Theft of service has been discussed
more in the boardroom in the last
three years than it has in the past
30. Why all the new interest?

In short, the boardroom is responding
to its perception of how some consumers
may react to skyrocketing energy and
resource prices. Unprecedented rates
are pushing more consumers than ever
before to resort to actions like stealing
service. There is even word of an emerging
market for professional “service tamperers,”
who are paid to assist consumers
in bypassing metered service.

Skyrocketing prices aren’t the only factor.
Sarbanes-Oxley, competition created
through deregulation, and the public and
state regulator’s demand for better corporate
governance all contribute to making
theft of service a 20th-floor topic.

No territory is immune and no service
is safe from theft. With a little knowledge,
a little nerve or the right contractor, anyone
can tap into electric, gas and water
services. The majority of theft occurs in
the residential sector, but the majority of
all revenues lost, estimated between 1 and
3 percent of total distribution revenues,
occurs in the commercial account sector.
A utility with $1 billion in revenues potentially
loses between $10 million and $30
million each year to theft, and more than
two-thirds of that loss is within the relatively
small commercial account sector.

We will focus on commercial accounts
to discuss how and where the most significant
theft occurs and what leading North
American utilities can do and are already
doing to curb such illicit activity.

Stealing Around the Meter

The smarter thieves do not steal all of
the service or tamper with the meter
directly. This is an ill-conceived act with
short-term payoff, since it’s likely to generate
a zero-read event at the utility.

Locking and sealing programs, solidstate
meters and automated meter reading
(AMR) tamper flags all help protect
the meter but not the energy. A smart
thief knows this and tampers around the
meter instead.

AMR tamper flags do not detect
bypasses that divert energy around
the meter, while locks and seals do not
prohibit theft outside of the meter box.
Analysis of theft cases among Detectent’s
AMR-enabled customers between
January and December 2006, revealed
that most thefts occur without any corresponding
tamper flag. Moreover, the
majority of tamper flags triggered don’t
even indicate a theft case or any other
spurious activity. Most of them are the
result of normal daily activities, such as
contractor-induced outages and maintenance
service calls, as well as external
factors like vibrations from nearby
machinery. The enormous volume of false
alarms tends to minimize the impact of
those few valid tamper flags.

So the question is, how can a utility
defend against theft of service when
the thieves are getting smarter and the
technology deployed does not address
the major losses that occur outside of the
meter box?

Protect the Service, Not the Meter

Service protection is not achieved through
locks and seals, although a locking and
sealing program is an important component
of meter security.

The service can only truly be protected
by recognizing anomalous consumption.
To do this, however, we first have to
understand what consumers do with the
energy and resources they receive. Then
models can be developed to establish
expected consumption patterns. Erroneous
consumption patterns can then
be detected as deviations from normal
expected behavior. As a result, both theft
around the meter and direct tampering
can be exposed.

Knowing the Consumer

Customer information systems (CIS) were
designed to facilitate customer identification
and business transactions such as
billing. They weren’t designed to store
the variety of data involved with getting
to know the customer better. A great
illustration of this is the classification of
customers by standard industry code. A
standard industry code of “full-service
restaurant” provides no indication of how
a restaurant uses energy and resources
to fulfill the needs of its customers. With
such simplistic coding, “full-service restaurant”
can include everything from
a large steak and seafood house to a
small sandwich deli. Both use energy and
resources in completely different ways to
serve their customers.

If, however, the data indicated that
the full-service restaurant was actually
a pizza parlor with an eat-in area, an
extrapolation on the energy needs of that
restaurant could be made. For instance,
you would expect to find a certain number
of pizza ovens to meet the demands of a
certain dining-area capacity. You would
expect seating areas to be air-conditioned
in the warm months and heated in the
cooler months. You would also expect
refrigeration in scale with the number of
potential patrons served.

One approach to protecting service collects
all of this information to create peer
groups and predictive models.

Consumer Modeling and Peer Grouping

Basic consumer models take into consideration
the standard industry code, but
more advanced models expect that code
to be broken down into greater detail.
These models are designed to understand
energy and resource consumption needs
based on in-depth knowledge of expected
usages. For the example shown in Figure
1, the broad category of full-service restaurant
is refined by cuisine, then by a
sub-type and, finally, into groups depending
on the environment where the food
product is served. When you break data
down into subcategories, you can determine
that the consumption of energy for,
as an example, a takeout pizza restaurant
is significantly different than that of an
eat-in pizza restaurant.

