On the impact of inventory accuracy improvements on sales Christoph Glock, Yacine Rekik, Aris A. Syntetos ECR, Paris, France – February 08, 2018 - 1 -
Background and objectives Inventory inaccuracies: major issue in retailing and apparel industry. Physical stock is (typically) less than what we think it is. Most reasonable assumption in retailing. Generally, stores are negative in terms of stock and distribution centres are positive. Thus, reconciling inventories may only lead to an increase in sales. ( We will see later that positive stock is also possible, still leading though to reduced sales! ) The problem has been established; we are not here to argue for its existence. - 2 -
Background and objectives But rather: Assess the implications of the problem, or rather the implications of fixing the problem (phase 1); Assess alternative ways of fixing the problem itself (phase 2). Phase 1: What is the impact on (increased) sales if inventory accuracy is increased by x%? How does inventory accuracy develop over time after a stock take? Is there an optimal number of stock takes? What exactly constitutes this problem of inventory discrepancies? Phase 2 (upon convincing everybody of the implications): what are the strategies to be employed (algorithmic driven, new identification technologies, counting, etc.) to fight the route causes of the problem? - 3 -
Error free ( r,Q ) inventory policy Order Quantity Q Q Reorder Point: r Safety Stock t L 2 L 1 Lead Time - 4 -
Impact of errors on the ( r,Q ) inventory policy This is the visible stock behavior: POS (Real Demand) + Replenishment Expected error free Order Quantity Q stock level Q Actual stock level subject to errors Reorder Point: r Safety Stock t L 2 L 1 Lead Time This is the invisible stock behavior: POS (Real Demand) + Replenishment + Skrink (Ghost Demand) - 5 -
Impact of errors on the ( r,Q ) inventory policy Wrong ordering Order Quantity Q Lost Sales are timing and more quantity frequent Q Reorder Point: r t Safety Stock L 2 L 1 And more importantly, Lead Time they are not detectable - 6 -
Impact of errors on the ( r,Q ) inventory policy Wrong ordering Order Quantity Q Lost Sales are timing and more quantity frequent Q Errors act as “ghost” demand decreasing the stock level without generating a revenue: The Inventory is controlled based on some visible wrong information, Reorder Point: r whereas the sales are satisfied based on some correct but invisible information t Safety Stock L 2 L 1 And more importantly, Lead Time they are not detectable - 7 -
Matched store experiment Stock Count Sales 12 weeks after the “Test” Sales (e.g.) 12 weeks inventory records have been before count “trued up” No Stock Count “Control” Sales (e.g.) over 12 weeks Sales 12 weeks after - 8 -
Empirical analysis Currently working with 8 Retailers across Europe: NDAs have been signed and we are in various phases with regards to data transfer and analysis; 4 Grocery retailers (supermarkets), 2 Apparel retailers and 2 other; Customised reports to be produced for all participating retailers. Initial results: 2 Grocery retailers: ALPHA and BETA - 9 -
Aim of the study There is some published evidence suggesting that as much as 65% of inventory records are wrong (Nicole DeHoratius, 2012). This is True for retailers ALPHA and BETA: Retailer BETA Retailer ALPHA Week 12 Week 1 2 3 4 5 Average Negative Discrepancy 29.93% Negative Discrepancy 26.46% 28.39% 27.35% 28.33% 35.42% 29.19% Positive Discrepancy 24.44% Positive Discrepancy 11.34% 12.99% 12.67% 13.33% 11.44% 12.36% Zero Discrepancy 45.63% Zero Discrepancy 62.20% 58.62% 59.97% 58.34% 53.14% 58.45% Our objective is not to prove the existence of Discrepancies, but to help answering the questions: Does more accurate inventory grow sales, if so by how much? What investment is required to improve it? - 10 -
Experiment at Retailer ALPHA TEST Stock Audit Store 12 12 weeks weeks CONTROL Store Sales in the Test and Control stores are compared and the impact of the stock audit on sales of the test store is investigated - 11 -
SKUs Clustering: ABC classification based on Turnover Middle Movers Slow Movers Fast Movers 40.91% of 37.81% of 21.26% of SKUs SKUs SKUs Turnover: 80.00% 69.87% 70.00% 60.00% 50.00% 40.00% 30.00% 20.00% Turnover: 10.00% 25.06% Turnover: 5.07% 0.00% Turnover Contribution - 12 -
