of service for every day it is set below the lead time, and raises no error. CodeCore Dynamics LLC. 9 September 2026. Summary Master planning decides how much of the future to plan and at what resolution. The whole horizon at full detail returns proposals for orders that need no decision for months. Too short a horizon and long lead time items arrive late. The two products agree almost entirely on lot sizing. One order per requirement, a floating time window and replenish-to-maximum exist in both. The window opens at the first demand in each, and neither product anchors it to the calendar. LN adds Fixed Order Quantity and Economic Order Quantity. F&O adds priority-based replenishment and a decoupling point. Neither product contains the other's full set. The setting to tune is how much demand one order covers. It is a working capital decision, not a service one. Doubling the window removes 39 per cent of the purchase orders and adds 126 per cent to the stock, at an unchanged fill rate. The right window length differs by business. 56 days suited the spare parts operation, 7 and 14 days the short lead time cases. Five business cases were built from two years of a UK wholesaler's order lines, and eleven lot-sizing settings ran in each. The floating window came top in all five. The two methods only LN documents never placed better than third of eleven. Four of the five cases were ties, decided by 0.13 to 2.81 points on a hundred point scale. Only the spare parts case separated, at 9.93 points, and it went to a setting both products offer. Service barely separates the time-phased rules at short lead times. Fill rate spread by under 0.2 points in three of the five cases, and by 2.8 and 4.3 points where the lead time reached 60 and 45 days. LN's coarse planning tier is optional, set per item, and Infor advises against maintaining it unless its extra functionality is needed. An LN item planned only by order planning behaves as an F&O item does. The coverage time fence is how many days ahead master planning reads supply and demand. Demand beyond it is not planned coarsely. It is not read, and nothing reports it. Microsoft's rule, that the fence exceed the total lead time, is correct, and ten days above it recovers almost all of the loss. Nothing here supports a migration from F&O to Infor LN on planning grounds. For Finance Leadership The current position can be established without opening the system. Three questions to the planning team answer it. Ask Warning flag Potential costs What is the coverage time fence, in days? A number at or below the longest purchasing lead time behind those items Roughly half a point of fill for every day it falls short, with no error raised Which items use the Period coverage code, and what period does each carry? One value applied across the whole catalogue The wrong stock and the wrong order count for whichever part of the catalogue it does not suit What review period was the demand classification measured in? Nobody knows, or it arrived with a reporting tool Coverage codes assigned from a class that would change if it were measured weekly instead of monthly The second answer converts to money from two figures already on the books. Moving the coverage period from 7 days to 28 days removed roughly two thirds of the purchase orders and roughly quadrupled the stock, at an unchanged fill rate. Multiply the orders saved by the fully loaded cost of raising one, and the stock added by unit cost and the holding rate. Those two inputs are yours and are not supplied here. Nothing in this article supports a migration to Infor LN on planning grounds. The two lot-sizing methods F&O lacks were tested. Neither improved on the coverage codes F&O already ships. Terms Dynamics 365 F&O Term Meaning Master planning The calculation that reads demand and supply and proposes orders Planning Optimization The current master planning engine, run as a service outside the application Coverage code The lot-sizing method for an item. Period, Requirement, Min./Max., Priority, Decoupling point or Manual Coverage period The length in days that one order covers under the Period coverage code Coverage time fence How many days ahead supply and demand are read Item coverage, coverage group Where coverage settings are held, the first overriding the second Infor LN Term Meaning Enterprise Planning The module holding LN's planning, both order-based and bucketed Order planning Planning recorded on a second by second basis, producing planned orders Master planning In LN this term means the optional bucketed tier only Order method The lot-sizing rule for a plan item Order interval The time window requirements are grouped into