Learning Curves in Implementing Automation

Table of Contents
  • Automation Training - expense or investment?
  • Sources of Training
  • Costs of Training
  • Benefits of Training
  • Priorities for New Breed Developers
  • The Learning Phenomenon
  • Learning Curve Models
  • Impact of Forgetting  

Automation Training - expense or investment?

Any approach to automation training must recognize the team approach already discussed in The Project
Executive and Project Teams.  The training program must recognize that there will be different roles, for example,
managers, technicians, developers, architects, and so on.

Second, training must coordinate with work assignments.  "Just-in-time training" is more than a cliché.  To get the
most return from the training investment team members finishing a course must be able to return to the job site
with immediate usable skills in the tool sets (development languages and other design and programming
products) available in their organization.  Delays in starting the project or acquiring development tools prevent
team members from being as effective as they can be.

Third, training must be both measurable and measured.  This includes independent, objective testing of the
participants and reviews by company personnel to measure effectiveness.

Finally, training must be intensive and accomplished in as short a time as possible.  It is simply too expensive to
let the training period extend over several months.  In a matter of weeks or days, the team members must have
intense exposure to the concepts and products with which they will be working.  This doesn't mean that all the
needed training courses should be taken in immediate succession.  Rather than stay in continuous "training
mode," participants could move directly from each training class to a project role, and back for additional training
as their project progresses.

Sources of Training

Training is normally provided by one (or more) of three sources: internal training departments, software and
hardware vendors, and independent training organizations.  Internal training organizations often are excellent
sources of training in courses that are repetitive in nature, or courses in which the instructor can develop subject
mastery in a relatively brief period.  For example, courses in supervisory skills or Excel spreadsheets are often
available internally.

Hardware and software vendors normally offer training in the specific technologies associated with their products.  
This training is offered through two channels:

  1. Directly from vendor
  2. From training organizations authorized by the vendor

For example, Novell training is provided almost exclusively by training organizations authorized and certified by
Novell.  Microsoft offers training both at its sites throughout the United States and at authorized network training
centers.  Oracle trains almost exclusively in its own centers.  You can get information about vendor sponsored or
authorized training by calling vendors directly and requesting information about their offerings.

If you need training on a specific product, you generally want to be sure that the training vendor is authorized or
certified by the product vendor.  When the product vendor authorizes the trainer, it normally maintains a quality
control interest in the training provided, ensures the trainer has the latest product changes and upgrade
information, and makes certain the training vendor is using the correct version of the software or hardware.

Independent training organizations can offer training in specific product technologies (such as in the use of Visual
Basic), interoperability issues (such as how to make Windows and NetWare work together), and platform-
independent skills (such as relational data analysis).  Independent training organizations may offer curriculum
planning and skills assessment services.  They should be willing to work with your organization to help develop
the objective performance measures you need in order to assess skills and develop a tailored training plan for
your organization.

Costs of Training

Training from product vendors and independent training organizations currently cost between $275 and $400 per
participant per day.  Most offer training at your site for six or more students, which lowers the per day cost for
each student.  Depending on the team role and experience of the team members, the total training requirement
for a person varies from just a few days to more than 40 days.

For purposes of estimating your transition to this environment, you can use $4,000 to $7,000 (1996 dollars!!!) per
person for individuals such as work group managers and LAN engineers, and $9,000 to $14,000 per person for
client/server applications developers and architects.  These costs are incurred either directly (as training dollars)
or indirectly (as project lost time, missed deadlines, network downtime, and so on.)  Usually, costs are much
higher when they are incurred indirectly.  These costs are real and must be factored in to the calculations used to
analyze the economics of automation

Benefits of Training

In a 1990 survey of Fortune 500 companies, the cost of a single hour of network downtime averaged $34,000 -
and this survey was taken before most of the participants had placed any mission-critical applications on the LAN.  
Recently, a Houston energy company undertook a right-sizing project as a cost-saving move.  Every day that it
remained on a mainframe system cost them $9,600 more than operating a LAN.  A delay of only one month in
project completion could cost that company $288,000.

Although quantification of benefits is elusive, the risks of inadequate training are clear: excessive downtime,
project delays,  and applications that fail to work because of platform incompatibility or performance issues.

Priorities for New-Breed Developers

After the team is assembled and trained, the company is at the point at which it is just then ready to begin full
commercial operation / right-sizing.  In the meantime, however, as the team has been assembled, the simple fact
is that the technology will continue to change unabated.

