The Impact of Item-level RFID on Product Availability in The ...

IntroductionContribution Of This Research

Summary

The Impact of Item-level RFID on ProductAvailability in The Retail Supply Chain

Gary M. Gaukler

Department of Industrial and Systems Engineering, Texas A&M University

INFORMS Seattle 2007

Gary M. Gaukler Impact of Item-level RFID on Retail


Summary

Outline

1 IntroductionItem-level RFID in SCMPrevious Research

2 Contribution Of This ResearchModel DescriptionResearch QuestionsResults



Summary

Item-level RFIDPrevious Research

Current Applications of RFID in SCM

Areas of implementation:

TransportationLogisticsWarehousing

Focus of improvements:

EfficiencyInventory record accuracy

Usually using RFID on case- and pallet level



Summary


Item-level RFID Will Become More Prevalent As RFID TagCost Decreases

Compare RFID tag prices (UHF passive)

4 years ago: $0.50Now: $0.15

Crucial to understand potential benefits of item-level RFIDMajor potential benefit: Retail operations

Preventing out-of-stock situations at the shelf (OOS)



Summary


Focal Points Of This Research

Retail supply chain - 1 manufacturer, 1 retailerLost salesItem-level RFIDModeling benefits from item-level RFID:

Improved in-store shelf replenishmentFewer out-of-stock situationsMore accurate demand forecasts



Summary


Previous RFID Research

Inventory accuracy:

Heese (2007)Atali, Lee and Ozer (2007)Kang and Gershwin (2005)

Inventory control policies:

Gaukler, Ozer and Hausman (2007)

Retail:

Wong and McFarlane (2003)Karkkainen (2003)Gaukler, Seifert and Hausman (2007)

This is just a small selection. For a more comprehensive overview,see Lee and Ozer, 2007.



Summary


Previous RFID Research

This research is closest to Gaukler, Seifert and Hausman (POM16(1), 2007).

GSH 2007:

Single period modelNo demand forecastupdatesImpact of RFID:

fewer OOS

This research:

Multi-period modelDemand forecasting basedon past salesImpact of RFID:

fewer OOSbetter demandforecasting



Summary

Model DescriptionResearch QuestionsResults

Backroom Vs. Shelf Stock

Manufacturer Retailerorder

deliver

Backroomstock

Shelf stock

replenish

RFID RFID



Summary


Customer Demand At The Shelf

Customer demand arrives according to Poisson processEvery demand arrival may see a shelf that is stocked, or emptyIdea: Remove those demands from the stream that correspondto an avoidable lost sale

Avoidable lost sale: the shelf is empty, but the backroom is notEffective demand = original demand less avoidable lost sales

Define 0≤ θ ≤ 1: probability that incoming demand sees anon-empty shelf, given ample backroom stock

θ determined by effectiveness of backroom-to-shelf process,misplacements, ...θ = 1: lost sales only occur if backroom stock exhausted

Effective demand can be approximated by a central limittheorem and expressed as a normal distribution N(θ µ,

√θσ)

(see GSH 2007)RFID: θR ; non-RFID: θN ≤ θR



Summary


Timeline Of Events

In each period, the timing of events is as follows:

1 Retailer establishes demand forecast Di based on Di−1 andprevious sales

2 Retailer calculates backroom stocking level Si and ordersquantity [Si −onhand inventory]+ from outside supplier.Product arrives instantaneously.

3 Real demand over the demand period is D and effectivedemand at retailer’s shelf over the demand period is Dθ

4 Retailer sells Vi over the whole demand period5 Retailer calculates profit for demand period i



Summary


Demand Forecasting

Demand forecasting done by one of three methods:

Standard one-equation exponential smoothingExponential smoothing with adjustment for censored demanddistributions (Nahmias 1994)Perfect forecasting: forecast = real demand (Benchmark)

Nahmias (1994) bases forecasts only on those salesobservations that are not censored by the available inventory,i.e. sales less than Si .



Summary


Inventory Policy

Periodic-review order-up to policy (for backroom stock)One-period profit function for period i :

πi = (r − c− t)E [Vi ]−h(Si −E [Vi ]) (1)

where E [Vi ] =∫ Six=0 xf θ (x)dx +

∫∞

x=SiSi f θ (x)dx

If retailer knew demands in each period:

Optimal order-up to level: S∗ = Si = θ µ +√

θσz , for all i ,where z is defined as Φ(z) = r−c−t

r+h−c−t

But in our model retailer forecasts demand:

Plausible order-up to levels: Si = θ µi +√

θσiz , where z asaboveOrder-up to level dependent on retailer’s demand estimate



Summary


Effects Of RFID in This Model

Effects of RFID:

Direct effect:

“Shelf RFID”Sales of product increasedue to fewer OOS

Indirect effect:

“Backroom+Shelf RFID”Changes to backroomstocking levels

Indirect effects are driven by updated demand forecasts



Summary


Research Questions

1 What is the relative magnitude of direct vs. indirect effects?2 What kinds of products are most conducive to attain benefits

under RFID?