Item master plan The optional per-item bucketed plan, holding a supply plan by plan period Order horizon The period, in working days, that order based planning covers Measurement Term Meaning Fill rate The share of demanded units served on the day they were required Days of cover Average stock held, divided by average daily demand Balanced score The mean of the categories rankable in a case, on a 0 to 100 scale within that case 1. Introduction 1.1 What the Module Does Master planning, or MRP, answers one question on a schedule. Given what is on hand, what is already on order and what is expected to be demanded, what should be ordered and when. It runs nightly in most implementations, discards the previous set of proposals and rebuilds them from current data. It returns planned orders for a planner to review, together with messages proposing that existing orders be advanced, postponed or resized. Read the entire future at full detail and the run returns thousands of proposals for orders that need no decision for months. Read only the near future and every item with a longer lead time than that window arrives late. 1.2 The Two Products Dynamics 365 Finance and Operations is Microsoft's enterprise ERP. Master planning sits in its supply chain module, and the current engine, Planning Optimization, runs as a service outside the application. A coverage code sets the lot-sizing rule for an item. Time fences bound the horizon. F&O plans everything inside that horizon at daily resolution. Infor LN is Infor's ERP for discrete manufacturing, where planning sits in Enterprise Planning. Every plan item is planned by order planning. Infor calls it "roughly comparable to traditional MRP planning". It records data "on a second-by-second basis" and produces planned orders. An item may additionally carry an item master plan, "roughly comparable to traditional MPS planning". It records data "in terms of time buckets (plan periods)" and produces a supply plan by period instead of orders. That second tier is a per-item choice and not the default. It is switched on by the Maintain Master Plan check box. Infor advises against it unless it is wanted, since item master plans "have an adverse effect on system performance. Therefore, you should only maintain an item master plan if you need specific master-plan functionality." 1.3 The Boundary flowchart LR subgraph FO["Dynamics 365 F&O"] direction TB A1[Today] --> A3[Planned at full daily resolutioncoverage code sets the lot size] A3 --> A4{{Coverage time fence}} A4 --> A5[Demand here is not read.No requirement transaction.No message.] end subgraph LN["Infor LN"] direction TB B1[Today] --> B3[Order planningplanned orders on exact dates,all materials and capacity] B3 --> B4{{Order horizon}} B4 --> B5{Maintain Master Planset for this item?} B5 -- No --> B7[Planning horizon equalsthe order horizon.Nothing further is planned.] B5 -- Yes --> B6[Master planningsupply plan by plan period,critical materials only,infinite capacity by default] B6 --> B8{{Planning horizon}} end classDef fine fill:#dcecdc,stroke:#3a7d3a,color:#0b0b0b classDef coarse fill:#fbe8cd,stroke:#b8761c,color:#0b0b0b classDef gone fill:#f6d8d4,stroke:#b0443a,color:#0b0b0b classDef bound fill:#ffffff,stroke:#52514e,color:#0b0b0b classDef dec fill:#e8eef6,stroke:#2a78d6,color:#0b0b0b class A3,B3 fine class B6 coarse class A5,B7 gone class A4,B4,B8 bound class B5 dec Figure 1. Colour marks the resolution demand is planned at. Green is full resolution, amber a bucketed plan, red a band that is not planned, blue a configuration choice. The two differ at the boundary. F&O excludes the far horizon, and Microsoft states both the intent and the mechanism. The coverage time fence prevents "noise" caused by supply suggestions that don't require attention for months, and beyond it "the system doesn't generate requirement transactions for any supply and demand that falls outside the coverage time fence." An LN item carrying a master plan lowers the resolution instead, in plan periods that may be short "for the immediate future and longer periods for longer-term planning". An LN item without one stops at its planning horizon, as F&O does. 2. Module Functionality 2.1 Lot Sizing flowchart TD ROOT[Lot-sizing methods] --> B[Same mechanism in both products] ROOT --> F["Dynamics 365 F&O only"] ROOT --> L[Infor LN only] B --> B1["Requirement = Lot-for-Lotone order per requirement"] B --> B2["Period = Order intervala window that opens at demand;neither is anchored to the calendar"] B --> B3["Min./Max. = Replenish to Maximumrefill when projected stock falls below a floor"] F --> F1[Priority-based replenishment] F --> F2[Decoupling point, DDMRP] L --> L1[Fixed Order Quantity] L --> L2[Economic Order Quantity] classDef both fill:#dcecdc,stroke:#3a7d3a,color:#0b0b0b classDef fno fill:#e8eef6,stroke:#2a78d6,color:#0b0b0b classDef ln fill:#fbe8cd,stroke:#b8761c,color:#0b0b0b classDef root fill:#ffffff,stroke:#52514e,color:#0b0b0b class B,B1,B2,B3 both class F,F1,F2 fno class L,L1,L2 ln class ROOT root Figure 2. Section 4 tests all five of these rules against each other. Lot-sizing mechanism Dynamics 365 F&O Infor LN One order per requirement Requirement coverage code Lot-for-Lot order method Group demand in a floating window Period coverage code. "The period starts with the first demand of the item" Order interval, "measured starting at the last generated order" Replenish to a maximum Min./Max., triggered below a minimum Replenish to Maximum Inventory, triggered below the inventory plan or safety stock Fixed order quantity No coverage code Fixed Order Quantity order method Economic order quantity No coverage code Economic Order Quantity, setting quantities "to at least the economic order quantity" Priority-based replenishment Priority coverage code Not established DDMRP decoupling buffers Decoupling point coverage code Not established. Allocation buffers are specification pegging, not DDMRP Order multiple, minimum, maximum Multiple, Min. and Max. order quantity on Default order settings Order Quantity Increment, Minimum, Maximum No automatic proposal Manual coverage code Order horizon of zero Table 1. The first five rows are tested in Section 4. 2.2 Capabilities Capability Dynamics 365 F&O Infor LN Optional bucketed planning tier None. One resolution across the horizon An item master plan, per item, holding a supply plan by plan period Plan periods of varying length No equivalent Shorter periods "for the immediate future and longer periods for longer-term planning". A period's quantity is spread across its days "proportional to the capacity on that day" rather than placed at a point Far horizon beyond the boundary Excluded from the calculation beyond the coverage time fence No supply plan or planned order generated beyond the planning horizon Capacity in the coarse tier Not applicable Infinite planning by default, taking no constraints into account. Workload control is the constrained alternative Finite capacity scheduling Supported Order planning takes all necessary materials and capacity into account Multi-level BOM explosion Supported, including scrap, subcontracting, co-products and yield Critical components only inside the master-planning horizon; all components once demand enters the order horizon Safety stock Minimum coverage, with pegging options per coverage group Safety stock on the plan item, optionally with a seasonal pattern Time fences Coverage, freeze, capacity Order horizon, planning horizon, forecast time fence and a time fence freezing the supply plan, all rounded to plan period ends Exception and action messages Action messages, futures messages, calculated delays Exception messages by type, resource, planner and item Simulation Multiple master plans Scenarios, one designated the actual scenario Capable and available to promise CTP supported from version 10.0.28 Standard ATP and component CTP without a master plan; ATP and CTP with one Multi-site and intercompany Forecast and downstream demand supported, cross-entity execution not yet Planning clusters Demand by sales channel No direct equivalent Channel master plans holding forecast, allowed demand and channel ATP Return orders Not considered by Planning Optimization Not established. LN documents sales and purchase return orders, but no planning page connects them to Enterprise Planning Table 2. Documented contrast. Planning Optimization lists finite capacity scheduling, freeze time fences, auto firming, kanban, item substitution, subcontracting, co-products and formula yield as supported. 2.3 Documented Gaps in Planning Optimization Not supported Status per Microsoft Return orders "Planning Optimization doesn't consider return orders" Freeze time fence set on item coverage Pending, and currently ignored BOM and formula lines with step consumption Pending, and currently ignored Sales line reservation using explosion Future wave Intercompany planning execution across legal entities Future wave Requirement types for skills, courses, certificates and titles Future wave Table 3. The whole of the documented gap, from the Planning Optimization fit analysis. 3. Data and Method 3.1 Sources Every number in this article comes from two public files. Demand comes from Online Retail II. It is the transaction file of a UK registered non-store online retailer selling giftware, largely to wholesale customers, published through the UCI Machine Learning Repository under CC BY 4.0. 