Inventory Control
Transcription
Inventory Control
Fifth session : "Inventory Control" Objectifs : Comprendre les techniques de gestion des stocks et les paramètres qui s’y rapportent. Contenu : - Exercice préliminaire à réaliser en salle d' info avant la séance d' exercice (pages 2 à 10). Attention, cet exercice contient certaines notions qui ne seront plus revues dans les autres exercices ; -Correction de l’exercice préliminaire réalisé en salle d’info (pages 12 à 14); - Enoncés des exercices (pages 15 à 17); - Solutions des exercices (pages 18 à 21). T.P. : - Exercice à réaliser impérativement par deux en salle d’informatique; - Plage horaire et locaux (voir annonce aux valves); - Démarrage: Double-cliquez sur “p:\prod2100\level1.xls”. Site WEB : - Il est possible de se procurer le fichier level1.xls sur la page WEB suivante : http://www.prod.ucl.ac.be/enseignement/notes/prod2100.html Fifth session : Inventory Control 1 Computer Laboratory : Inventory Control Problem Description This game aims at getting familiar with the inventory control policies. The problem is simple : one warehouse, one retailer and one product. The items flow from the warehouse to the retailers where they are sold. The average demand at the retailers is 0.5 item per day (half a kilo for example). You are the retailer and have to decide when and how many items should be ordered from the warehouse. Your goal is to maximise the profit over a period of time. Costs are due : -for placing orders at the warehouse (order cost); -for holding items as soon as they are ordered (pipeline and holding cost). Revenues are generated by selling items. Software Description You should get a similar screen to the following one. CON T R OL Warehouse -> R e se t va lue s S imula te 1 da y S imula te 1 ye a r Supply lead time (days) Demand Distrib. (U, N, C) Review period (days) Initial inventory Reorder point Order quantity Order up to level 7.0 0 <- order transit(-4) transit(-3) transit(-2) transit(-1) for tomorrow <- inventory demand sales IAG 1997 S ta tistics on da y: #sales 0 #demands 0 fill rate 0 #stockouts 0 #cycles Retailer -> 0 stockout freq today's 1 pipeline cost today's 0 holding cost Lt = 4 order cost Type = u total profit Rp = 10 1 year = (days) Io = 2 New random list(True,Fal.) R= Unit Profit ($/item) Q= Unit Holding cost ($/item/day) Qmax = 6 Unit Order cost ($/order) Inventory Control : Description 5.0 100 46 53 0.87 3 10 0.30 7.04 8.16 10 1354.8 100 FALSE 30 0.04 1 IAG/93-94 This3.0 game aims at getting familiar with the inventory control policies. The1.0 problem is simple : one warehouse and one retailer. The-1.0 items flow from the warehouse to the retailers where they are sold. The average daily demand is 0.5. The distribution is either Constant (0.5) or Uniform (either 0 or 1). The screen is divided into 5 different parts: the control part (top left), the daily status (top center), the global statistics (top right), the data (middle) and a chart (bottom). These different parts are detailed in the following four pages. If you want, you can immediately move to the cases and refer to those explanations when needed only. Fifth session : Inventory Control 2 Help page 1: Data ( middle - green on the screen ) Meaning The lead time is the number of days you will have to wait before getting your order. It should remain in [0, 5] days. A lead time of 0 means that the order placed on the evening comes for the next day morning. If your lead time is 1, your order will arrive one day later. So, if you order on Monday evening, your order will arrive on Wednesday morning. Demand Distribution C = constant (the daily demand is 0,5); (U, N, C) U = uniform (the daily demand is variable and equals 0 or 1 with Type= probability 1/2); N = normal (the daily demand is variable and distributed as a normal function with a daily average demand of 0,5). Review period (days) Number of days between two checks of the inventory level. An Rp= order can be placed only at the end of a review period. If review period = 0, this is equivalent to a permanent review. If review period > 0, then you are in a periodic review. In this case, you check your inventory every (integer) Rp days. Initial inventory The initial inventory is the inventory you have at day zero and Io= that you can use at the beginning of the game. It allows you to avoid problem at the start of the