High-margin?High-variability?

3 What are tag cost levels that allow for a profitable item-levelRFID implementation?



Summary


Profit Improvement Vs. Tag Cost

Figure: Profit Improvement vs Tag CostBackroom+Shelf RFID: Profit Improvement vs.

Tag Cost

-8.00%

-4.00%

0.00%

4.00%

8.00%

$0.00 $0.01 $0.02 $0.05 $0.10 $0.20 $0.30 $0.50

Tag Cost [$]

Perfect Forecast Sophisticated Forecast Simple Forecast



Summary


Profit Improvement Vs. Tag Cost (II)

The analysis used θR = 0.97, θN = 0.92The analysis does not consider fixed costs of RFID

Can include these in NPV analysisClearly, fixed cost per product decreases as sales rate increases

Insights:

RFID implementation most conducive to fast-selling productsThe more sophisticated the existing forecasting methods are,the harder it is for RFID to be profitableBreak-even tag cost seems readily attainable (excluding fixedcosts)



Summary


Direct Vs. Indirect Effects

Figure: Profit Improvement vs. Tag Cost t and RFID Effectiveness θR

$0

.00

$0

.01

$0

.02

$0

.05

$0

.10

$0

.20

$0

.30

$0

.50

0.90

0.94

0.99

-12.00%

-8.00%

-4.00%

0.00%

4.00%

8.00%

Tag Cost [$]

R

Shelf RFID: Profit Improvement vs. Tag Cost and R

(a) Shelf RFID

$0.0

0

$0.0

1

$0.0

2

$0.0

5

$0.1

0

$0.2

0

$0.3

0

$0.5

0

0.90

0.94

0.99

-15.00%

-10.00%

-5.00%

0.00%

5.00%

10.00%

15.00%

Tag Cost [$]

R

Backroom+Shelf RFID: Profit Improvement vs. Tag Cost and R

(b) Backroom+Shelf RFID



Summary


Direct Vs. Indirect Effects (II)

Observations:

Additional benefit from updating the demand forecast can besubstantialAn RFID implementation that updates demand forecasts ispotentially substantially more profitable than an RFIDimplementation that does not

Insight:

RFID most useful for products that are stocked continuallyover many replenishment periodsLess beneficial for one-season, promotional products that mayonly be stocked once



Summary


RFID Benefits Vs. Profit Margin

Figure: Profit Improvement vs. Profit Margin; Simple Expo SmoothingProfit Improvement vs. Profit Margin

-8.0%

-4.0%

0.0%

4.0%

8.0%

90% 80% 70% 60% 50% 40% 30% 20%

Profit Margin

Pro

fit

Imp

rov

em

en

t

0.0%

20.0%

40.0%

60.0%

80.0%

100.0%

% d

ue B

ackro

om

% due Backroom Shelf RFID Backroom+Shelf RFID

(a) Profit Improvement



Summary


RFID Benefits Vs. Profit Margin (II)

Without adjusting backroom stocking: (“Shelf RFID”):

Sales and profit improvements highest for high-margin products

With adjusting backroom stocking (“Backroom+Shelf RFID”):

Sales and profit improvements highest for low-margin productsReason: censored demand forecasting (simple expo smoothing)

Insight:

In real-world RFID implementations, a paradoxical effect maysurface: low-margin products (typically B and C items) mayshow more improvement than high-margin productsWhen sophisticated forecasting techniques are used thatcorrectly account for censored demand observations, maximumRFID benefit is obtained for high profit margin products



Summary

Summary

Accounting for indirect effects of RFID is crucialMulti-period model captures the true value of RFID inreducing OOS, which the one-period model could notExpected benefits from RFID are - somewhat surprisingly -very dependent on the type of forecasting algorithm usedPopular wisdom says to implement item-level RFID first forhigh-margin products. This is only true if sophisticatedforecasting methods are already in place. If not, the reversemay be true



Summary

Outlook on Future Research

Perform analysis of the decentralized system

Market power structure, incentives for RFID deployment, ...Include substitutable products

Compare item-level RFID system to a case-level RFID system

Perpetual inventory, POS information



Summary

Thank you

Questions?

Contact information:

Dr. Gary M. [email protected] and Systems EngineeringTexas A&M University


The Impact of Item-level RFID on Product Availability in The ...

Business

salesimpact of rfid

gaukler impact of item

potential benefits of

original demand

shelfcustomer demand

noteffective demand

shelf oosgary

store shelf replenishmentfewer