1,067,371 rows, 1 December 2009 to 9 December 2011, across 5,131 stock codes. Each row is an invoice line carrying stock code, quantity, unit price and a timestamp. The row count matches the repository's own figure. ONS series J596 is used for one check and does not enter the results. It is the Retail Sales Index for non-store retailing, value not seasonally adjusted. The seasonally adjusted version of the same series was downloaded and deliberately not used, since the check turns on whether the seasonal pattern matches. Public demand history is rare. This is a distributor's order book, not a manufacturer's. Limits says what that rules out. 3.2 Filters Applied Stock codes matching five digits with an optional letter suffix are retained. That excludes 61 non-item codes covering postage, adjustments, bank charges and samples. Credit invoices, and the 3,457 negative quantities outside them, are excluded from demand. An item enters the replay only with 12 or more demand days and a 180 day active window. That leaves 3,498 of the 4,773 items carrying any demand. 3.3 Classification Items are classified on average demand interval and squared coefficient of variation of demand sizes, against cutoffs of 1.32 and 0.49 from Syntetos, Boylan and Croston. Those authors derived the cutoffs for selecting a forecasting method, not a lot-sizing rule. Measured weekly, the population divides into 120 smooth, 131 intermittent, 1,344 erratic and 1,903 lumpy items. 3.4 Constructed Inputs Three inputs are constructed rather than observed. Forecast error is generated, and Section 4.6 varies it. The five business cases are combinations of lead time, forecast error and item population, and no source claims they are typical. The opening stock, the tie margin and the remaining parameters are chosen values, listed in Limits. 3.5 Checks One firm's order book should move with the national series for its sector. Against the ONS Retail Sales Index for non-store retailing, value not seasonally adjusted, the correlation over 23 complete overlapping months is 0.858 in levels and 0.807 in logs. Both series peak in November. The firm's November runs at 1.75 times its own yearly mean, the sector's at 1.35. Structural check What it asserts What it catches Wagner and Whitin bound No lot-sizing rule may cost less than the exact optimum A rule that reports a cost below the true optimum Unit bucket identity A one day window must equal one order per requirement A window that groups demand it should not Zero-noise identity With a perfect forecast, a plan whose boundaries do not depend on demand must not move An order timed off its release date Reproducibility A rerun in a fresh process must reproduce stored results exactly A seed that is not stable across machines Table 4. Each check is an identity rather than a plausibility test, so a defect cannot satisfy one by chance. All four hold for the published run. 3.6 Business Cases and Configurations Case Lead time Forecast error Items drawn from Fast-moving distributor 5 days 10% Smooth and erratic Long-lead importer 60 days 20% All classes Spare parts operation 45 days 30% Intermittent and lumpy Seasonal wholesaler 20 days 20% All classes Volatile demand 14 days 40% Erratic and lumpy Table 5. Each case replans weekly over 104 runs, after a warm-up that is discarded. No coverage time fence is set in any of them. Eleven configurations run in each case. Six are shared: one order per requirement, the floating window at 7, 14, 28 and 56 days, and replenish-to-maximum. Five are LN only: Fixed Order Quantity at 7 and 28 days of mean demand, and Economic Order Quantity at three setup-to-holding ratios. That ratio is a property of the business and is not in the data, so three values are swept. The replenish-to-maximum minimum is set to cover the lead time. A minimum below the lead time cannot serve demand during replenishment. Each configuration is scored on three outcomes, normalised from 0 to 100 against the best and worst achieved in that case. Service comes from fill rate, working capital from days of cover, planner workload from purchase orders raised per year. Scores are comparable within a case and not across cases. 