simulation, before your policy is effective, otherwise you could miss some sales before your first order is delivered. Reorder Point Used when your are in permanent review. The order point is the R= level of your inventory position at which an order of size Q is automatically placed (when simulating a whole year). Order Quantity Used when you are in permanent review. The order quantity is Q= the quantity which is ordered when the reorder point is reached. Order up to level Used when you are in periodic review. Every review period, an order is placed. The size of the order is the difference between Qmax= Qmax and the current inventory position. This difference is calculated for you in the soft. You only have to choose Qmax. 1 Year (days) The period of time during which your inventory control policy will be simulated. A normal (but modifiable) value is 100 days. New random list If true, then a new random sequence is generated for the run. (true or false) Otherwise, a same pre-computed random sequence will be used. Unit profit ($/item) The difference between the selling price and the purchase cost gives the profit you make when selling 1 unit. Set to 30 $/item. Unit Holding cost Cost generated by holding one item one day in your inventory or ($/item/day) in the pipeline (truck). Set here to 0.04 $ / day / item. Order cost ($/order) Cost generated by each order you place at the warehouse. Set here to 1 $ / order. On your screen, the light green part (on the left) corresponds to the data (Lt, Type, Rp) you have to enter according to your assignments and to the inventory control policy you want to simulate (Io, R, Q, Qmax). The dark green data are the simulation parameters and the cost data. These parameters should normally not be modified. DATA Supply lead time (days) Lt= Fifth session : Inventory Control 3 Help page 2: Control, daily status and Statistics (Top of screen) Click control buttons (top left - grey) Reset values The simulation restarts at zero, all the former statistics are erased. The retailer inventory is set to Io. Simulate 1 day The program simulates the demand for 1 day, determines the daily sales and the new inventory position, and updates all the statistics. Simulate 1 year The program simulates your strategy for one year (100 days). Every day, it determines the demand, the sales and the new inventory position. According to your policy, an order is possibly placed. All the statistics are recorded. Warehouse Order Transit from warehouse to retailer Retailer Inventory Demand Sales Daily Status (top center; white and red) is the quantity you order at the warehouse. You only have to type in your order when you work day by day. describes the transit of the order when you work day by day. Enter for example a lead time of 5 days and a warehouse order of 2 items and click several times on ""Simulate 1 day"". Look at the transit: the ordered items will go from "transit 4" to "for tomorrow" before being in your stock. shows what you have in your inventory (on the evening of the current day). It is used to serve the demand and is replenished by your orders. shows the retailer demand of the current day. show the retailer sales of the current day. Statistics (top right ; yellow) Statistics on day Number of days you already simulated. All the statistics correspond to the period from day 0 evening up that day, on the evening. # sales Number of sales made (up to the current evening). # demands Number of demands placed (up to the current evening). fill rates The fill rate observed over this period: = # sales / # demands. # stockouts Number of cycles where a stockout has been observed. # cycles Number of cycles observed in the simulated period. A cycle is counted each time an order is received. stockout fr. The percentage of cycles with stockout: = # stockouts / # cycles. Pipeline cost Cost generated up to now by all the ordered items when they are "in (Pc) transit". Per day, it is equal to (# units in transit that day × 0.04). Holding cost Cost generated up to now by the items that remained in your (Hc) inventory. Per day, it is equal to (average # of items in inventory that day × 0.04). The average is taken between the morning and the evening. Order cost Cost generated by the orders you placed at the warehouse. Per day, it (Oc) equals 1 if an order is placed, 0 otherwise. Total Profit = (unit profit) * # Sales - Order cost - Holding cost - Pipeline cost. Fifth session : Inventory Control 4 Help page 3: Enter the Data First you reset the data by clicking the corresponding button. Then you set the values for the parameters: • supply lead time (Lt) • demand distribution (Type) Then you can decide to play either manually (day by day) or to let the computer simulates your strategy for a whole "year". How to Play: manually Every day, you must decide how much to order and enter that quantity as a Warehouse order (top center). Then, you click on "SIMULATE 1 DAY". The program simulates the demand for 1 day and updates all the statistics. You then repeat this process day by day. Here below is an example where 4 units were ordered on day 0 evening, and 3 more units on day 6 evening. The data here were (zero lead time, and constant demand = 1/2 per day). W arehouse -> CO N T R O L R e se t va lue s S imula te 1 d a y S imula te 1 ye a r Supply lead time (days) Demand Distrib. (U, N, C) Review period (days) Initial inventory Reorder point Order quantity Order up to level 0 <- order S ta tistics o n d a y: transit(-4) #sales transit(-3) #demands transit(-2) fill rate transit(-1) #stockouts 0 for tomorrow #cycles Retailer -> 3.5 <- inventory stockout freq today's 0.5 demand pipeline cost today's 0.5 sales holding cost Lt = 0 IAG 1997 order cost T ype = c total profit Rp = 1 year = (days) Io = New random list(T rue,Fal.) R= Unit Profit ($/item) Q= Unit Holding cost ($/item/day) Qmax = Unit Order cost ($/order) Inventory Control : Description 4.0 3.0 2.0 T his game aims at getting familiar with the inventory control policies. 1.0 T he0.0 problem is simple : one warehouse and one retailer. T he-1.0 items flow from the warehouse to the retailers where they are sold. 7 3.5 3.5 1.00 0 2 0.00 0 0.75 2 102.25 100 FALSE 30 0.04 1 IAG/93-94 T he average daily demand is 0.5. T he distribution is either Constant (0.5) or Uniform (either 0 or 1). Order curve Inventory curve Lost sales Help: CHART (bottom) The set of black (positive) columns correspond to your orders. The green columns (positive, grey on the figure) indicate the evening inventory positions The red columns (negative, grey on the figure) show when sales are missed. For this example, all the statistics are quite obvious, except for the costs. Two orders were placed and this explains the order cost. To determine the holding cost, you should compute the average inventories on each day. They amount to 3.75 for day 1, 3.25 for day 2, and so on. All together, 18.75 unit day. This explains the holding cost. The profit is obtained by subtracting the costs from the sales revenues. Fifth session : Inventory Control 5 Help page 4: How to Play automatically First, you have to set the review period Rp. A. * If you want to simulate a whole year with permanent review (Rp=0), you have to choose and enter the reorder point R and the order quantity Q. * Then, you click on "SIMULATE 1 YEAR". The program simulates your strategy for a whole year: an order of size Q is generated each time the inventory position hits the order point (R). Only integer values are allowed for R and Q. Here below is an example of results for the indicated data. The graph on the bottom shows the orders that were automatically placed for you. Also, below the horizontal axis are indicated the sales you missed. Try now to understand the statistics related to the service performances (observed fill rate and stockout frequency). Warehouse -> CON T R OL R e se t va lue s S imula te 1 da y S imula te 1 ye a r Supply lead time (days) Demand Distrib. (U, N, C) Review period (days) Initial inventory Reorder point Order quantity Order up to level 7.0 0 <- order transit(-4) transit(-3) transit(-2) transit(-1) for tomorrow <- inventory demand sales IAG 1997 S ta tistics on da y: #sales 0 #demands 6 fill rate 0 #stockouts 0 #cycles Retailer -> 0 stockout freq today's 1 pipeline cost today's 1 holding cost Lt = 4 order cost Type = u total profit Rp = 0 1 year = (days) Io = 3 New random list(True,Fal.) R = 1 Unit Profit ($/item) Q = 6 Unit Holding cost ($/item/day) Qmax = Unit Order cost ($/order) Inventory Control : Description 100 45 53 0.85 6 7 0.86 7.2 10.82 8 1323.98 100 FALSE 30 0.04 1 IAG/93-94 5.0 This3.0 game aims at getting familiar with the inventory control policies. The1.0 problem is simple : one warehouse and one retailer. The-1.0 items