Slide1: Resource Allocation and Contract Pricing Brenda Dietrich


Slide2: Motivation Examples Overnight Delivery Business Recovery Desk-side Service IT hosting Product Repair/Upgrade Services Known factors that could/should be considered Scale, Variability, Predictability Issues and Opportunities Outline


Slide3: Contracts, especially contracts for services, are becoming increasingly complex bundling of goods and services delivery over time/space base cost, plus numerous adders fixed + variable beyond some limit rebates The total value of the contract includes factor not explict in the contract "drag" after sale service and parts PR Motivation


Slide4: Goods or services are to be provided over time on a specified schedule e.g. PC's, some application hosting as required business recovery, warranty service as needed by your customer's customer web catalogues, video on demand application hosting transportation Contract may specify min and max, but allow flexibility in timing mix location Services contract delivery terms


Slide5: Contracts for services typically specify service levels and penalties for not meeting them Multiple service levels, with higher unit cost for premium services Penalties for missed service average service level (per defined measurement period) penalty per miss above threshold (per period) Improvements over past performance, or against industry benchmarks Services are marketed based on reliability and flexibility, as well as price Contracts may specify "entitlements" per measurement period surplus may be forfieted additional charges for exceeding entitlement in supply chain "flexibility" may be a specified entitlement. Service levels and entitlements


Slide6: Fulfillment requires the deployment of resources, but there are choice Which resources to use When to use them Typically multiple contract are being fulfilled from the same pool of resources When there are no scare resources: serve all contracts at maximum profit or minimum cost When there are shortages: select which penalties to incur or which requests beyond entitilemnt to serve to maximize profit define "profit" - current period, long term expected, ?? Resource allocation decisions can impact profit in operations. Can resource allocation tools help in pricing? Role of Resources


Slide7: Motivation Examples Overnight Delivery Business Recovery Desk-side Service IT hosting Product Repair/Upgrade Services Known factors that could/should be considered Scale, Variability, Predictability Issues and Opportunities Outline


Slide8: TransportationContract Pricing Problem: Determine per lane, per unit price for a long term contract, involving large, but unknown volumes and multiple lanes, and service guarantees Tools: Customer demand models and resources optimization Model the demand variability by day, based on similar customers Evaluate many scenarios of demand to determine peak and likely resource requirements Rapidly solve large optimization models representing allocation of resources against demand scenarios Benefits: Accurate cost models for pricing of contracts, and fast evaluation of opportunities. expected cost, confidence intervals, identification of risks Issues: Cost of new contract depends on current contract base


Slide9: Business Continuity & Recovery Services BCRS - insurance for computer facilities: If customer's computer facilities are unavailble, IBM provides comparable facilities for them Also provides data vaulting, copies of application software, and testing facilities Cost of serving a customer depends on their equipment their applications their geographic location and the pool of customers already under contract Pooling and redeployment of IBM resources is possible, given adequate leadtime.


Slide10: Business Continuity & Recovery Services The system is modeled as a queueing system with 4 arrival streams hurricanes, earthquakes, floods, power outages. Can compute the inventory risk exposure Can Compute the target inventory levels Can compare new customer to existing pool (.7x, 12.5x) Can compare risk and inventory levels with and without new customer # of Computers Exceedance Probability (in percentage) Client server DEC IBM Large IBM Midrange ------------------------------------------------------------------------------ 1 47.60 3.20 91.72 98.28 2 30.64 1.44 77.64 96.72 ...... 6 6.28 0.12 25.08 71.08 7 4.72 0.12 19.96 61.16 8 3.44 0.04 16.60 52.68


Slide11: Deskside Support Subject to Multiple SLAs Each arriving call has a specific Service Level attached to it. Contract specifies response times (x) and service levels (y) for each class of calls SLA specified as: Probability{time to service the call= %y. Penalties associated with not acheiving specified service levels Cost is propostional to number of service technicians Historical data on calls, some limits and averages specified in the contract Allocation rules play a signifcant role in determining technician requirements dedicated "gold service" techs vs adaptive re-allocation. solvable with queueing model Extensions multiple skill classes geographies - location insert delay into schedulingt


Slide12: Large enterprises haveing 100's of IT applications seek to reduce costs by outsourcing the operation of some of these applications For each application know: HW requirements Skills Time Zone, Risk, Security needs, etc For each location know: HW availability and operating costs Skill availability anc costs Time Zone, Risk level, Security level, etc Some linkages between applications Hosting proposal includes proposed locations for each application and total cost IT application Hosting


Slide13: IT application Hosting


Slide14: Spare-Parts Planning Challenges: Develop model and solver to minimize inventory and transportation costs, and satisfy stringent time-based service criterion Replace hierarchical model with one that emphasizes rapid delivery Solver must execute within 1 day (on a full national problem instance) to work within a 1-week planning cycle Cost of fulfillment depends on inventory costs and service technician costs Profit is made by amortizing these costs across multiple customers with similar requirements


Slide15: Contract duration relative to delivery period Varibility of demand mix, location, timing Predictability of demand how long is the "prediction" interval relative to the "deployment" interval can prediction be improved Resource flexibility Resource uniformity Switching costs for redeployment of resources QOS Measurement intervals Partial list of factors to consider


Slide16: Disconnect between sales and fulfillment inconsistent cost assumptions inconsistent resource availability assumptions inconsistent assumptions on resource allocation Role of demand forecast in pricing a contract Must each contract "stand alone" as profitable what is services analogy of "available to promise" "capable tp promise" "available/profitable to sell" Issues and Opportunities