4. Results 4.1 Service Figure 3. The spread the eleven settings produce within each case, on a log scale. The service bars span all eleven, replenish-to-maximum included. Among the time-phased rules alone the spread is a fraction of a point, as the numbers below give. Among the time-phased rules the fill rate spread is 0.00, 0.19 and 0.04 points in the three short lead time cases. It opens to 2.84 points at a 60 day lead time and 4.30 at 45 days with intermittent demand. Below those lead times the choice of rule is not a service decision. Replenish-to-maximum is the exception. It served between 82.9 and 88.9 per cent across the five cases, where every time-phased rule served above 94 per cent. At short lead times it was dominated outright. In the fast-moving case it held 15.3 days of cover to serve 88.9 per cent, where a 7 day window held 5.5 days to serve 100. A reorder point cannot use forward visibility, and on this demand that costs more than it saves. Wemmerlov reported the same in 1989: "when demands cannot be predicted, the choice of lot-sizing technique is not a very important issue", qualified to hold "when the procedures are applied under equal conditions, and when only holding and ordering costs are considered." 4.2 Scores by Business Case Figure 4. Balanced score within each case. A solid box is a clear winner and a dashed pair is a tie. Orange labels mark the methods only LN documents. Business case Working capital Planner workload Balanced Margin Fast-moving distributor One per requirement 56 day window 7 day window 0.50 Long-lead importer Replenish to max Replenish to max 56 day window 2.81 Spare parts operation Replenish to max 56 day window 56 day window 9.93 Seasonal wholesaler One per requirement 56 day window 7 day window 0.13 Volatile demand One per requirement 56 day window 14 day window 1.02 Table 6. Margin is the gap between first and second. Anything under three points is reported as a tie. The floating window, F&O's Period coverage code and LN's order interval, comes top in all five cases. Four of those are ties, at 0.50, 2.81, 0.13 and 1.02 points. The spare parts case separates, at 9.93. The two methods only LN documents never placed better than third of eleven. Fixed Order Quantity reached third in three cases and fourth and fifth in the others. Economic Order Quantity placed lower again in every case. It took the service category in three cases, but only at the lowest setup-to-holding ratio tested. At that ratio it approaches one order per requirement and pays for the service in order count. Two columns follow from the definitions rather than the results. The longest window wins planner workload because it raises the fewest orders. One order per requirement wins working capital because it holds the least stock. They are the two ends of one trade. The right window length differs by business. 56 days for the two long lead time cases, 7 and 14 days for the three short ones. 4.3 Orders Against Stock Figure 5. The floating window sits below and left of both LN-only methods at every length tested. At matched stock each raises more orders, and at matched order count each holds more stock. Across the 795 item-configurations where both outcomes are defined, doubling the window changes purchase orders by −39.1 per cent and days of cover by +125.9 per cent. Simple theory predicts a halving and a doubling. Orders fall by less than half because demand at this granularity is lumpy, so many buckets sit empty. Fixed order quantity and economic order quantity sit outside that trade at every setting tested. Economic order quantity sits further out than fixed order quantity. LN sets order quantities "to at least the economic order quantity", so it raises small requirements to the EOQ and leaves large ones alone. On lumpy demand that adds orders without removing stock. 4.4 Demand Classification and the Review Period Review bucket Smooth Intermittent Erratic Lumpy Daily 0.0% 4.5% 1.5% 94.0% Weekly 3.4% 3.7% 38.4% 54.4% Monthly 22.5% 2.7% 56.0% 18.7% Table 7. The same 3,498 items in every row. Only the period the measurement was taken in changed. Figure 6. The source paper defines the interval as "expressed in number of forecast review periods", so the measure carries the review period inside it. Demand classifications are used to assign coverage codes. No item is smooth or lumpy on its own account. It is smooth or lumpy relative to a review period. Neither product names that period. The tool that produced the classification chose it. 4.5 Coverage Time Fence Every setting above trades one outcome against another. The coverage time fence withholds demand from the calculation instead. Figure 7. Four window lengths at a 60 day lead time. All four recover through the same fence. The dotted markers show where the transition would sit if the window length governed it. Microsoft asks that the fence be "longer than the total lead time", and that rule is correct. Taking the 14 day window, a 60 day fence returned 89.6 per cent at a 60 day lead time, 65 days returned 94.9 per cent and 70 days returned 97.5 per cent against a plateau