The addition of chain affiliation can
make these models even more precise
in determining expected energy and
resource usage, because most locations
of the same chain will have standard
equipment specifications. In Figure 2,
expected consumption varies not only
by the restaurant environment but also
by the national chain that dictates the
installed equipment specification.

Applying the Logic to Other Business Types

The restaurant example may seem like
an obvious case for consumer modeling,
but in reality, consumer modeling can be
applied to all business types. Hotels and
motels can be classified by number of
rooms and leisure and business amenities
provided. Gas stations can be classified by
complimentary services such as car wash,
shopping and restaurant facilities.

Gathering the Data

The obvious next question then is, if the
data does not reside in the customer
information system, then from where does
all this information about the consumer
come? It might surprise you to learn that
the information is gathered from a huge
variety of publicly available sources.

The Internet offers access to vast
repositories of information on commercial
consumers. These are commercial
entities that can create good will, and
benefit in other ways from having the
public know about them. All we have to
do is know how and where to obtain that
information for our own purposes.

There are numerous databases you can
purchase that offer a wealth of information
about commercial consumers, such
as type of business, products and services
offered; environment in which the
products and services are offered; size
of business; location; number of employees;
hours of operation; sales volumes;
and so forth. This information, once
paired with data from the utility’s own
computer information system, provides
quite a clear picture of how commercial
consumers utilize the energy and
resources delivered to them.

Establishing Peer Groups

Gathering data on a single consumer is
only valuable if there is a like set of data
on similar consumers available for purposes
of comparison. Therefore, peer
groups must be established and similar
data elements must be gathered for each
account in a peer group.

Peer groups can be developed for each
service territory as well as nationally. Geographically,
local peer groups may represent expected energy and resource usage
more reliably, but national peer grouping
offers a larger base with which to make
comparisons, develop trends and establish
expected norms. Figure 3 shows plots of
consumption for various businesses of
the same type against an expected norm
calculated from the peer group’s data. The
expected norm is illustrated by the dark
magenta line and outliers are considered
those that do not follow the same pattern.
Actual monthly consumption is not a significant
factor, as this particlar grouping is
not concerned with service capacity.

Eliminating False Alarms

It’s possible to become overwhelmed with
increases in false alarms as the volume
of data being processed increases. This
is especially true when comparing businesses
based on expected similarities. Any
number of things can change a business’s
consumption pattern such that it appears
anomalous – changes in ownership or
management, a shift in operating hours,
an incorrect business classification, multiple
meters under different
names, seasonal conditions,
closures for remodeling, even
the simple act of upgrading
equipment. The key to
successfully identifying
anomalous consumption is to
filter out these false alarms
through a process of screening
and analysis.

Before a seemingly anomalous
account is identified for
investigation, it gets reviewed
and verified for accuracy and
any potential changes in how
business is conducted. This
process tends to catch most of the false
alarms and eliminates false cases from
investigation. Furthermore, it corrects the
account’s peer grouping to ensure integrity
in other comparisons.

This process is illustrated in Figure 4.
The confidence factor on the right side
indicates the likelihood of a theft being
discovered. The more thorough the
screening and analysis, the higher the
likelihood of success in the field.

Other Sources of Data

Another significant data source is to look
at other services delivered. Whether
delivered by one utility or by separate
utilities, multiple service data can add
tremendous value to the analysis process.
A great example would be a Laundromat.
A Laundromat typically has gas dryers
and electric washers. If we know the
usage of either gas, electric or water, then
consumption projections can be made for
the other services too. It is very safe to
extrapolate this way because most Laundromat
customers will use an amount of
water and electricity in direct proportion
to the amount of gas used for drying. If
one of the services is not metered properly,
it will show up as anomalous based
on its ratio to the other metered services.
Even if all services had been tampered
with, the ratio of stolen service is not
likely to be consistent, so it would still
appear as anomalous usage.

These same energy use ratios can
be developed and modeled for almost
all business types, whether they’re
restaurants, hotels, service stations, public
facilities, etc. Energy use ratios have
consistently identified significant theft
cases that date back years and would otherwise
have remained undetected.

Summary

Theft is not going away and is only likely
to increase in the coming years. That’s the
bad news. The good news is that there are
effective tools for identifying anomalous
usage patterns. Data gathered in different
service areas can be used to build
consumer models and peer groups. By
integrating data from AMR tamper flags
with those models, valid theft cases can be
identified more accurately.