flow from the warehouse to the retailers where they are sold. The average daily demand is 0.5. The distribution is either Constant (0.5) or Uniform (either 0 or 1). Order curve Inventory curve Lost sales Help: CHART (bottom) The set of black (positive) columns correspond to your orders. The green columns (positive, grey on the figure) indicate the evening inventory positions The red columns (negative, grey on the figure) show when sales are missed. B. * If you want to simulate a whole year with periodic review (Rp>0), you have to choose and enter the "order up to level" Qmax. * Then, you click on "SIMULATE 1 YEAR". the program simulates your strategy for a whole year: an order of size : ( Qmax current inventory position ) is generated at each review period. An example of results was given with the software description (page 2). Fifth session : Inventory Control 6 Cases 0 and 1 If you do not understand some variables, some performance or some manipulation, please refer to the former three help pages. Case 0 : Manual simulation : understanding the game * Set the inital inventory Io to 0 and reset by clicking on the "reset-restart" button. It's now day 0, evening. * Enter the following data : Constant demand (distribution = C), lead time = 1. * Enter an order of size 10 (as a warehouse order) and click on "Simulate 1 day". It's now day 1, evening. Explain : 0.1. why your order did not arrive; 0.2. the demand of this day; 0.3. the sales of this day and the inventory position on the evening ; 0.4. all the statistics values; * Click on "Simulate 1 day". It is now day 2, evening (see the day count). Explain again : 0.5. the demand and sales of this day, the inventory on this evening; 0.6. all the statistics from day 0 up to day 2 (a holding cost is charged for the average inventory during this day. In the morning, you had 10 units; in the evening, 9,5 units are left. This means an average of 9,75 units); * Click several more times on ""Simulate 1 day"" and each time, be sure you understand all the figures. If you missed something, restart the case 0 completely and refer systematically to the help pages. Leave the case 0 only when you are sure you understood everything. Case 1 : EOQ with zero lead time Constant demand (distribution = C), lead time = 0, review period = 0 (permanent review). With those assumptions: 1.1. Which are the best reorder point (R) and order quantity (Q) ? Determine those values mathematically. Perform an automate simulation. With these parameters, explain all the performance. 1.2. Once you've found the optimal values, record your results, then observe what you obtain when you add or substract 1 to Q [look at the statistics]. 1.3. At the optimum, how are the holding and order costs related ? 1.4. If you increase the reorder point by 1 (set R to 1), what do you gain, what do you loose ? Check with the software. Note : If you do not understand what happens, examine attentively the chart at the bottom of the screen. Fifth session : Inventory Control 7 Cases 2 and 3 Case 2 : EOQ with lead time Constant demand, lead time = 4 days, review period = 0 (permanent review). With those assumptions : 2.1. Which are the optimal order point and order quantity ? 2.2. Don' t you think an initial inventory is necessary to avoid stockouts at the beginning ? If yes, which one should be the best one ? 2.3. Compute the pipeline cost (still on a one year basis) mathematically and check with the soft. Case 3 : Safety stock with permanent review Variable demand (distribution = U), lead time = 4 days, review period = 0. With those assumptions: 3.1. Explain the demand distribution 3.2. Simulate one year with the same R, Q and Io you used in case 2. Examine the chart. What do you observe in terms of number of sales ? Explain the fill rate and the stockout frequency. 3.3. The reorder point R is normally defined as R = Dd*Lt + SS where Dd is the evening daily demand, Lt the lead time and SS the safety stock. Which SS did you use in case 2 ? 3.4. Set the safety stock to 2, what is then R ? Is there any risk of loosing sales in this case ? What about the risk when R equals 3? 3.5. Check with the soft and record for each R value the observed fill rate and stockout frequencies. 3.6. Check again with the soft (with R = 3) when working with a year length of 200 days. Don't you loose some sales in this case ? Why ? 