of about 98.1. Ten days above the lead time therefore recovers almost all of the loss. The replan cadence here is weekly, and the demand has to be visible for at least one run before the order is released. The window length does not govern where service returns. Swept in five day steps at windows of 7, 14, 28 and 56 days, the four curves are indistinguishable through the transition, and the smallest fence within half a point of the best fill runs from 70 days at a 7 day window to 85 days at a 56 day window. Were the window length in control, those would fan out across the full 49 day span of the windows. Across the 5,880 item-scenarios, each day the fence falls short of the lead time costs 0.451 percentage points of fill, at a robust t of −34.7. 4.6 Sensitivity to Forecast Error Figure 8. Fill rate, order count and the order elasticity are drawn on scales wide enough to show movement and do not move. Plan stability, in red, moves across the whole range. Forecast error is a parameter, not a property of the data. Its size was varied from 0 to 40 per cent and its persistence between runs across 0, 0.5 and 0.8. The three outcomes behind the scores held throughout. Fill stayed between 99.93 and 99.99 per cent, orders between 11.95 and 12.11 a year, and the order elasticity between −39.1 and −40.6 per cent. One result moved in the direction the literature predicts. Days of cover rose from 11.73 to 14.33 across the sweep. Wemmerlov reports that forecast errors "not only lead to stockouts, they also induce larger inventories." Plan stability was measured throughout. No claim is made from it. Counting a planned order as changed by a fixed number of units, and by a fixed percentage, point in opposite directions. That is a property of the measure and not of the products. Only the physical outcomes are reported. 5. Discussion 5.1 Where Demand Is Dropped flowchart TD D[Demand line, due date D] --> Q{Is D within thecoverage time fence?} Q -- No --> X[Not loaded into the engine.No requirement transaction.No message.] Q -- Yes --> C{Coverage code} C -- Requirement --> R[One planned order per requirement] C -- Period --> P[Window of L days opensat the first demand] C -- Min/Max --> M[Replenish to maximumwhen below minimum] R --> L{Is D minus lead timein the past?} P --> L M --> L L -- No --> OK[Planned order released on time] L -- Yes --> DEL[Delay. Order placed at theearliest feasible date] X -.-> LATE[Demand becomes visible onlywhen it crosses the fence] LATE --> DEL classDef ok fill:#dcecdc,stroke:#3a7d3a,color:#0b0b0b classDef warn fill:#fbe8cd,stroke:#b8761c,color:#0b0b0b classDef gap fill:#f6d8d4,stroke:#b0443a,color:#0b0b0b classDef dec fill:#e8eef6,stroke:#2a78d6,color:#0b0b0b class OK ok class DEL,LATE warn class X gap class Q,C,L dec Figure 9. Blue is a decision the engine takes, green a demand served on time, amber one served late, red the path where the demand is never seen. Observations from the diagram: The red path raises nothing. No requirement transaction, so no action message, no futures message and no delay. The demand stays invisible until it crosses the fence. The fence is measured from today, not from the item's lead time, and in calendar days. A week is seven days regardless of the working time calendar. An item with a 60 day lead time behind a 30 day fence is a configuration no validation prevents. 5.2 Where the Fence Resolves flowchart TD S[Master planning run starts] --> M{Coverage time fenceset on the master plan?} M -- Yes --> MP[Master plan value wins.Item and group are ignored.] M -- No --> I{Override set onitem coverage?} I -- Yes --> IC[Item coverage value wins] I -- No --> G{Set on thecoverage group?} G -- Yes --> CG[Coverage group value wins] G -- No --> DF[Global default, 100 days] classDef win fill:#dcecdc,stroke:#3a7d3a,color:#0b0b0b classDef trap fill:#fbe8cd,stroke:#b8761c,color:#0b0b0b classDef dec fill:#e8eef6,stroke:#2a78d6,color:#0b0b0b class IC,CG,DF win class MP trap class M,I,G dec Figure 10. Amber marks the path that defeats an audit carried out at item level. Planning Optimization "first checks the item coverage line, then the coverage group, and finally the global coverage time fence", and a value on the master plan overrides all three. A review has to start at the master plan and work down. 5.3 Configuration Find the exposure first. It is a report against setup tables and needs no transaction history. For each item, resolve the coverage time fence the override chain returns, by working down Figure 10. The master plan wins if it carries a value, then item coverage, then the coverage group, then the 100 day default. Compare that against the item's total lead time. Every item whose fence does not clear its lead time is exposed. Rank them by lead time, longest first, since the loss grows with it. Two settings decide the outcome. Both sit in the same two places. Setting Where What to put in it Coverage time fence Master planning > Setup > Coverage > Coverage groups, General tab. Item coverage for a single item Above the longest total lead time behind the group, with room to spare. Ten days above recovered almost all of the loss in these tests, and the 100 day default already clears every lead time tested here Coverage period The same two places By business, not by policy. 