Dynamics 365 Master Planning & Optimization: Migration, Configuration & Best Practices

Master Planning Evolution

Dynamics 365 F&O has undergone significant changes to its planning engine over the past five years:

Phase 1: Native Master Planning Engine (2016–2021)

The original MPS (Master Planning Schedule) engine was built into the core Dynamics 365 database. It processed demand, safety stock calculations, and MRP logic synchronously. The engine was monolithic, not cloud-optimized, and struggled with scale beyond 100K plan lines.

Characteristics: Synchronous processing, single-threaded, integrated with D365 DB, prone to lock contention, limited reporting.

Phase 2: Planning Optimization (2021–Present)

Microsoft released Planning Optimization as a separate, cloud-native add-on service built on Azure optimization engines (notably Microsoft’s solver technologies). Planning Optimization runs asynchronously in the cloud, returns results to D365 via API, and scales to millions of SKUs and millions of plan lines.

Characteristics: Asynchronous, cloud-native, scalable, separated concern (planning logic lives in Planning Optimization; D365 manages master data), built-in scenarios, rapid innovation.

Current Status

The deprecated MPS engine still exists but is in no-fix mode. Microsoft committed to sunsetting it by end of 2024. All organizations using MPS must migrate to Planning Optimization by that date or lose planning functionality.

Planning Optimization Architecture

Planning Optimization is a SaaS service separate from your D365 instance. The architecture works as follows:

Data Flow & Synchronization

Step 1 – Master Data Sync: D365 pushes product master data, BOM structure, routing, lead times, demand forecasts, safety stock policies, and item coverage settings to Planning Optimization via APIs. This sync is near-real-time (within minutes of configuration change).

Step 2 – Plan Execution: You trigger a plan run (manually or scheduled via batch job). Planning Optimization receives the plan request, loads master data from cache, and executes the optimization algorithm.

Step 3 – Results Return: Planning Optimization returns planned purchase orders, manufacturing orders, and transfer orders. D365 inserts these into the ReqPO and ReqTrans tables. Plan generation time: typically 5–30 minutes depending on complexity (100K–1M+ plan lines take 20–60 minutes).

Step 4 – Firming & Release: Planners review action messages and firm planned orders manually or via automated rules. Firmed orders convert to purchase orders (POs) or work orders (WOs) for execution.

Planning Optimization vs. Deprecated MPS

Speed: Planning Optimization is 2–5x faster. A plan with 500K lines that took 4 hours in MPS now takes 30–60 minutes.

Scale: Planning Optimization scales to 2M+ plan lines. MPS struggled beyond 500K.

Optimization Quality: Planning Optimization uses advanced optimization algorithms that can consider multiple objectives (cost, service level, inventory) simultaneously. MPS used simpler sequential logic.

Scenarios: Planning Optimization supports multiple scenarios (Base, Competitive, Demand Surge) with independent parameter sets. MPS did not.

Real-Time Capabilities: Planning Optimization can be called via API for real-time planning (order promising, what-if analysis). MPS was batch-only.

Migration from Deprecated Master Planning

If you are currently running the deprecated MPS engine, migration to Planning Optimization is mandatory. The process is not automatic; it requires careful planning and testing.

Migration Approach

Phase 1 – Assessment (Weeks 1–4):

• Document current MPS configuration (all item coverage settings, safety stock policies, demand forecasts, scenarios).
• Identify custom MPS extensions, user exits, or table extensions that rely on MPS tables.
• Catalog all MPS reports and dashboards; identify Planning Optimization equivalents.
• Run sample MPS plans and archive results for comparison.
• Assess data quality in master data (item master, BOM, routing, lead times). Planning Optimization is less forgiving of data issues.

Phase 2 – Pilot Planning Optimization (Weeks 4–12):

• Enable Planning Optimization add-on in your sandbox environment.
• Replicate MPS configuration into Planning Optimization parameters (item coverage, safety stock, demand forecasts).
• Run Planning Optimization on the same date range as your archived MPS plans.
• Compare results: planned orders, quantities, timings. Expect 5–20% variance due to different algorithms. If variance is large (>30%), investigate configuration differences.
• Identify missing MPS customizations and decide: replicate via Planning Optimization API, or accept operational change.

