RFID-Enabled Visibility and Inventory Record Inaccuracy: Experiments in the Field Bill Hardgrave (presenter)John AloysiusSandeep Goyal

Information Systems DepartmentUniversity of Arkansas

Research QuestionsWill RFID technology improve inventory

record accuracy?

Can RFID technology ameliorate the effects of known causal predictors of inventory record inaccuracy?

What are the characteristics of product categories for which RFID technology is effective in reducing inventory record inaccuracy?

Hypothesis 1

RFID-enabled auto-adjustment will decrease inventory record inaccuracy over and above existing inventory management systems

Study 1 All products in air freshener category

tagged at case level Interrupted Time-series design Data collection: 23 weeks 13 stores: 8 test stores, 5 control

stores Mixture of Supercenter and Neighborhood

Markets Daily physical counts 10 weeks to determine baseline Same time, same path each day

Study 1 Results: Linear Mixed Effects (Pre-test/post-test comparison for test stores)

Results of Linear Mixed Effects

Variables Effect (Intercept) 8.004*** Sales Volume -0.953** Variety

-0.003

Item cost

-0.040* Dollar Sales 0.000 PRE

0.138**

TRANS

-1.875*** POST -0.345*** Notes: *p < .05, **p < .01, ***p < .001

Velocity = Number of units sold per day; Item Cost =Cost of an item in cents; Sales Volume = Item Cost X Velocity; Variety = Number of unique SKUs carried in a store; PRE: Periods numbered consecutively for 40 day

window around the adjustment; POST: Periods numbered 0 for 20 days before the adjustment, numbered consecutively after; TRANS: Numbered 0 before the adjustment, numbered 1 after

From DeHoratius and Raman (2008) Item level

Item cost Sales volume Dollar volume sales Distribution structure (fixed)

Store level SKU variety Audit frequency (fixed) Inventory density (fixed)

Study 1 Results: Discontinuous growth model (Interrupted time series for test stores)

Study 1 Results: Linear Mixed Model for Test versus Control stores

Variables Fixed effects (Intercept) 5.654*** Sales volume 2.356*** Variety 0.000 Item Cost 0.001 Dollar sales -0.002 Test -1.630** Period -0.008 Notes: *** p < .001, ** p < .01

Sales volume = Number of units sold per day; Item Cost =Cost of an item in cents; Dollar Sales = Item Cost X Velocity; Variety = Number of unique SKUs carried in a store; Test: Dummy variable coded 1 for test stores and 0 for control stores; Period: Day 1 starting when RFID auto-adjust was made available in test store.

Study 1: Discussion

• PI accuracy improved 23%•Results were essentially what we

expected• Insight from DeHoratius and Raman

(2008) variables•Raises the question: what about other

categories?

Hypothesis 2 (Study 2)

RFID-enabled auto-adjustment will ameliorate the inventory record inaccuracy due to high sales volume, low item cost, high SKU variety, high dollar volume of sales, and inventory density (across multiple categories)

PI: Perpetual Inventory

Study 2

• Untreated Control Group design with pretest and post-test • Matched Sample• 62 stores: 31 test stores, 31 control stores• Mixture of Supercenter and Neighborhood Markets• Spread across the United States

• Looked at both understated PI and overstated PI • Control stores: RFID-enabled, business as usual▫ Test stores: business as usual, PLUS used RFID reads (from

inbound door, sales floor door, box crusher) to determine count of items in backroom

Auto-PI: adjustment made by system For example: if PI = 0, but RFID indicates case (=12) in backroom, then PI

adjusted

PI: Perpetual Inventory

Study 2 (contd.)