3.7. A good exercise at home, consists in comparing the gains and losses when the order point changes. In the computer room, just perform a simulation with R equals to 2, 3 and 4 and each time, record all the statistics. Here are indications on how to compare the solution R=2, R=3, R=4. The basic question is: "Is it worthwhile to increase R, for example, from R to R + 1 ?" By increasing R by one unit, we increase the average inventory by one unit. How much does this cost per year ? This money represents my "loss". On the other hand, by increasing R by one unit, we could hope to save 1 sales (one lost sales less). My gain is here : = Prob (saving 1 sales in a cycle) * number of cycles per year * profit per sales = Prob (DLt > R) * number of cycles per year * profit per sale We give you this probability distribution : Prob [demand during the lead time = 4] = 1/16 Prob [demand during the lead time = 3] = 4/16 Prob [demand during the lead time = 2] = 6/16 Prob [demand during the lead time = 1] = 4/16 Prob [demand during the lead time = 0] = 1/16 (Try also to understand where these values come from). Fifth session : Inventory Control 8 Decide finally (after having filled the next table), which R value is the best one, economically. Did you find the same result with the simulation ? 2 ------------> 3 3 ------------> 4 4 ------------> 5 Loss Gain 3.8. At home, for R = 2, 3 and 4; try to calculate the corresponding stockout probability and fill rate. Remember that the fill rate is given by backorder model: Q − n( R) Q R=2 R=3 R=4 SS Stockout Probability (per cycle) = X = P( DLt > R) Number of cycles in 100 days = Y Number of stokouts in 100 days = X*Y Time between 2 stockouts = 100 days/(X*Y) n(R) = P( DLt = X )( X − R) X >R fill rate = 1 − n( R ) Q Again, compare the mathematical results with the simulation results. Fifth session : Inventory Control 9 Case 4 Case 4 Safety stock with periodic review Variable demand (distribution = U), lead time = 4, review period of 10 days. With those assumptions: 4.1. Set Qmax = 8 and an initial inventory of 4. Simulate one year (100 days). 4.2. Examine the chart. Note when an order was placed and how much was ordered. 4.3. Did we loose some sales ? 4.4.The vulnerability period is the period of time during which we cannot interact on the inventory. If we do not order now, we will have to wait for the next review period (10 days), then order and wait for the lead time (4 days). How long is the vulnerability period here ? 4.5. What is the maximal demand during the vulnerability period ? 4.6. Set Qmax to this value and simulate for one year (think at the initial inventory). Can we loose sales ? 4.7. Decrease several times Qmax by 1 until you observe some lost sales. 4.8. Which Qmax gives the best profit ? Try several values and record all the performances. 4.9. Again, as in case 3, you can try to calculate at home the different probabilities of having a demand of 10, 11, 12, 13 and 14 items during the vulnerability period. You can estimate it using a normal distribution (see case 3 / at the end of the page) average daily demand is 0,5 item and that your vulnerability period lasts 14 days, what is the average demand during this period ? Knowing that the daily standard deviation is of 0,5 item, what is your standard deviation during the vulnerability period ? Once you have those two variables, you can calculate the different probabilities asked here above. Fifth session : Inventory Control 10 ANSWERS COMPUTER LABORATORY: Case 0: 0.1. because the lead time is 1 day. The order placed on day 0 evening will arrive on day 2 morning. 0.2. The demand of the day is 0,5. 0.3. There is no sale at day 0. The inventory position is at zero but, there are 10 units in the pipeline. 0.4. demands = 0,5 sales = 0 pipeline cost = 0,4 = 0,04 * 10 * 1 holding cost = 0 order cost = 1 profit = -1,4 0.5. demands = 0,5 sales = 0,5 0.6. demands = 1 = 0,5 + 0,5 sales = 0,5 pipeline cost = 0,4 holding cost = 0,39 = 9,75 * 0,04 order cost = 1 profit = 13,21 = 15 - 0,4 - 0,39 - 1 Case 1: 2 * 1 * 50 ; 0,04 * 100 R = 0 (because the lead time is zero). 1.1. Q = 5 = 1.2. 4 5 6 Holding cost 8,16 10 12,32 Order cost 13 10 9 Total cost 21,16 20 21,32 Hc (R = 5) = 0,04 * 100 * 5/2 = 10; Oc (R = 5) = 1 * 10 (indeed, number of cycles on one year = D/Q = 50/5 = 10). 1.3. at the optimum, Oc = Hc = 10. 