Developers can take steps to keep current and survive in this environment.  
  1. First, they must adopt a modular, reusable, function-specific approach to code implementation.
  2. Second, developers must evaluate new tools on the job and incorporate them when appropriate.  Even
    though a specific tool selection step is included in the downsizing/rightsizing/automation methodology,
    constant review of new tools is necessary to keep abreast of the best options available.
  3. Third, developers should be encouraged to choose tools from reliable vendors that have a likelihood of
    surviving over the longer term.  Simply because a standard was developed more than a few years ago
    does not mean that it's an option for today.  For example, an advantage of COBOL is that it has
    demonstrated remarkable staying power.  Compared to a language such as C++ or SmallTalk, however,
    COBOL is not well designed to handle the development requirements of most GUI-oriented applications
    today.
  4. Fourth, developers and their management should make a commitment to self-development:  subscriptions
    to technical journals, time for skills upgrades, attendance at user group meetings, and technical
    presentations at department meetings.
  5. Last, management and the development staff must realize that training and retooling are ongoing
    processes.

The Learning Phenomenon

In automated mechanized facilities projects, including enterprise automation projects, the effects of learning on
productivity and product quality cannot be understated. The impact of forgetting or interruption of learning in a
production environment is also of major concern.

Learning curves are important for automated productivity analysis, cost estimation, and resource allocation
decisions.  Learning curves present the relationship between cost (or time) and level of activity on the basis of the
effect of learning.  An early study by Wright (1936) disclosed the "80% learning" effect, which indicates that a
given operation is subject to a 20% productivity improvement each time the activity level or production doubles.  
A learning curve can serve as a predictive tool for obtaining time estimates for tasks in a project environment.
Typical learning rates that have been encountered in practice range from 70% - 95%.  Learning curves are also
referred to as progress function, cost-quantity relationship, cost curve, product acceleration curve, improvement
curve, performance curve, experience curve, and efficiency curve.

Several alternate models of learning curves have been presented in literature.  Some of the most notable models
are the log-linear model, the S-curve, the Stanford-B model, DeJohg's learning formula, Levy's adaption function,
Glover's learning formula, Pegels' exponential function, Knecht's upturn model, and Yelle's product model.  The
univariate learning curve expresses a dependent variable (e.g., production cost) in terms of some independent
variable (e.g., cumulative production). The log-linear model is by far the most popular and most used of all the
learning curve models.

Learning Curve Models

There are several models of learning curves in use in business and industry. The log-linear is, perhaps, the most
extensively used.  The log-linear model states that the improvement in productivity is constant (i.e., it has a
constant slope) as output increases.  There are two basic forms of the log-linear model: the average cost function
and the unit cost function.

Average Cost Model

The average cost model is used more that the unit cost model. It specifies the relationship between the
cumulative average cost per unit and cumulative production. The relationship indicates that cumulative cost per
unit will decrease by a constant percentage as the cumulative production volume doubles.

Unit Cost Model

The unit cost model is expressed in terms of the specific cost of producing the xth unit.  The unit cost formula
specifies that the individual cost per unit will decrease by a constant percentage as cumulative production doubles.

Impact of Forgetting

As has been demonstrated in literature, interrupting of the learning process can adversely affect expected
performance.  Forgetting can occur at different times and in different forms as outlined below:

  • Forgetting can occur intermittently based on a scheduled pattern (e.g., scheduled production breaks).
  • Forgetting can occur randomly and governed by some probability distribution (e.g., machine breakdowns).
  • Forgetting can occur continuously due to some natural process (e.g., effect of aging).
  • The psychology literature contains a good coverage of forgetting in the general sense. But extension to
    production research has been limited. A worker can learn and forget at different rates.   

In automated projects, the learning curves take on major significance.  To ignore these curves and to ignore
resistance to change which will dramatically impact these curves is to do so at your own peril and the project's.  
For a good set of articles of how to overcome resistance, you can check out Rick Maurer's website:  http://www.
beyondresistance.com and...

Further information on Learning Curves - refer to Can You Overcome Resistance to Change? This page describes
what you must do even begin to successfully climb the learning curve and it contains a solution for making the
climb easier and much more successful.
AllenWeb Site - since 1995
Learning Curves in Implementing Automation - Jun 1998
Automation
Disclaimer: This is Paul Allen's personal business library. Articles and papers contained therein are for his use and reference.
While this information is being made public by being published on the internet, any use of any material or guidance contained
herein is at one's own risk. Neither the author Paul Allen nor Project Executive Group have any responsibility if anything is taken
from this website and utilized. While we believe these articles and papers make a significant contribution to learning, beware that
anyone who uses this information does so at their own risk. We make no representations to accuracy or completeness.
Return
HOME