Phase 3 – Parallel Running (Weeks 12–16):

• Run both MPS and Planning Optimization monthly for 1–2 months. Compare results weekly.
• Build planner confidence in Planning Optimization results; train planners on new UI and workflows.
• Identify any systematic differences (e.g., Planning Optimization is more aggressive on safety stock). Tune parameters to align.

Phase 4 – Cutover (Weeks 16–17):

• Disable MPS planning. Enable Planning Optimization as production planning engine.
• Archive final MPS plan; document cutover rationale.
• Monitor first two production planning cycles closely (daily standups with planning team).

Common Migration Pitfalls

Pitfall 1 – Data Quality Issues: Planning Optimization is less forgiving of invalid master data (missing item lead times, malformed BOMs, orphaned forecast lines). Conduct a data quality audit before migration. Fix issues in MPS first to establish baseline cleanliness.

Pitfall 2 – Custom MPS Extensions: If you built custom demand forecast adjustments, demand pattern recognition, or planning rules on top of MPS, those don’t automatically port to Planning Optimization. You must either replicate them via Planning Optimization APIs or accept the operational change. Budget 4–8 weeks for this work.

Pitfall 3 – Parameter Tuning: Safety stock factors, service level targets, and coverage policies may need tuning in Planning Optimization due to different algorithms. Don’t expect 1:1 parameter mapping. Run a 2–4 week parallel period to identify and adjust parameters.

Pitfall 4 – Scenario Management: If you used MPS scenarios for forecasting or what-if analysis, replicate this workflow in Planning Optimization. Planning Optimization supports scenarios natively but with different mechanics.

Demand Forecasting & Integration

Demand forecasting is a critical input to master planning. Inaccurate demand forecasts lead to either excess inventory or stockouts. Planning Optimization can work with demand forecasts from multiple sources: historical sales data (D365 built-in), external forecasting systems (Demand Planning app, Azure Synapse), or manual adjustments by sales and marketing.

Demand Forecast Sources

Option 1 – Built-in Demand Forecasting (D365 Forecast Module): D365 provides basic statistical forecasting based on historical sales. It computes moving average, exponential smoothing, and seasonal decomposition. Accuracy typical: 60–75% for stable products.

Use this if: You have 2+ years of stable historical sales; demand patterns are reasonably predictable; you don’t have budget for advanced forecasting software.

Option 2 – Demand Planning App (Microsoft): Microsoft’s dedicated Demand Planning application provides advanced forecasting (machine learning-based). It ingests sales history, external data (weather, economic indicators, web traffic), and uses gradient-boosted models (similar to XGBoost) to predict demand. Accuracy typical: 75–90%.

Use this if: You have volatile or seasonal demand; multiple data sources (web traffic, promotions, competitor pricing); you want to blend AI-generated forecasts with human judgment.

Option 3 – External Forecasting System: Some organizations use Salesforce Demand Planning, SAP Integrated Business Planning, or custom Python/R forecasting models. D365 can ingest forecasts via import or API.

Use this if: You have significant investment in specialized forecasting software; complex supply chain (multi-echelon, complex dependencies); you want to maintain system-of-record separation.

Demand Planning App Integration with Planning Optimization

The Demand Planning app is the native Microsoft solution. Integration workflow:

• Sales History Import: Demand Planning ingests 2–5 years of historical sales data from D365 sales orders and invoices.
• Forecast Generation: Machine learning models generate monthly or weekly demand forecasts.
• Forecast Review & Adjustment: Sales and marketing teams review AI-generated forecasts and adjust for promotions, new customers, or market events.
• Demand Forecast Push: Approved forecasts are pushed to D365 (Forecast module) and automatically synchronized to Planning Optimization.
• Plan Execution: Planning Optimization reads forecasts and generates supply plans.

Key Demand Forecasting Parameters

Forecast Bucket: Time period for demand aggregation: Day, Week, or Month. Daily forecasts provide more granularity but are noisier. Weekly or Monthly forecasts are more stable for planning.

Demand Time Fence: Period (e.g., 30 days) during which actual sales orders take precedence over forecasts. If you have a customer PO for Week 3, it overrides the statistical forecast for Week 3.