1.4. Increasing R aims at reducing the number of lost sales. Here we do not loose any sales (lead time being 0), so, we cannot gain anything. It would just cost us the holding cost of 1 unit during 1 day. Fifth session : Inventory Control 11 Case 2: 2 * 1 * 50 ; 0,04 * 100 R = 0,5 * 4 + 0 = 2 2.2. Yes, an initial inventory Io is necessary . Io = 2, Lt being of four days and the demand of 0,5 per day; 2.3. pipeline costs = 8 = 50 * 4 * 0,04 (= yearly demand(items in transit) * Lt * Hc). 2.1. Q = 5 = Case 3: 3.1. Prob[Dj = 0] = 0.5; Prob[Dj = 1] = 0,5; 3.2. Number of demands = 53 Number of sales = 49 4 lost sales 3.3. SS = 0 (case 2) because the daily demand was constant (C); 3.4. R = 4 = 0,5 * 4 + 2; no risk of loosing sales in this case. If R = 3, there is 1/16 chances to loose sales. Sales will be lost if (D > R) during the lead time, i.e. prob (DLt>3) = Prob (DLt = 4) = 1/2 * 1/2 * 1/2 *1/2 = 1/16. On a year length of 100 days, we will run 10 (Q = 5, year = 100 days, Dday = 0,5 => 10 orders a year) times the risk to loose a sale with probabilité 1/16. With the soft, no sales are lost, the probability of loosing sales being too small for a year length of 100 days (number of demands = 53 = number of sales). 3.6. Number of demands = 105 Number of sales = 104 1 lost sale 3.7. 2 ------------> 3 3 ------------> 4 4 ------------> 5 Loss(1) 0,04 * 100 * 1 = 4 0,04 * 100 * 1 = 4 0,04 * 100 * 1 = 4 Gain 5/16 * 10 * 30 1/16 * 10 * 30 0 = 93,75 = 18,75 (1) on néglige le fait que cette unité supplémentaire quittera quelque fois le stock. 3.8. R=2 SS 0 (= 2 - 0,5 * 4) Stockout Probability 5/16 per cycle (=1/16 + 4/16) Number of cycles in 10 100 days Number of stockouts 5/16 * 10 = 3,125 in 100 days Time between 2 100/3,125 = 32 stockouts (days) Fifth session : Inventory Control R=3 1 (= 3 - 0,5 * 4) 1/16 R=4 2 (= 4 - 0,5 * 4) 0 10 10 10/16 = 0,625 0 100/0,625 = 160 100/0 = ∞ 12 n(R) = average units which are backordered during a cycle fill rate 1*Prob (Dlt = 3) + 1*Prob (Dlt = 4) 2*Prob (Dlt = 4) = 1/16 = 1*4/16 + 2*1/16 = 6/16 1 - 6/80 = 0,925 1 - 1/80 = 0,9875 0 1-0=1 Case 4: 4.2. One order every 10 days with a Qi = Qmax- Ii. 4.3. Number of demands = 53 Number of sales = 52 1 lost sale. 4.4. Vulnerability Period = 14 days. 4.5. Dmax = 14 units (1 unit per day). 4.6. With Qmax = 14, no lost sales. 4.7. First lost sale with Qmax = 8. 4.8. Best profit with Qmax = 9. 4.9. µ = 14 * 0,5 = 7 σ = 14 * 0,52 = 1,87 => N(7; 1,87) Prob (D≥8) Prob (D≥9) Prob (D≥10) Prob (D≥11) Prob (D≥12) Prob (D≥13) Prob (D≥14) = Prob (D>6,5) = Prob (Z>(6,5-7)/1,87) = Prob(Z>-0,26) = 1 - Prob(Z>0,26) = 1 - 0,39 = 0,61 = Prob(Z>0,5/1,87) = Prob(Z>0,26) = 0,39 = Prob(Z>1,5/1,87) = Prob(Z>0,80) = 0,21 = Prob(Z>2,5/1,87) = Prob(Z>1,33) = 0,09 = Prob(Z>3,5/1,87) = Prob(Z>1.87) = 0,03 = Prob(Z>4,5/1,87) = Prob(Z>2,40) = 0,008 = Prob(Z>5,5/1,87) = Prob(Z>2,94) = 0,0016 = Prob(Z>6,5/1,87) = Prob(Z>3,47) = 0,00026 Prob (D=7) Prob (D=8) Prob (D=9) Prob (D=10) Prob (D=11) Prob (D=12) Prob (D=13) Prob (D=14) = Prob (D≥7) - Prob (D≥8) = 0,22 = Prob (D≥8) - Prob (D≥9) = 0,18 = Prob (D≥9) - Prob (D≥10) = 0,12 = Prob (D≥10) - Prob (D≥11) = 0,06 = Prob (D≥11) - Prob (D≥12) = 0,022 = Prob (D≥12) - Prob (D≥13) = 0,0064 = Prob (D≥13) - Prob (D≥14) = 0,000134 infime... Prob (D≥7) Fifth session : Inventory Control 13 QUESTIONS Les questions 2 à 5 sont relativement courtes et importantes. La question 1 est relativement longue mais vous permet de manipuler les techniques de calcul de stock. QUESTION 1 The BELGIAN DRUGS COMPANY manufactures pharmaceutical products in Europe. This company is one of the most important subsidiary of the Drugs Company. The corporate office is located in Dallas. The B.D.C. produces 600 products using more than 100 different raw materials. 1. GENERAL DATA Promoted “general manager” of the planning department, Mr Kozari was commissioned to set up the best possible inventory management systems. The first term spent in the company, allowed him to gather the needed information to set up an effective inventory policy for one family of products. Here are the data : • B.D.C. works 5 days a week, 48 weeks a year (1 month = 4 weeks). • According to the accounting manager, the annual holding rate of raw materials is of 20 % of the purchasing price. The order cost is estimated at 650 francs. 