Pennsylvania Profitability Analysis Identifies Individual High-Value Electric Utility Customers With 83-96% Accuracy

Summary

While most electric rates reflect the cost-of-service incurred by an entire customer class throughout the year, individual customer diversity in diurnal and seasonal energy use means that some customers in the same customer class contribute more to revenue than they cost to serve while others contribute considerably less.

Profitable customers will be the most attractive targets for energy suppliers in competitive markets because there is more leeway to offer price breaks to win new customers while still maintaining a profitable customer relationship. Target marketing will simultaneously make these high-value utility customers also the most at-risk customers. The extent to which energy suppliers can "cherry pick" these most desirable utility customers depends largely on their ability to identify and influence these profitable customers.

This article establishes the importance of identifying individual high-profit customers even before markets open to competition. Cost-of-service and profitability analyses are conducted for a sample of Pennsylvania retail store customers. This analysis reveals inter-customer cost-of-service variations of more than 2 cent/kWh and profit variations of more than 3 cents/kWh. In more concrete dollar terms, untargeted marketing to commercial customers in the 1.5 - 2.0 million kWh size is just as likely to net a $ 10,000 annual profit as it is to net a $10,000 annual loss. Fortunately, scoring models can be used to improve the odds of selecting profitable customers to a range of 83-96%.

Profitability

Customer profitability is the difference between revenue received and the cost of providing products and services. Transmission, distribution and stranded costs are not included in this analysis because these costs will be assessed by the local utility regardless of the energy supplier. This analysis focuses entirely on cost and revenue from electricity sales excluding related products and services.

From an energy supplier's perspective, the cost of service for each individual customer is determined by the customer's hourly energy use profile and the portfolio of energy and capacity contracts and/or generation capacity used to meet that customer's energy demand.

To provide a more general analysis, we estimate energy supply costs from publicly-available data sources. Electricity costs are composed of variable and fixed costs; variable costs include operating costs associated with the generation of electricity (e.g., fuel costs, maintenance) and fixed costs cover additional costs associated with generating equipment. Variable costs for Pennsylvania customers are estimated with PJM market-clearing electricity prices. Fixed capacity costs are estimated from utility rate schedules excluding transmission, distribution and stranded costs components (rate structures are designed to provide revenue to cover variable and fixed costs).

Revenue generated by each customer is determined by applying a Pennsylvania electric rate schedule to each customer's monthly electricity.

Pennsylvania Customer Data

Profitability analysis undertaken with market-segment average load profiles provides relatively little value because profit variations of customers within a single market segment are typically much greater than profit variations that occur between the market segments. Profitability analysis requires detailed information on a sample of individual customers.

A representative sample of commercial sector customers was extracted from the Pennsylvania MAISY State-Level Utility Marketing and Hourly Loads Databases. A single business category was selected for the analysis in part to omit obvious cost-of-service variations that exist between different business activities (e.g., hospitals and retail stores). The objective of this analysis is to evaluate the more subtle differences that exist among customers in similar market segments. The small and medium retail store business market segments were selected for analysis because the average cost-of-service and load characteristics of these segments are consistent with the entire Pennsylvania commercial sector; consequently, the results of this more focused analysis have applicability to the entire commercial sector. Retail stores include building and garden supplies, department stores and other general merchandise stores, auto dealers, apparel, furniture, drug and miscellaneous retail activities. Retail stores housed in shopping malls are excluded (Retail Malls is a separate MAISY business category). Customers in this market segment range from small retail stores (3,000 square feet, 15000 kWh is the smallest) to 300,000 square feet, 2 million kWh department stores (by way of reference, the typical new Wal-Mart store is about 100,000 square feet).

Information provided in the database for each customer includes detailed building, occupant, equipment and operating characteristics as well as annual energy, energy use by end use, and hourly loads.

Cost of Service

PJM market clearing prices from April 1997 through March 1998 are used to estimate variable energy supply costs. Prices vary significantly across hours of the day and by season. Figure 1 shows average market-clearing prices for July and October.

Hourly loads for each of the individual Pennsylvania retail store customers in our sample are multiplied by market-clearing prices to compute the variable cost of supplying electricity to each customer. These costs are presented in Figure 2 where each dot represents a single customer. In terms of variable costs, customer size (i.e., total energy use) suggests somewhat greater cost-of-service for larger customers; however, variable costs differ by as much as 0.3 cents per kWh regardless of annual energy use (Note: in Figures 2-4, lower cost customers are shown in the upper portions of the graphs, ie., as negative cents/kWh).

In competitive energy markets, the price of contracted capacity reflects both daily demand requirements and the volatility of daily demand. A simple algebraic representation of both factors is provided in the following cost of capacity (C_c) relationship:

C_c = P_kW*kW_av + P_vol*(kW_peak-kW_av) (1)

where P_kW= capacity prices faced by the energy supplier, kW_av= average daily demand, P_vol = a volatility price premium and kW_pk = monthly peak demand .

Since both commercial class customers and small-medium retail stores, on average, return a profit close to zero, total revenues (determined from energy and capacity charges of a typical 1998 Pennsylvania commercial rate structure) are used along with total variable costs computed above to calibrate the cost of service equation (1). Figure 3 shows customer differences in the cost of capacity, C_c ($/kW) with respect to average monthly peak kW. No obvious correlation exists between capacity costs and customer size.

Figure 3 illustrates the fact that there is considerable variation in monthly peak demand volatility across retail store customers. While capacity costs vary with alternative specifications of equation 1, most results are similar to those presented in Figure 3. (In some short-run situations, for some energy service providers, capacity costs can be negligible. However, for any significant period in a growing market, there must be some return to generating capital represented here as capacity costs).