Forecast Accuracy Measure: Typical metrics are Mean Absolute Percentage Error (MAPE) or Mean Absolute Error (MAE). Track accuracy by product family or service line. MAPE <20% is excellent; >40% is poor and signals need for process change.

Demand History Review: Quarterly, audit demand forecast accuracy against actual sales. Identify high-error forecasts (MAPE >50%) and investigate root causes. Was the forecast wrong, or did actual demand shift unexpectedly?

Supply Planning & Coverage Groups

Coverage groups are the core mechanism for controlling how Planning Optimization generates planned orders. A coverage group is a set of rules that defines: What demand signals trigger planning? At what inventory level do we order? How much lead time do we assume?

Coverage Group Settings

Period – Code: Time unit for planning (Period, Day, Week, Month). Most organizations use Month (or Day for very high-velocity products).

Coverage Time Fence – Days: Number of days into the future for which we generate planned orders. Example: 90 days. Planning Optimization will only generate planned orders if the need occurs within 90 days of today.

Demand Time Fence – Days: Period during which actual confirmed demand (sales orders) replaces forecasts. Example: 30 days. Any sales order within 30 days overrides the statistical forecast.

Include Safety Stock: Should Planning Optimization factor in safety stock when generating planned orders? Almost always Yes. If No, you run the risk of stockout when demand exceeds forecast.

Include Safety Lead Time: Should Planning Optimization assume additional lead time variability (safety lead time) when scheduling planned orders? Typically Yes if your suppliers have inconsistent lead times.

Planned Order Type: Should Planning Optimization generate Purchase Orders (POs), Manufacturing Orders (WOs), or Transfer Orders (TOs)? Typically configured per item or item family.

Order Quantity Method: How does Planning Optimization calculate lot size?

• Fixed Order Quantity: Always order the same quantity (e.g., 100 units). Simple but can lead to unbalanced inventory.
• Lot-for-Lot (LFL): Order exactly what is needed in each period. Most flexible; requires stable demand.
• Period Order Quantity (POQ): Order to cover demand for N periods (e.g., 3 months of demand). Reduces order frequency; increases carrying cost.
• Economic Order Quantity (EOQ): Order quantity that minimizes total cost (ordering + carrying cost). Complex; rarely used in modern organizations.

For most organizations, Lot-for-Lot or Period Order Quantity work best.

Item Coverage Configuration

Item coverage overrides allow you to deviate from coverage group defaults for specific items. Example: Your default coverage group uses Monthly buckets, but high-velocity items (like fasteners) need Daily planning.

Configure item coverage on the Product Master Data screen: Product → Plan → Item Coverage. Specify:

• Coverage group assignment
• Lead time (supplier + internal processing)
• Minimum order quantity
• Replenishment order quantity (fixed lot size if using Fixed Order Quantity method)
• Safety stock or safety lead time override

Safety Stock & Variability Factors

Safety stock is inventory held to protect against demand or supply uncertainty. It’s a key lever for balancing service level with carrying cost. Too little safety stock leads to stockouts; too much wastes cash in excess inventory.

Safety Stock Formula

The standard formula is:

Safety Stock (SS) = Service Level Factor × Average Demand × Lead Time Variability Factor

Service Level Factor (z-score): Reflects your target service level (probability of not stocking out in a given period). Common values:

• 95% service level = z-score of 1.65
• 97.5% service level = z-score of 1.96
• 99% service level = z-score of 2.33
• 99.9% service level = z-score of 3.09

Higher service levels require more safety stock (and higher carrying costs).

Average Demand: Mean daily or weekly demand. Planning Optimization calculates this from historical sales or forecasts.

Lead Time Variability Factor: Coefficient representing how much demand or supply variability exists. Formula:

LTVF = sqrt(Lead Time Days + Safety Lead Time Days)

This reflects the principle that longer lead times accumulate more variability.

Tuning Safety Stock Parameters

Step 1 – Measure Actual Variability: Over a recent 12-month period, calculate:

• Coefficient of Variation (CV) for demand: Standard Deviation of Daily Demand / Mean Daily Demand. High CV (>0.5) indicates volatile demand.
• Lead Time variance: Standard Deviation of actual supplier lead time.

Step 2 – Set Service Level Target: Business decision. Typically 95–99% depending on product criticality and margin.