2. LIST OF THE PRODUCTS, OF THEIR PURCHASING PRICES AND OF THE SUPPLIERS' LEAD TIMES The following table only lists the most important products. The list was determined by an ABC classification of all the raw materials. Products Denatured alcohol Ammonic chloride Acetone Impalpable sugar Chocolate flavour Paraffin Glycol propylene Glycerine Sodic glumitate Coconut oil Soja flour Sodic citrate Suppliers Monsanto Hoffman Merck Merck Rhône-Poulenc Merck RIT Capsulit Rhône-Poulenc Merck Rexolin Monsanto Fifth session : Inventory Control Yearly average consumption (in KGs) 3500 16080 13500 8500 85 5500 18240 5500 32640 5000 2000 28080 Lead time Unitary (working days) Price (FRS/kg) 40 3 672 2 20 4 36 2 6500 5 49 4 70 var. 76 2 110 1 60 5 16 4 94 5 14 3. QUESTIONS 1) Establish an ABC classification on annual value criteria. Annual Value = annual consumption * unitary price Products Ammonic chloride Sodic glumitate Sodic citrate Glycol propylene Chocolate flavour Glycerine Impalpable sugar Coconut oil Acetone Paraffin Denatured alcohol Soja flour 2) Annual Value ? 3.590.400 2.639.520 1.276.800 552.500 418.000 306.000 300.000 270.000 269.500 140.000 32.000 20.600.480 % of annual cumulated % of Class value annual value ? ? ? 17.43 69.88 12.81 82.69 6.2 88.89 2.68 91.57 2.03 93.60 1.48 95.08 1.45 96.53 1.31 97.84 1.31 99.15 0.68 99.83 0.17 100 100% ? ? Assume the daily demand of the ammonic chloride is deterministic and constant, when and how many KGs of ammonic chloride would you order ? What is the total cost of this policy ? 3) Consider now that the daily demand follows a normal distribution with a weekly variance of 180 kg2. 3.1.) Because of a frequent use of ammonic chloride, Mr Kozari believes it is unacceptable to be out of stock more than once every five years. When and how many KGs of ammonic chloride would you order in this case ? What is the total cost of this policy ? 3.2.) Mr Kozari assumes now a 99 % fill rate is needed. He would like to adopt quickly a stock up policy to reduce the annoyances provoked by this product. Once more, determine when and how many KGs of ammonic chloride you would you order in this case ? What is the total cost of this policy ? 4.1.) Till now, we considered Mr Kozari had no or very low inventory control costs. Time has changed and these turn out to be far more expensive. Mr Kozari cannot bear it anymore, he thus decides to check the level of his inventory at some fixed instants. When do you advice him to check his inventory level ? Knowing that he doesn't want to be out of stock more than once every five year (case 3.1), what security stock would you build up ? Is this policy more or less expensive than the one chosen in case 3.1. ? Why ? Fifth session : Inventory Control 15 4.2.) Mr Hoffman changes once more its agreement with Mr Kozari and compels him now to place one order exactly every month. Determine when and how many KGs of ammonic chloride you would you order in this case ? What is the total cost of this policy ? QUESTION 2 Votre firme utilise des plateaux garnis au rythme de 10 unités par jour, 360 jours par an. Chaque plateau garni est composé d'un plateau et de quatre tasses. Le fournisseur FA fournit vous facture: 10 FB par plateau, 20 FB par tasse et 450 FB pour le transport (indépendamment de la composition et des quantités commandées). Le fournisseur FB ne fabrique pas de plateau mais peut vous fournir des tasses au prix de 20,25 FB sans aucun frais de livraison. Vous ne pouvez néanmoins commander moins de 500 tasses à la fois. Vous estimez un coût de détention équivalent à 40% l’an de la valeur stockée. Chez qui vous fournissez-vous, de combien et pourquoi? QUESTION 3 Vous achetez des quotidiens au prix de 20 francs et les vendez au prix de 30 francs. Les journaux invendus sont à jeter. La demande journalière est distribuée selon une normale de moyenne 100 et d'écart type 20. Combien de journaux commanderez-vous chaque jour? QUESTION 4 Votre firme dessert ses clients à partir de 4 dépôts régionaux. Chacun de ces dépôts traite approximativement le même volume. Votre firme envisage maintenant de regrouper ses 4 dépôts en un seul. Peut-elle s'attendre à une économie en terme de stock de sécurité? QUESTION 5 Vous avez constitué un stock de sécurité afin de garantir en moyenne une rupture tous les 5 ans. Vous décidez maintenant pour des raisons économiques d'augmenter les quantités de commandes (Q) tout en gardant le même point de commande (R). Cela at-il une incidence sur le