• Five general merchandise categories• Floorcare• e.g., Powerforce vacuum, tough stain pretreat, Woolite gallon

• Air freshener• e.g., Glade plugin, Febreeze paradise, Glade oil

• Formula• e.g., Pediasure chocolate, Nutripal vanilla

• Ready to assemble furniture• e.g., computer cart, pedestal desk, executive chair

• Quick cleaners• e.g., wood floor cleaner, Readymop, Swiffer floor sweeper

PI: Perpetual Inventory

Study 2 (contd.)•Data collection

• Two waves (Pre and Post implementation), two months apart•Same time, same path each wave•Stock physical counts

• conducted over 5 days in each wave by an independent company•Dependent variable: PI Absolute = |PI – Actual|

•Looked at both understated and overstated PI

Pre-implementation Post-implementation

RFID Implementation

5 days 2 Months 5 days

Study 2 (contd.)

• Data collection (contd.): Measures• Item cost• Cost of the item to the retailer

• Sales volume• Quantity of item sold for two month preceding measurement

• Dollar sales• Dollar amount of items sold for two month preceding

measurement• Density• Total number of units in a category divided by linear feet of shelf

space for that category• Variety• Total number of unique SKUs in a category

PI: Perpetual Inventory

Study 2 Results: Ameliorating effects of RFID (Pre-test/Post-test)

PI~TREAT + COST + SALESVOL + DOLLARSA + DENSITY + CATVAR + TREAT_XXX

Treatment X Cost 0.0280 *Treatment X Sales Vol. -0.0092 ***Treatment X Dollar Sal. -0.0004 *Treatment X Density -0.0132 ***Treatment X Variety -0.0015 ***

*** p < .01, ** p < .05, * p < .10

Study 2 Results:Effect size for Treatment, Linear Mixed Model

PI = β0 + β1*Treatment

Category Control Stores Test Stores DifferenceFloorcare -0.208 * -0.899 *** 0.691**Airfreshener -1.099 * -2.729 *** 1.63***Furniture -0.061n.s. 0.168 n.s. -0.229 n.s.Formula 0.894 n.s. -2.004 *** 2.898 ***Cleaners 1.692 ** 1.319 ** 0.373**

1. *** p < .01, ** p < .05, * p < .102. Significance of difference assessed by

interaction term of treatment (pre-post) and group (test-control)

Study 2 Results:Characterization of Categories

CategorySales

Volume Item CostDollar Sales

SKU Variety

Inventory Density % Improve

Floorcare 16.06 20.21 366.52 736 24 45.15%**

Airfreshener 91.47 2.59 232.14 1123 224 29.56%***

Furniture 9.69 51.94 586.32 384 4 -60.64%n.s.

Formula 127.53 10.46 1499.14 282 130 81.60%***

Cleaners 80.46 6.14 559.47 120 72 16.86%**

*** p < .01, ** p < .05, * p < .10

Contributions

▫RFID technology with case-pack tagging demonstrated to improve inventory inaccuracy by 16% to 81% depending on category characteristics

▫Evidence that RFID technology is effective in ameliorating the effects on inventory inaccuracy of item cost, sales volume, dollar sales, density, and variety

PI: Perpetual Inventory

Contributions (contd.)

RFID technology is more effective in reducing PI inaccuracy in product categories which have: higher sales volume, lower item cost, higher dollar sales, greater SKU variety, greater inventory density

Future Research Directions

What is the economic impact of improving inventory accuracy (with RFID)?

Imagine inventory accuracy with item-level tagging …

Bill Hardgravebhardgrave@walton.uark.edu479.575.6099

John Aloysiusjaloysius@walton.uark.edu479.575.3003

Sandeep Goyalsgoyal@walton.uark.edu479.575.6961

For copies of white papers, visithttp://itri.uark.edu/researchKeyword: RFID

Business Problem and Motivation Perpetual inventory (PI) record

inaccuracy affects forecasting, ordering, replenishment PI is inaccurate on 65% of items (Raman et

al. 2001)

At any given time the retailer in this study manages about $32 billion in inventory

Scientific Motivation

Firms are skeptical about implementing new technologies based on pure faith, but need value assessments, tests, or experiments (Dutta, Lee, and Whang 2007)