Step 3 – Calculate Initial Safety Stock: Use formula above. Planning Optimization will apply this to every demand period.

Step 4 – Monitor Service Level & Stockout Frequency: After 2–3 months of planning with new parameters, calculate actual service level achieved:

Actual Service Level = (1 – (Units Short / Total Demand)) × 100%

If actual service level is lower than target, increase safety stock. If higher, reduce it.

Safety Lead Time

Safety lead time is additional lead time (in days) added to supplier lead time to account for variability. It reduces the risk of late orders.

Example: Supplier quotes 30 days lead time, but actual lead times range from 25–40 days. Add 7 days safety lead time (total lead time: 37 days). This ensures planned orders arrive before stockout risk.

Configure on item coverage: Product → Plan → Item Coverage → Safety Lead Time.

Action Messages & Futures

Action messages are planning signals that communicate to planners what changes are needed to the supply plan given new demand or changes in lead time / safety stock settings.

Types of Action Messages

New: Create a new planned order to meet demand.

Expedite: Move an existing planned order forward in time to meet accelerated demand.

Postpone: Move a planned order backward in time to reduce inventory investment when demand is lower than planned.

Increase: Increase the quantity of an existing planned order to meet additional demand.

Decrease: Decrease the quantity of an existing planned order because demand is lower than planned.

Cancel: Cancel a planned order that is no longer needed.

Action Message Thresholds

Planners don’t want to see action messages for every 1-unit change. Instead, configure thresholds to suppress trivial messages.

On the Coverage Group, set:

• Quantity Tolerance – Decrease %: Don’t send a “Decrease” message unless the quantity drops more than X% (e.g., 10%).
• Quantity Tolerance – Increase %: Don’t send an “Increase” message unless quantity increases more than X% (e.g., 10%).
• Time Fence for Action Messages – Days: Only send action messages for planned orders due within N days (e.g., 30 days). Ignore changes to distant future orders.

Properly tuned thresholds reduce planner noise and focus attention on material changes.

Futures (Intercompany Planned Orders)

“Futures” in Planning Optimization refers to planned transfer orders between legal entities or warehouses. When you have a multi-entity supply chain (e.g., Regional Distribution Center supplies local branch), Planning Optimization can generate transfer orders between entities.

Configuration: Link items via Intercompany Trade Relations; enable “Planned Transfer Orders” in coverage group.

Planned Order Firming

Planned orders are hypothetical; they become real (purchase orders or work orders) only when firmed. Firming is the manual or automatic process of committing a planned order.

Firming Methods

Manual Firming: Planner reviews planned orders in the Master Plan screen and clicks “Firm” for each one. Time-consuming but gives complete control.

Automatic Firming Rules: Planning Optimization can automatically firm planned orders based on rules:

• Firming Time Fence: All planned orders within N days of today are automatically firmed. Beyond N days, they remain planned.
• Firm to Lead Time + Horizon: Firm any planned order that is due within (Lead Time + Safety Lead Time). This ensures supply orders are placed far enough in advance.

Example: Item with 14-day lead time, 7-day safety lead time. Set Firming Time Fence to 21 days. All planned orders due within 21 days are automatically firmed into purchase orders.

Firming Parameters

On the Coverage Group, set:

• Firm Planned Orders: Yes/No. If Yes, planned orders are auto-firmed based on time fence.
• Firming Time Fence – Days: Number of days within which planned orders are automatically firmed (e.g., 30 days).

Review firmed orders weekly. Monitor whether firmed quantities and timings align with actual demand.

Intercompany Planning

If your organization has multiple legal entities (subsidiaries, regional branches, manufacturing plants), Planning Optimization can generate transfer orders between entities to meet demand.

Intercompany Trade Relations Setup

For each item that can be replenished from another entity:

1. In the Inventory module, create an Intercompany Trade Agreement between supplying entity and purchasing entity.
2. Specify item, lead time (transfer time), and replenishment order quantity.
3. In the source and destination item coverage settings, enable Planned Transfer Orders.

Planning Intercompany Orders

When Planning Optimization generates a plan, it considers both external suppliers and intercompany sources. It will generate a planned transfer order (PTO) if:

• Intercompany supply is cheaper than external supply, or
• Intercompany supply has a shorter lead time, or
• Intercompany supply is the only configured source.