nombre moyen de ruptures par an? Fifth session : Inventory Control 16 Réponses QUESTION 1 1) Products Ammonic chloride Sodic glumitate Sodic citrate Glycol propylene Chocolate flavour Glycerine Impalpable sugar Coconut oil Acetone Paraffin Denatured alcohol Soja flour Annual Value 10.805.760 3.590.400 2.639.520 1.276.800 552.500 418.000 306.000 300.000 270.000 269.500 140.000 32.000 20.600.480 % of annual cumulated % of Class value annual value 52.45 52.45 A 17.43 69.88 A 12.81 82.69 B 6.2 88.89 B 2.68 91.57 C 2.03 93.60 C 1.48 95.08 C 1.45 96.53 C 1.31 97.84 C 1.31 99.15 C 0.68 99.83 C 0.17 100 C 100% 2) When to order? R = 134 kg How many to order? Q* = 395 kg Total cost? TC = 10858764.7 frs 3) 3.1.) Permanent review When to order? Vulnerability period = 2 σ2 days = 8.48 kg P(z) = 0.00491 => z = 2.6 SS = 22 kg R = 156 kg How much to order? Q* = 395 kg Total cost? TC = 10.861.721 frs 3.2.) Permanent review When to order? Vulnerability period = 2 E(z) = 0.46 => z= - 0.1 SS = 0 Fifth session : Inventory Control 17 R = 134 kg How much to order? Q* = 395 kg Total cost? TC = 10858764.7 frs 4.1.) Periodic review When to order? T* = 0.024 year => ± 6 days F* = 40 orders / year How much to order? Qmax - Current inventory Vulnerability period = 2 + 6 = 8 days σ 8 days = 16.97 kg z = 2.6 SS = 45 kg Qmax = 581 kg Total cost? TC = 10.864.822 frs (more than 3.1) 4.2. ) Periodic review When to order? T* = 1 month F* = 12 orders / year How much to order? Qmax - Current inventory Vulnerability period = 2 + 20 = 22 days σ 22 days = 28.14 P(z) = 0.0166 => z = 2.1 SS = 60 kg Qmax = 1534 kg Total cost? TC = 10911672 frs QUESTION 2 Fifth session : Inventory Control 18 FA uniquement: (on fait une commande groupée c-à-d des plateaux garnis): calcul de Q optimal: (D = 3600; Oc = 450; Hc= 0.4*90 = 36) Coûts associés: détention : 36 * 300 / 2 = commande 450 * 3600/300 = Total Q=300 5400 5400 10800 FA et FB: Chez FA: Q optimal pour les plateaux: (D= 3600; Oc= 450; Hc= 0.4*10= 4) Q=900 Coûts associés: détention : 0,4 *10* 900 / 2 = 1800 commande 450 * 3600/900 = 1800 Chez FB: aucun coût de commande on commande le plus souvent possible Q = 500 Coûts associés: détention : 0.4*20.25 * 500 / 2 = Supplément de coût pour l’achat des tasses 3600*4*0.25 = Total 2025 3600 9225 NB : 0,25 = 20.25F(coût des tasses chez FB) - 20F(coût d’achat chez FA) On choisit donc la solution FA(plateaux par 900) + FB (tasses par 500) QUESTION 3 Nous effectuons un raisonnemnt marginal. Ce raisonnement est par ailleurs plus important que la calcul numérique qui s’en suit. Supposons que l’on commande Q journaux et que l’on se pose la question : “est-il intéressant de commander Q + 1 journaux ?” Gain Perte Q -------------------------------> Q + 1 si je vends le journal supplémentaire => 10 francs de gain si je ne vends pas le journal supplémentaire => 20 francs de perte gain : Prob (D > Q) * 10 perte : Prob (D ≤ Q) * 20 J’augmente Q tant que le gain est > à la perte et je m’arrête aux environs de l’égalité : Prob (D > Q) * 10 = Prob (D ≤ Q) * 20 <=> Prob (D > Q) * 10 = (1 - Prob (D > Q)) * 20 <=> Prob (D > Q) = 20 / 30 = 0,66 => Z = - 0,44 (cfr. table normale réduite). Autrement dit, le dernier journal que j’achète doit être vendu avec une probabilité de 2/3 (et perdu avec un probabilité de 1/3) pour qu’en moyenne, ni je n’y gagne, ni je n’y perde. Si la demande est distribuée selon N(100, 20), alors je commanderai une 91 journaux. Q−µ En effet, Z = - 0,44 = σ Fifth session : Inventory Control 19 <=> Q = - 0,44 * 20 + 100 = 91 QUESTION 4 Oui, car l’écart-type global sera inférieur à la somme des écarts-types individuels : • si 4 dépots séparés : La demande dans chaque dépôt est de (µ, σ). Ainsi, dans chaque dépôts on aura un SS = k * σ. Au total, cela donnera : SS = 4 * kσ • si regroupement des 4 dépôts : La demande globale pour les 4 dépôts est de moyenne : µ + µ + µ + µ = 4 µ et d'écart type σ = σ 2 + σ 2 + σ 2 + σ 2 = 2σ. Le SS commun aux 4 dépôts sera donc de k * 2σ pour le même service, inférieur à la somme des SS des 4 dépôts pris de manière séparée. QUESTION 5 La probabilité de rupture par cycle ne changera pas car celle-ci ne dépend pas du point de commande R qui ne change pas. Par contre, le nombre de cycles par an va diminuer puisque l'on commande par plus grande quantité. Le nombre moyen de ruptures par an va donc diminuer. Fifth session : Inventory Control 20