56 days suited the spare parts case, 7 or 14 days the short lead time ones. One group-wide value across a mixed catalogue will be wrong for part of it Set the fence first. Lengthening it trades nothing away, and the exposure is created by shortening it. The coverage period trades constantly, at 39 per cent of the orders against 126 per cent of the stock per doubling. That is a treasury decision, not a planning one. Planning Optimization does not consider return orders. An expected return will not reduce a net requirement and the replenishment goes out regardless. Where return volumes are material, that stock has to reach the plan by another route. Where a demand classification is used to assign coverage codes, record the review period it was measured in. Section 4.4 shows the class is not a property of the item alone. 5.4 For Microsoft Request Case for it Evidence behind it Validate the fence against the lead time A fence shorter than the total lead time cannot serve the demand behind it, and the system already holds every number needed to say so Measured. Section 4.5 prices the shortfall Consider return orders in the planning calculation A documented gap with a direct operational consequence Microsoft's own fit analysis A coarse far horizon Seeing a long lead time item's far demand today means admitting it at full daily resolution. That is the noise the fence exists to suppress None The first costs little and the evidence is in Section 4.5. The second is Microsoft's own documented gap rather than an inference from anything here. The third carries no measurement. LN's optional item master plan puts a coarse plan beyond the order horizon where F&O has nothing. Its purpose is visibility, not replenishment, so replaying independent demand cannot price it. Infor advises against maintaining it unless the functionality is wanted. Fixed Order Quantity and Economic Order Quantity should not be requested. On this demand neither improved on the coverage codes F&O already ships. 5.5 What the Analysis Supports Claim Where it comes from The floating window tops every business case, and the two LN-only methods never place better than third Data Four of five cases are ties. The one that separates goes to a shared setting Data Doubling the window costs 39 per cent of orders and adds 126 per cent to stock Data Service barely separates time-phased rules below a 45 day lead time Data A fence short of the lead time costs 0.451 points of fill per day Data The demand classification moves with the review period it is measured in Data Which lot-sizing rules each product offers Documentation That LN's second tier is optional, and Infor advises against it by default Documentation Everything in Table 2 Documentation Table 8. Six findings from the data. The rest from the two vendors' published manuals. Neither an LN nor an F&O system was run in this modelling exercise. Both sets of rules were reimplemented from documentation. Therefore a behaviour that differs from the documentation is invisible to this method. Where Infor's manuals are silent, the assumption made is recorded in the repository. Limits This is one wholesaler's order book, with no bills of material, routings or capacity. Nothing here tests dependent demand, multi-level netting or capacity scheduling. That bounds the findings to purchasing and distribution. The lot-sizing literature places uncapacitated results there too. Wemmerlov notes that such work is "applicable to purchasing/distribution situations but not necessarily to decisions related to manufacturing lot sizes." The fence result is derived at a single level, and a multi-level bill would extend the requirement. No safety stock was modelled. F&O's minimum coverage is a standard setting any implementation carrying a 60 day lead time would use, so the fence results are the cost of the exclusion with nothing buffering it, and the magnitudes are an upper bound. The direction of the omission is known. Introducing safety stocks "generates even larger inventories and also more orders", so both the stock and the order counts