🚀 go-pugleaf

Many thanks to those people who responded to the question I posted last week.
This posting provides a summary of the responses.  Any additional comments 
are appreciated

------------------------------

The question was as follows,


>I need help on the following: I need to know how certain organizations charge
>for use of their network gateways (gateways join two systems/networks
>together, usually at the midrange or mainframe level e.g. at the Front-end
>Processor). For instance, do they charge by the number of sessions a client
>signs up for, by the number of sessions the client actually used, do they have
>a one-time membership fee with a smaller monthly maintenance fee which allows
>users a large but undefined number of sessions, do they not charge at all and
>if so then what is their justification? All these questions ask: HOW is the
>gateway charged. But I also need to know, if possible, WHAT the charge is
>based on: what cost items do they include in maintenance of the gateway
>and what percentage of these cost items is used. Cost items include: human
>resources, CPU time, memory, software development, and other overhead items.


>Any ideas from anyone who is familiar with this area would be appreciated.
>Any information on books or papers covering this topic, or firms doing similar
>work, would also be appreciated.
>


------------------------------------

Mr. Vielmetti writes,

we run an internet gateway here.

we charge what the market will bear :).  that means some flat rate
services, and some metered services.  in any case there is a fixed monthly
charge and (if it's metered) a charge based on connect hours.

the idea is to give people proper incentives to not use up scarce
resources without paying for them.  the most obvious scarce resource here
is the modem, so we charge for connect time. 

hope this helps.

  Edward Vielmetti, vice president for research, Msen Inc. emv@Msen.com
        Msen Inc., 628 Brooks, Ann Arbor MI  48103 +1 313 998 4562

----------------------------------------

Mr. Winsor at WINSOR_L@WIZARD.COLORADO.EDU writes,

I do cost allocation for the central administrative computing service for the 
University of Colorado.  We serve four campuses plus a central administration 
office.  We allocate all costs (equipment depreciation, software, personnel, 
space) to about 19 different services, including communications.  We would like 
to base the communications allocation on connect time and megabytes transferred, 
but don't have the data.  Instead we use a "logon day" basis.  If someone 
logs on once or 100 times in a day, they incur one logon day that gets added 
into the allocation basis for their campus.

------------------------------------------

Mr. Israelit writes,

First, let me explain that I am in business school at the MIT Sloan School
and that, as a result, I am probably going to suggest something that is not
in common usage. I have an interest in Microeconomics and feel that a
practical appplication of this subject could help you out.

To maximize the profits for a project of this nature you should charge a
two-part tariff. Basically, this will be a one-time fee and per connection
charge. This is because there are probably users with different levels of
demand for your services. Using basic micro, you can set the tariff at a
low enough level such that you will attract a large number of subscribers
and set a fee such that you will have a high level of the usage profits.
These levels are selected to maximize the total profits of the project
rather than maximize the gains from either the tariff or the fee alone.

I don't know if this sounds like something you would like to pursue. It
does require that you have an estimate of the market demands from different
users. If you would like to discuss this analysis further (free of charge)
let me know. By the way, I am only at Intel for an internship, and I will
be bac in Boston after the middle of August.

Sincerely,



Sam Israelit
Engineer, Businessman, . . . Brewer
Portland, OR
sami@scic.intel.com

----------------------------------------------------

Mr. Hines writes,

This question is very closely associated with how organizations
budget.

If the organization views the computing resource as "infrastructure"
then it usually is funded centerally - that is to say the providor
pays for everything, do not recharge, and their operating budget is
adjusted upward to absorb the costs.  This is the most simple
arrangement.

If the organization views the computing resource as a "utility" then
it usually expecte the provider of services to establish a fee
structure to recover the operating costs (i.e. break even).  In this
case, the provider has to make up the rules.  It is most "user
friendly" if it is in terms the user can relate to.  Charging for
background resources (disk space, people, etc) are more difficult.
If you can base fees on "every time you hit <enter>" is costs you 25
cents -- that is something they can relate to.  Of course to do this,
the provider has to have a very good understanding of cost accounting
and fixed and variable costs.

Of course there are all kinds of variations to the "utility" scheme.
Part of the costs may be subsidized by central services (a hybrid
"infrastructure" and "utility").

Our data center charges for teleprocessing by establishing classes of
service based on "number of <enter>'s pressed.  They have established
low, medium, and high use classes, and set fixed prices for each.
The class a person is charged for, is determined from prior year
history.  The first year the person is placed in the low category.

We are using a hybrid system - with some costs supported as
infrastructure, and some recharged as a utility.  The allocation
seems somewhat arbitrary, but they maintain it is negotiated.

Again, it depends on your budget structure.


----------------------------------------------------------------------
Internet:  mshines@ia.purdue.edu      |  Michael S. Hines
Bitnet:  michaelh@purccvm             |  Internal Auditor-EDP
Purdue WIZARD Mail: MSHINES           |  Purdue University
GTE Net: (317) 494-5845               |  1065 Freehafer Hall
Compuserve: Mike Hines (73240,1631)   |  West Lafayette, IN 47907-1065
----------------------------------------------------------------------

and Mr. Laidley confers,

 I agree with most of what Michael Hines said, with a one addition.