SKUs Clustering: ABC classification based on Turnover Middle Movers Slow Movers Fast Movers 40.91% of 37.81% of 21.26% of Fast Movers (17.78% of SKUs SKUs SKUs SKUs) are generating Turnover: 80.00% 69.87% of turnover. They 69.87% also represent 57.93% of 70.00% Discrepancies 60.00% Discrepancy: 50.00% 28.23% Discrepancy: 40.00% A more accurate 57.93% Discrepancy; inventory system will 30.00% 13.83% highly benefit the Fast 20.00% Movers! Turnover: 10.00% 25.06% Turnover: 5.07% 0.00% Discrepancy Contribution Turnover Contribution - 13 -
SKUs Clustering: ABC classification based on Discrepancy Discrepancy Discrepancy Class % of SKUs Contribution Mean € Min € Max € High Discrepancy 1.83% 69.92% 94.30 -14485.75 14163.01 Middle Discrepancy 17.29% 25.08% -1.07 -365.21 357.12 Low Discrepancy 80.88% 5.00% -0.41 -19.87 19.87 The Fast/High SKUs (Fast Less than 2% of But 45.6% of High Mover class for Sales and SKUs are generating Discrepancy Class High Discrepancy class for 70% of total also belong to the Discrepancy) needs a discrepancy Fast Mover Class careful analysis of inaccuracy sources and operations improvements - 14 -
SKUs Clustering based on the Discrepancy sign Discrepancy Sign SKUs Discrepancy Mean Discrepancy Min Discrepancy Max Class % € € € 1 4.03% 555.59 66.9 14163.01 2 4.46% 39.97 23.07 66.34 Discrepancy SKUs Discrepancy 3 4.91% 14.98 9.59 23.03 Sign % Mean € 4 6.22% 6.09 3.48 9.57 Negative 29.93% -80.79 5 51.88% 0.16 -1.28 3.47 Positive 24.44% 103.88 6 8.49% -3.25 -5.52 -1.3 Zero 45.63% 0 7 6.09% -8.31 -11.96 -5.53 8 5.21% -17.39 -25.06 -11.97 9 4.66% -39.38 -61.27 -25.07 10 4.05% -510.22 -14485.75 -61.44 Positive discrepancy is Discrepancy is not Positive or Negative not negligible and always negative. It is discrepancy: the generally is caused by not all about impact on sales is Information Systems Shrinkage the same manipulations and errors - 15 -
Comparison Test vs. Control The Turnover in the Test store is 5.22% higher than in the Control Store Turnover comparison Test vs Control Store Fast Mover Middle Mover Slow Mover 8.00% 6.11% 5.90% Fast and Middle Mover 6.00% SKUs which account for 4.00% more than 86% of 2.00% discrepancies seem to 0.00% benefit from the stock -2.00% audit taking place in the Test Store -4.00% -6.00% -8.00% -8.21% -10.00% - 16 -
Sales increase in the Test store, after the Audit, compared to the Control Store 6.00% 4.91% 5.00% All Discrepancy classes 4.00% 3.14% (High, Middle and Low) 2.52% 3.00% benefit from the stock 2.00% audit with a higher 1.00% benefit for the “High Discrepancy” Class 0.00% High Discrepancy Middle Discrepancy Low Discrepancy 12.00% 10.42% 10.00% 8.01% 8.00% Both positive and 6.00% 4.80% negative discrepancy 3.61% classes benefit from 4.00% 2.56% 1.91% 1.71% stock audit 2.00% 0.43% -0.22% -0.54% 0.00% Sign Sign Sign Sign Sign Sign Sign Sign Sign Sign - 17 - Class 1 Class 2 Class 3 Class 4 Class 5 Class 6 Class 7 Class 8 Class 9 Class 10 -2.00%
Experiment at Retailer BETA 1 1 1 1 4 1 1 1 1 1 1 week week week week week weeks week week week week week Stock Audit Physical Discrepancy: Computer Discrepancy: what was found vs what what was found vs what is Computer and Counted Physical stock during the should be in the stock shown in the computer audit are contrasted with what they should be given last week audit and before the audit stock movements (last counted physical stock + stock input – stock output) - 18 -
SKUs Clustering: ABC classification based on Turnover Turnover vs Discrepancy Contribution 80.00% 69.97% Fast Movers (14% of 70.00% SKUs) are generating 69.97% of turnover. They 60.00% 48.65% also contribute to 48% of 50.00% 29.23% Computer Discrepancies 46.14% 40.00% and 46% of Physical 29.05% Discrepancies 30.00% 20.00% 25.02% 22.12% 10.00% 24.80% 5.00% 54% of Fast Mover class 0.00% belongs to “High Physical Fast Mover Middle Mover Slow Mover Discrepancy” class. Computer Discrepancy Contribution Turnover Contribution 33% of Fast Mover class Physical Discrepancy Contribution belongs to “High Computer Discrepancy” class - 19 -
Correlation stock movement-discrepancy Average Physical Discrepancy as a function of the Stock Output 40 A Strong correlation between 35 y = 0.4037x + 6.0082 30 stock movements (Input and R² = 0.7622 25 Output) and the Discrepancies 20 (Computer and Physical) 15 10 5 0 ( 10) 0 10 20 30 40 50 60 Average Physical Discrepancy as a function of Stock Input 60 50 y = 0.4332x + 3.647 40 R² = 0.8274 30 20 10 0 ( 10) 0 10 20 30 40 50 60 70 80 90 - 20 -
Recommend
More recommend