Figure 4 shows total cost-of-service (i.e., total variable plus capacity costs) of serving individual Pennsylvania small and medium retail customers graphed against customer kWh.

Estimating Profitability

The commercial class rate structure used in the previous section is applied to customer energy and hourly loads to compute revenue from each customer; the cost-of-service data developed in the preceding section is then subtracted from revenue to derive customer profit.

Figure 5 shows customer profitability in cents/kWh units graphed against kWh. This chart reveals a lack of correlation between customer size and profitability. Figure 6 presents annual profitability, in dollars, for these same customers. This figure shows that marketing to commercial customers in the 1.5 - 2.0 MWH size is just as likely to net a $10,000 annual profit as it is to net a $10,000 annual loss.

Scoring Cost-of-Service and Profitability

Given the variation in profitability shown in Figure 6, the critical question is: can lower cost and more profitable retail stores be identified prior to their retention as customers?

The MAISY Databases provide extensive information on customer building, equipment, occupancy and DSM information for each customer (i.e., each point) in the preceding figures. Cost of service and profitability data on each customer are statistically related to customer characteristics in scoring models which are then used to identify other low-cost, high-profit customers. Scoring models developed in this study relate customer cost-of-service and profitability to easily-measured customer characteristics such as billing kWh, kW, floor space, space heating fuel and several other variables.

Table 1 below shows the probability of identifying low-cost small and medium retail sales customers with untargeted marketing programs and as well as the probability of identifying low-cost customers with scoring models.

The first row of data in Table 1 shows that an unfocused marketing program will gain profitable customers 66 % of the time in the small retail store market and 45% of the time in the medium retail store market. The second row in Table 1, however, shows that these odds can be improved to 91% and 83% by employing the results of scoring models to identify customers with the most attractive cost-of-service characteristics.

Table 1. Probability of Gaining Lower-Than-Average-Cost Customers With Untargeted and Targeted Marketing Programs

Small Retail Customer

Medium Retail Customer

Untargeted

66%

45%

Targeted

91%

83%

Similar results are obtained when focusing on profitability as indicated in Table 2. The odds of finding a profitable customer are only about 50% with untargeted programs; however, applying the results of profitability scoring models increases those odds to 84% and 96% for the small- and medium-sized customers, respectively. The implications, in light of Figure 6, are especially clear for larger customers: using the scoring models almost guarantees the selection of the individual $7,000 - $10,000 profit customers over customers who represent losses of $8,000 - $10,0000 each.

Table 2. Probability of Gaining Profitable Customers With Untargeted and Targeted Marketing Programs

Small Retail Customer

Medium Retail Customer

Untargeted

46%

54%

Targeted

84%

96%

Other Marketing Objectives and Analysis Issues

Analysis in preceding sections focuses only on cost-of-service and profit in providing the electric energy commodity to customers. Energy conservation, telecom, consolidated billing, security and other services can be evaluated in an identical manner using MAISY data in an integrated analysis.

While not addressed in this application, strategic pricing analysis should also be integrated with cost-of-service and profitability analysis.

Metering Versus Local Utility Generic Load Profile Costs

Most small and some medium-sized utility customers do not currently have time-interval meter technology required to provide individual customer cost-of-service settlements with the local distribution utility. For these customers, energy suppliers have the option of installing a time-interval meter or accepting the local utility's generic customer-class load profiles to estimate customer cost-of-service. Any currently unmetered customer expected to be less costly to serve than indicated by the relevant utility load profile can be offered a more competitive electricity rate and then metered with a time-interval meter to avoid the generic load profile assessment. Amortizing time-interval metering costs, including the potential benefits of new customer services possible with new metering technologies, makes this an attractive option at even modest cost-of-service differentials.

Similarly, allowing more-costly customers to be assessed at lower local utility load-profile rates shifts commodity costs to the local utility while providing cross-selling opportunities in telecom, security, and other services of the new energy supplier.

Using Scoring Model Information in Marketing Programs

Analysis results reported here for retail stores are similar to those conducted for other market segments. The most striking conclusions to be drawn from these results is that traditional energy sector marketing approaches which use customer contact databases and then target customers by SIC codes and size measures (e.g., number of employees or estimated energy use) can expect, on average, to have an equal mix of profitable and unprofitable customers while an application of the information developed in scoring models can typically increase the percentage of profitable customers to the range of 85 - 95 percent.

Scoring models using information on as few as three or four general customer characteristics are often sufficient to distinguish profitable from unprofitable customers. Applying the scoring models to individual customers requires information on the scoring variables for each customer. Some of these variables may already exist in customer or marketing information systems. Customer values of these variables may also be identified as part of a customer contact process including inbound and outbound telesales, direct marketing mail-in responses as well as proprietary data sources.

MAISY Energy Marketing and Hourly Loads State-Level Databases

All cost-of-service and profitability analysis and scoring model estimation were developed with detailed individual Pennsylvania retail store customer data from the MAISY Energy Marketing and Hourly Load Databases.

MAISY State-level Energy Marketing and Hourly Load Databases include extensive building structure, end-use energy use, hourly loads, equipment, conservation, and other energy-related variables along with traditional economic and demographic information for electric and gas utility customers throughout the US.

MAISY clients include more than 90 electric and gas utilities, power marketers, energy service companies and other energy-related organizations throughout the US and Canada. A sample of MAISY clients include Cinergy, Southern Company, PG&E Energy Services, DukeSolutions, New Energy Ventures, Entergy, and Enron.