Such empirical-based research requires “a well-designed sample, with appropriate controls and rigorous statistical analysis”

Research Model: Making the Business Case for RFID Technology

RFID Technology

Inventory Visibility

Inventory Record Inaccuracy

Costs/Profitability

Research Gap

Delen et al. 2007

Research Gap

There is little empirical research in the field that demonstrates and quantifies the ability of RFID technology to improve inventory inaccuracy

There is no empirical research that characterizes product categories for which RFID technology may be effective in reducing inventory record inaccuracy

Key Terms Inventory visibility

Retailer’s ability to determine the location of a unit of inventory at a given point in time by tracking movements in the supply chain

Inventory record inaccuracy Absolute difference between physical inventory and the

information system inventory at any given time (Fleisch and Tellkamp 2005)

RFID-enabled auto-adjustment A system that leverages RFID technology to correct for

the absolute difference between physical inventory and the inventory management system inventory at any given time

How does inventory inaccuracy occur?

Mechanisms which result in record inaccuracy

Results in overstated PI?

Results in understated PI?

Can case-level RFID reduce the error?

Incorrect manual adjustment

Yes Yes Yes

Improper returns

Yes Yes No

Mis-shipment from DC

Yes Yes Yes

Cashier error Yes Yes No

PI: Perpetual Inventory Source: Delen et al. (2007)

Research Gap There is evidence that RFID technology improves

inventory visibility

Researchers assume that improved inventory visibility will result in improved inventory record inaccuracy and consequently impact costs and profitability

The current research experimentally manipulates inventory visibility in field conditions (by means of an RFID enabled auto-adjustment system) in order to assess the effect on inventory record inaccuracy

Study 1 (contd.) Looked at understated PI only

i.e., where PI < actual Treatment:

Control stores: RFID-enabled, business as usual

Test stores: business as usual, PLUS used RFID reads (from inbound door, sales floor door, box crusher) to determine count of items in backroom▪ Auto-PI: adjustment made by system▪ For example: if PI = 0, but RFID indicates case

(=12) in backroom, then PI adjusted – NO HUMAN INTERVENTION

Read points - Generic Store

Backroom Storage

Sales FloorSales Floor

Door Readers

Backroom Readers

Box Crusher Reader

Receiving Door Readers

Study 1: Statistical Analyses Two comparisons:

Discontinuous growth model (Pre-test/Post-test)

PI = b0 + b1*PRE + b2*POST + b3*TRANS

Linear mixed effects model (Test/Control)

Random effect: Items grouped within stores

Statistical software: R

Hardware: Mainframe

Study 1 Results: Descriptive statistics (all stores, pooled across pre-test/post-test periods)

Variable Mean Std. Dev. 1 2 3 4 5 1. Sales Volume 1.13 1.18

2. Item Cost 171.89 75.71 -0.305** 3. Dollar Sales 21.78 20.26 0.650*** 0.125***

4. Variety 294.08 74.15 0.078*** 0.146*** 0.160*** 5. Treatment 0.52 0.50 -0.038 0.001 -0.076** 0.059***

6. PI-Inaccuracy 5.01 8.38 0.076*** -0.080*** 0.121*** 0.182*** 0.030 Notes:

*** p < .001, ** p < .01

Study 1 Results: Post Hoc Analysis

Study 2: Statistical Analyses Comparisons:

Linear mixed effects model (Pre-test/Post-test)

Random effect: Items grouped within stores

Statistical software: R

Hardware: Mainframe

Study 2 Results: Descriptive Statistics

Mean Std. Dev.

1 2 3 4 5

1 PI_ABS 3.16 11.382 Cost 47.99 11.96 -.049*

*3 Category

Variety795.31 464.01 .015** -.198**

4 Sales Volume 52.40 184.95 .400** -.032**

-.037**

5 Dollar Sales 735.31 2786.83

.201** .356** -.177** .648**

6 Density 100.84 93.10 .159** -.217**

.263** .170** -.114**