The Infosys global supply chain management blog enables leaner supply chains through process and IT related interventions. Discuss the latest trends and solutions across the supply chain management landscape.

« Is Supply Chain Planning still the top most priority of investment during uncertain times? | Main | Fearing supplier bankruptcy? Take the steady steps for procurement collaboration with your competition »

Resolving the puzzle of achieving Higher Service Levels and Lower Inventory Levels through Faster Demand Sensing!!

I had talked about how to do forecasting during recession in my earlier blog here. During my discussions with end users the common feedback has been that frequent forecasting can help in faster sensing of demand, it doesn’t help in reducing inventories. Some of the most common questions emerging from end users on short term frequent forecasting are –  

• How to deal with lead times which are longer than the forecasting period?
• How do I change my safety stock levels? Do I need to?
• How will my service levels will be affected?
• How shall I deal with larger demand variation in granular level forecasting?

It is of utmost importance for Supply Chain Practitioners to provide resolution to these pertinent end user concerns. Hence I thought I would share my perspectives at answering these questions through this blog. For explanation, I have designed an illustration taking a cue from a real business scenario, and skinning out all complexity to bare minimum in an effort to explain the main concept.

For understanding, let’s assume that business planning is done on quarterly basis. Demand for the quarter is 300 units. Forecasting frequency is once a month and monthly demand is arrived at by multiplying quarterly demand with weighted average of a month. Let’s say, the monthly demand is 100 units. Variation in demand is 20%. Existing service levels are at 90%. Lead time for the order delivery is 15 days. Safety stock definition is arrived at looking at lead time, and hence, kept as two weeks (~15 days).

The inventory held for this scenario by the planning manager was, 50 units (inventory to meet demand for two weeks)+ 20 units (on account of variation on 100% demand) + 50 units (safety stock calculated for two weeks) = 120 units. With this calculation, the planner was able to achieve 9 to 10 inventory turns and desired service levels. Hence, he believed that this was an efficient arrangement and will not benefit significantly from weekly forecasting.

Let’s investigate, if the planning manager starts doing weekly forecasting where variation in demand jumps from 20 t0 30%, how can he still reduce his inventory and increase the service levels? The demand for Week 1 is 15, week 2 is 20, week 3 is 30 and week 4 is 35. Weekly forecast is not equal to monthly forecast divided by 4. Each week’s forecast is, in effect, is calculated through history.

Earlier, Safety stock = Lead Time = 2 weeks. Lead time is a key factor affecting safety stock but so as service levels. In fact, Safety stock can be determined as a product of Service Factor x Variation in Demand. For the value of Service Factor (SF) = 1, we can achieve 84% of service levels. For 98% service level, the value of SF = 2.

Note on Service Factor : Service factor is calculated by using inverse of the standard normal cumulative distribution with mean zero and standard deviation of 1. Too difficult to understand? Just use NORMSINV function in excel. If your desired service levels is 92%, take NORMSINV (92%) = 1.41.

For the new inventory scenario, we will be ambitious and aim at achieving 98% of service levels. The Service Factor value will be 2. So, Safety Stock = Service Factor x Variation in Monthly Demand = 2 x 20 = 40 units. We will also have to factor lead time in the calculations. Earlier, lead time of two weeks was less than the forecasting frequency of four weeks, hence, it is agreeable to hold safety stock equivalent to the lead time. However, now the forecasting period is less than the lead time, we need to account this using Lead Time Factor. Hence, we will replace Lead Time stock of two weeks with Lead Time Factor = Square root (Lead time / Forecast period). For this illustration, this factor will be square root (2/1) = 1.414
Lead Time Factor x Safety Stock = 1.414 x 40 = 57 units.

Summarizing both scenarios in a snapshot:

Scenario 1:

 Monthly ForecastInventory Units
Forecast100 50 (2 weeks)
Demand Variation20% 20
Order Lead Time2 weeks -
Safety Stock2 weeks 50
Total Inventory120 Units

Scenario 2:

Weekly Forecast Forecast Demand Variation (30%) Safety Stock Lead Time Factor * Safety Stock Total Inventory Order Qty
Week – 1 15 5 40 57 77 Q1
Week – 2 20 6 40 57 83 Q2
Week – 3 30 9 40 57 96 Q3
Week – 4 35 10 40 57 102 Q4

The calculation shows that little back-of-the-page work can instantly realize the benefit of frequent forecasting through 15-20% inventory savings, yet improving service levels. It accounts for demand variation on both – weekly and monthly levels and provides enough inventory buffer to meet the demand. A planner can choose to consider only one of the demand variation and tighten the inventories further. The concept of demand pacing can be used very efficiently with such planning. Forecast for every week will be tweaked upwards or downwards based on the actual consumption data. Order quantity and frequency can be changed every week to pace the demand for the month and subsequently, for the quarter.

For discussion, I have tried keeping the illustration as simple as possible. More complexity can be added by introducing lead time variation, constraints on order frequency and order lot size, order cost, different service levels for different customers etc. However, it is certain that organizations doing more frequent forecasting will be benefited through it. As shown in the illustration above, in spite of reducing inventory, it allows organizations to achieve higher service levels. Both of them are, critical during recessionary times!!


Hi Mitul,
Please let me know how the total inventory was calculated in Scenario 2.

Post a comment

(If you haven't left a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Thanks for waiting.)

Please key in the two words you see in the box to validate your identity as an authentic user and reduce spam.

Subscribe to this blog's feed

Follow us on

Blogger Profiles

Infosys on Twitter