Monitor intercompany transfer orders separately from external purchase orders. They have different approval workflows and logistics (transfer vs. external receipt).

Demand-Driven MRP (DDMRP)

Demand-Driven MRP (DDMRP) is a modern planning discipline that challenges traditional MRP logic. Rather than pushing supply based on forecasts, DDMRP “pulls” supply based on actual market demand. It simplifies planning by decoupling dependent demand and reducing the “bullwhip effect” (demand variability amplification upstream).

DDMRP Core Concepts

Decoupling Point: The point in the supply chain where you decouple forecast-driven planning from demand-driven planning. Example: In a Make-to-Order manufacturer, the decoupling point is the sales order. In a Make-to-Stock distributor, it’s finished goods inventory.

Buffer Levels (instead of Safety Stock): DDMRP replaces “safety stock” with a three-level buffer:

• Min Buffer (Green Zone): Minimum acceptable inventory. If inventory drops below Min, it’s a problem.
• Max Buffer (Red Zone): Reorder point. When inventory reaches Max, place a replenishment order.
• Top of Buffer (Yellow Zone): Target maximum inventory to hold.

Demand Sensing: Use real-time market signals (actual orders, point-of-sale data, web clicks) rather than statistical forecasts. Responds faster to demand shifts.

Planned Order Execution Profiles (POEPs): Instead of ordering in fixed cycles, order as soon as buffer is breached. Reduces inventory; increases order frequency.

DDMRP Benefits

• Faster Response to Demand Changes: Pull-based planning reacts within days; forecast-driven planning lags by weeks.
• Lower Inventory: Decoupling reduces safety stock because demand variability is localized to each decoupling point (not amplified upstream).
• Simpler Planning: Less reliance on statistical forecasting; more reliance on actual demand signals.
• Better Service Level: By focusing on decoupling points, buffer stock is positioned where it’s most needed.

DDMRP Implementation in Planning Optimization

Planning Optimization doesn’t have a “DDMRP mode,” but you can approximate DDMRP logic:

• Set Demand Time Fence = 0 days: Don’t use forecasts; rely only on confirmed demand (sales orders).
• Set Min/Max inventory levels manually: Configure minimum and maximum stock levels on item master instead of relying on safety stock formulas.
• Set Order Quantity = Lot-for-Lot: Order only what is needed, when it is needed.
• Use Demand Planning App for real-time visibility: Pull actual POS or order data into Demand Planning to sense demand in real-time.

DDMRP is most suitable for high-velocity, predictable demand products (consumer packaged goods, retail, distribution). It’s less suitable for complex manufacturing with long lead times or custom orders.

Performance Tuning & Optimization

As your product portfolio grows and supply chain complexity increases, Planning Optimization can slow down. A plan run that took 30 minutes with 100K SKUs might take 60+ minutes with 500K SKUs.

Performance Monitoring

Measure plan generation time: Time from “Run Plan” click to results returned. Log this weekly. If trending up, investigate.

Typical Benchmarks:

• 100K SKUs, 100K plan lines: 5–10 minutes
• 500K SKUs, 500K plan lines: 20–40 minutes
• 1M+ SKUs, 1M+ plan lines: 45–90 minutes

If your plan run is 2–3x slower than benchmark, investigate root causes.

Performance Optimization Strategies

Strategy 1 – Reduce Plan Scope: Instead of planning all 500K SKUs, plan only high-velocity items (e.g., top 80% of sales volume). Use static safety stock for slow-movers. Reduces plan lines by 30–50%; cuts plan time proportionally.

Strategy 2 – Reduce Coverage Time Fence: Instead of planning 365 days forward, plan 180 days. Beyond 180 days, demand is too uncertain anyway. Reduces plan scope; improves responsiveness.

Strategy 3 – Optimize Master Data: Remove obsolete items from item master. Orphaned BOMs, dead suppliers, or invalid item coverage settings confuse the planner. Data cleanup can improve plan time by 10–20%.

Strategy 4 – Reduce Demand Variability (via Demand Planning): More accurate forecasts lead to simpler, faster planning. Invest in Demand Planning app to improve forecast accuracy. Reduces safety stock needs and plan complexity.