 We've recently gone through a lengthy and painful experience deciding how to
 charge for usage of the EDI system. In the end we adopted a scheme similar to
 what the VANs use - costs charged are based on usage, measured as the amount
 of data sent/received. Our EDI service is MVS batch oriented, so this was the
 easiest way to measure.

 My key issue is that the cost of collecting data for charging costs not exceed
 the revenue to be generated. It seems silly, but it is easy to lose sight of
 this.

 Regards, Bill Laidley
 EDI Expert Team
 BC Systems Corp. Vancouver (604) 660-9705
 WLAIDLEY@BCSC02.BITNET (Bill Laidley)

-----------------------------------------

Lastly, Mr. Mcguffin writes,

re: your friend's questions.

I helped one of my clients develop a costing / charging model for use
of their multi-platform service bureau and national computer network, so I
have some experience in this area.  Some specific responses, and general
comments:
  
> I need help on the following: I need to know how certain organizations charge
> for use of their network gateways (gateways join two systems/networks
> together, usually at the midrange or mainframe level e.g. at the Front-end
> Processor). For instance, do they charge by the number of sessions a client
> signs up for, by the number of sessions the client actually used, do they have
> a one-time membership fee with a smaller monthly maintenance fee which allows
> users a large but undefined number of sessions, do they not charge at all and
> if so then what is their justification?

This depends on the type of business, and their relationship with their clients
(e.g. is it just internal charging, which is more like cost recovery, or is it
a fee for service business).  I am almost certain there is no "standard" way for
doing this, since it really depends on what the "business objective" is of
charging, plus any business or market constraints.  This will naturally vary
between companies.

>                                         All these questions ask: HOW is the
> gateway charged. But I also need to know, if possible, WHAT the charge is
> based on: what cost items do they include in maintenance of the gateway
> and what percentage of these cost items is used. Cost items include: human
> resources, CPU time, memory, software development, and other overhead items.
>

Again, "it depends".  A charge doesn't actually have to be based on anything,
other than what the market will bear.  You can simply charge $2/whatever,
regardless of what it costs you to provide each whatever.  This is different
from what it _costs_ you to provide each whatever. Cost is important to know 
in order to run your business, but may only slightly influence what you charge.

To figure out what items are included in cost (and maybe, therefore, the charge),
cost accounting principles come into play.  All of the things she mentioned are fair game in determining what it costs to provide the gateway services.  Some have a nice direct relationship to each unit of service, while others (like fixed overheads) can only be assigned arbitrarily.

Assigning costs between services can be a real headache, and depends on what "resource measurement units" there may be kicking around for the resource in question.  For example, you may have a record of CPU usage by user, so you
could apportion that way.  Network packets may or may not be accounted for by
network devices, and may or may not have adequate identification as to who originated them. Humans may keep time sheets, but how do you allocate the "operating system upgrade" project to the gateway versus the other projects
using the machine.  There's lots of room for different opinions here.

>Any ideas from anyone who is familiar with this area would be appreciated.
>Any information on books or papers covering this topic, or firms doing similar
>work, would also be appreciated.
>

I haven't run across any reference material in this area.  Bean counters (I am
one) tend to have great difficulty with the technical issues surrounding cost
accounting for computer resources.  Info systems people (I'm one of these too)
tend to be put off by the whole process of accounting.  So I don't think
that there has been anything definitive put together on either side.

-----------

My advice for your friend would be to start with the business reasons for
"charging" for the gateway.  Is it really cost recovery?  Is it to make
a profit from gateway services?  In either case, what are we trying to
do (e.g. allocate costs to other business units? justify new hardware?
get people to use the gateway?  not use it?  attract new customers?
get customers to use it differentyl, like more off-peak time vs. peak time?)
These and other questions should drive the process a great deal more than
what other companies do.

Next, look at the technical constraints.  What resource usage is actually
measurable, and how can they be attributed to different users?  Will
measurements (e.g. transaction counters) adversely affect performance?
Do we need to be able to prove measurements (as in an audit)?  What will
doing all this cost in equipment, software, people's time.  These may
dramatically limit what can be done.

Then, devise a model for how the charging mechanism should work, and submit
it to many sanity checks, walk throughs, etc.  Once something is decided on,
planning for its implementation will invariably take longer than everyone
thinks.  Things like beginning data collection (resource usage, costs),
developing systems to process the data collected, etc. can take time.

Also, make sure all affected parties are involved..the accountants and the
systems people..in making a decision, and implementing the results.


----------

I have no idea if any of this helps.  I've made all sorts of assumptions,
none of which may be valid.  But perhaps it may give your friend some
structure or direction to work with.

Regards,

Craig McGuffin, C.A.
C.R. McGuffin Consulting Services
Tel: (416) 485-0623
crmcguffin@crmcg.com (Craig R. McGuffin)

==============================================================================

Thanks again to those who responded or expressed interest,

Robert Williams
CIMLab
Mechanical Engineering Dept.
University of Toronto
william1@drill.me.utoronto.ca