reported here would rise. Forecast error is the assumption the plan stability results rest on. Section 4.6 reports the range they move across. No stability figure here should be read as a prediction for a live system. The structural checks cover the lot-sizing rules and not the whole replay, since the Wagner and Whitin bound applies to the rules alone. The fine sweep behind Figure 7 sits at a single 60 day lead time. The 0.451 point per day figure spans lead times of 5, 15, 30, 45, 60, 75 and 90 days against seven fences. The ten day recovery margin is close to the seven day replan cadence used throughout, so a different cadence would move it. The balanced score is the mean of the categories rankable in a case. It weights them equally, and that weighting is a judgement. Table 6 carries the per-category winners alongside it. Chosen parameters. Each moves some conclusion if moved. The 1.32 and 0.49 cutoffs, the weekly review period behind the classification, the activity filter of 12 demand days and a 180 day active window, the 28 day visibility window, the weekly replan cadence, opening stock of 14 days, the warm-up period, the three point tie margin, the minimum spreads for ranking a category, and the setup-to-holding ratios swept for economic order quantity. No organisation is accused of anything. The wholesaler is not named in the source and is not identified here. Neither Microsoft nor Infor is alleged to have documented anything incorrectly. Reproducing This The source code and datasets are at github.com/kingomnivore/mrp-fno-vs-ln. Clone it, pip install -r requirements.txt, then python run.py. Both data files are committed, the workbook through Git LFS, so it runs as cloned. The recipe below describes the same analysis independently of the code, for anyone who would rather rebuild it. 1. Get the data. Online Retail II from the UCI Machine Learning Repository, both sheets of the workbook. ONS series J596, value not seasonally adjusted, for the check in Section 3.5 only. 2. Aggregate. Reduce the order lines to daily demand per item. Drop non-item codes, credit lines and negative quantities, then apply the activity filter in Section 3.2. 3. Classify. Compute the demand interval and the squared coefficient of variation of demand sizes in the review period the business intends to plan in, and split on 1.32 and 0.49. 4. Implement the rules. Code the lot-sizing rules from both vendors' documentation, including the ones the product under test does not offer, giving each rule the same treatment where the documentation is silent. 5. Replay. Define business cases as combinations of lead time, forecast error and item population. Run every rule in every case, replanning weekly, and commit each order on its release date rather than its receipt date. 6. Score. Discard the warm-up. Score each category only where the underlying outcome differs across configurations, and report the margin alongside the rank. 7. Sweep. Vary the forecast error before reporting anything that depends on it. 8. Check. Run the four identities in Table 4 before reading any result. Chosen parameters. As named in Limits. On Method The analysis code and the initial draft were produced with Claude Code. The dataset and the configuration claims were verified against the sources listed below before publication. The code is public on the repository. Sources Dynamics 365 Supply Chain Management documentation Coverage settings Replenishment methods and quantity modification Coverage time fences Planning Optimization fit analysis Differences between Planning Optimization and the deprecated master planning engine Parameters not used by Planning Optimization Finite capacity planning and scheduling Infor LN Enterprise Planning documentation Master planning versus order planning Master planning, an overview Demand and Inventory Planning Infor LN Enterprise Planning, User Guide for Order Planning. Source for the order interval, the order methods and the plan period distribution rules Plan periods in Enterprise Planning Order interval Items - Planning (cprpd1100m000) Order horizon Allocation buffers Glossary for Enterprise Planning Analysis code and data, GitHub About Author Dean Fachrie is a Functional Analyst at CodeCore Dynamics LLC, working on Microsoft Dynamics 365 F&O architecture and enterprise system design. Configuring master planning, or evaluating Dynamics 365 F&O against another ERP? We help enterprise supply chain and finance teams set coverage codes and time fences so master planning reads the whole horizon it should, and size replenishment against the working capital the business will carry. Contact us or connect on LinkedIn for architecture reviews.