The 'learning curve' has entered into everyday usage. We hear that somebody has a 'steep learning curve' when they need to take up new knowledge, particularly when dealing with a new job or activity. At an organisational level, studies in learning rates can orient companies towards the best ways of learning, to boost economies of scale or unit cost reductions. Inversely, such studies can show where an organisation has gone wrong and how initiatives relying on learning to succeed may fail due to a misguided focus on where and how learning takes place.
In this 2004 Sjhingo Research Prize winning paper, Managing Learning Curves in Factories by Creating and Transferring Knowledge, published in the California Management Review, INSEAD's Henry Ford Chaired Professor of Manufacturing, Luk Van Wassenhove and his former Phd student and now Vanderbilt University's Assistant Professor of Operations Management, Michael Lapre, carry out a thorough analysis of the learning process at NV Bekaert SA, the world's largest independent producer of steel wire. Within a research project spanning a number of years, learning processes were studied within Bekaert's steel cord division. This section produces a third of the world's wire used in steel-belted radial tyres through a four-step production process.
The company's production people faced a number of hurdles to learning within the production process, including:
- Detail complexity due to the hundreds of machines, process variables and customer specifications, often needing simultaneous processing;
- Dynamic complexity impacting how the effects of problems experienced on any given machine or production stage could be transferred to the overall process;
- Ambiguity due to disagreements regarding the actual sources of a given problem;
- Incomplete technological knowledge in specifying overall process settings to prevent defects happening.
Bekaert's learning is first analysed at a local level across three domains: quality improvement projects, the impact of learning processes on the factory's learning curve and production lines created specifically to learn about production processes in three factories. Major areas of question concern the impact of locally acquired knowledge on a factory's learning curve, and in particular the impacts of conceptual and operational learning.
Production improvement projects are classified into four categories and conceptual and operational learning scores are compared according to four knowledge stocks:
- Fire-fighting: low conceptual and low operational learning, teams hardly touch on cause-effect relationships and at most, implementing minor changes;
- Artisan skills: low conceptual but high operational learning, projects yielding results but for which the reasons are ill-understood and are therefore difficult to transfer;
- Non-Validated Theories: high conceptual but low operational learning, teams use science and statistics to arrive at a highly designed solution, but cannot ensure operational success when implemented;
- Operationally Validated Theories: high conceptual and high operational learning, teams use scientific models and statistical experiments, resulting in working solutions using scientific principles.
Operationally validated theories bring the most value to the learning curve and can lead local projects to easier global implementation. For this, high levels of both conceptual and operational learning take place involving those who intellectually and technically design solutions and those who implement and operate the improvement solutions. Many companies privilege either one or the other type of learning and as such are destined to fail in reaching any meaningful systemised global improvement.
Lapre and Van Wassenhove then look at Bekaert's efforts to design a production line to accelerate the learning curve, keeping in mind that at the time, within the study area only 25% of improvement projects were operationally validated theories that actually accelerated the learning curve. Bekaert put together a Model Line A (MLA) for one key product and basically changed the organisational structure around this product.
The MLA team worked to a number of aims, both quality and productivity focussed and problems were solved through scientific models and followed though by testable hypotheses. The levels of technical knowledge reached through this process were systematically higher than typical improvement projects. The tried and tested production was then transferred to a new US plant that subsequently reached the highest levels of productivity and quality performance of all Bekaert's plants.
While this was certainly a successful learning curve experiment, Bekaert discovered that wide replication of this MLA business model was not an automatic success and results varied widely across other plants. Once back to 'local' within a 'global' replication things are bound to get more complicated, and this is where local knowledge comes in and the models are readjusted to fit.
Another key success factor is management buy-in and the level of authority given to teams to define their own projects-where the players are implicated, the results speak for themselves. And where a 'thick management' ethic prevails, the project manager would be deep within the process in question, fully aware of all the parameters, but strongly linked to senior management to get support and push changes through. Equally important is the hierarchical environment and the need for projects to be run cross or inter-departmentally to allow complex problem solving.
Managing, and indeed improving learning curves itself requires some learning. But considering the enormous amounts spent on productivity and quality improvement projects, and the not always brilliant results, firms would do well to invest a little more in learning to learn. Lapre and Van Wassenhove's paper will get them off to a good start!
California Management Review, Vol. 46, No 1, Fall 2003