Strategy 5 – Adjust Planning Bucket: If using Daily buckets, switch to Weekly. Fewer time periods = faster planning. Trade-off: less granular (orders are rounded to weeks instead of days).

Planning Optimization Limitations & Workarounds

Planning Optimization is powerful but has some constraints:

Limitation 1: No Built-In Demand Prioritization

Planning Optimization doesn’t natively support “VIP customer” or “critical order” priorities. If you have 1,000 units of a constrained product and two customers (one strategic, one small), Planning Optimization doesn’t know to prioritize the strategic customer.

Workaround: Pre-allocate supply to VIP customers manually. Create a separate item variant or planning pool for VIP demand. Adjust their demand forecast upward in Demand Planning.

Limitation 2: No Constraint-Based Optimization

Planning Optimization doesn’t consider capacity constraints (machine hours, labor hours) when generating work orders. It will happily generate 1,000 work orders for Week 3 even if your plant can only handle 200.

Workaround: Use Capable-to-Promise (CTP) module to validate orders against capacity. Or use Advanced Planning and Scheduling (APS) tools like Logility or o9 for constraint-aware planning.

Limitation 3: Limited Scenario Comparison

Planning Optimization supports multiple scenarios (Base, Competitive, Demand Surge) but comparing them is clunky. It’s not designed for interactive what-if analysis.

Workaround: Use Power BI or custom reports to compare scenario results. Or use Planning Optimization API for programmatic scenario analysis.

Limitation 4: Slow for Real-Time Order Promising

If you want to promise a customer delivery date in real-time (when they place an order), Planning Optimization is too slow. A full plan run takes 20+ minutes; you need an answer in seconds.

Workaround: Use Capable-to-Promise (CTP) or Available-to-Promise (ATP) logic for real-time promises. These use available inventory and known supply (firmed orders) instead of running full optimization.

Frequently Asked Questions

Do we have to migrate from the deprecated MPS engine to Planning Optimization?

Yes. Microsoft has committed to sunsetting the MPS engine. All organizations must migrate by end of 2024. If you continue using MPS past that date, it will stop receiving support and functionality updates.

Will Planning Optimization generate the exact same planned orders as our old MPS configuration?

Not necessarily. Planning Optimization uses different algorithms and may have subtle differences in how it rounds quantities, interprets lead times, or prioritizes multiple demand signals. Expect 5–20% variance in planned order timing and quantities. If variance is large, investigate configuration differences (safety stock, demand time fence, coverage time fence).

How do we determine the right service level (95%, 97.5%, 99%)?

Service level is a business decision. Higher service levels require more safety stock and higher carrying costs. Start with 95% for commodity items and 99% for critical components. Over time, measure actual stockout frequency and adjust service level target up or down based on actual vs. target performance.

Should we use Planning Optimization API for real-time planning?

Planning Optimization API exists but is slow (20–60 minutes per run). It’s suitable for batch scenarios or what-if analysis, not real-time order promising. For real-time order promising, use Capable-to-Promise (CTP) logic instead.

Can Planning Optimization handle multi-level BOMs with deep hierarchies?

Yes. Planning Optimization handles complex multi-level BOMs efficiently. However, if BOM structure is invalid or contains orphaned components, plan generation will be slow or produce unexpected results. Validate BOM integrity before running plans.

How often should we run the master plan?

Weekly is most common (e.g., every Monday morning). Some organizations run twice weekly if demand is very volatile. Monthly is too infrequent and leads to stale supply plans. Daily is usually overkill unless you have highly perishable products or extreme demand variability.

What’s the best way to manage demand forecast errors?

Track forecast accuracy (MAPE or MAE) monthly by product family. Identify items with MAPE >50%. For high-error products, increase safety stock or use Manual forecasting (expert judgment) instead of statistical forecasting. Review demand sensing accuracy (how often actual sales exceed forecast) every quarter.

Can we use Planning Optimization for multiple entities (multi-company planning)?

Yes, Planning Optimization considers all legal entities and generates intercompany transfer orders based on cost and lead time. However, master data (item master, forecasts, coverage settings) must be set up per entity. Ensure consistent coverage policies across entities to avoid surprising intercompany behavior.