In 1985, Herman Miller, Inc. (HMI) introduced Ethospace, a radically new type of systems furniture. Each frame contained multiple tiles; each tile could have one of 75 different finishes...on each side. The flexiblity of the system came at the cost of complexity: in the past, an office could be specified with 10 line items. With Ethospace that blossomed to over 100. HMI knew the only way to specify this new system was using CAD.
HMI approached me to train their elite "Pavilion" dealerships on the use of PC-based CAD and specification systems, thinking the new technology would be their biggest challenge. As it turned out, the Pavilion dealers' problems went much deeper: specification errors. Historically, to mitigate errors in specification, they ordered as much as 20% additional product.
With only 10 line items per office, warehousing a few extra parts may have made sense, but with Ethospace, companies were storing 1000s of parts and pieces to cover contingencies. All at once, the Pavilion dealerships faced three massive challenges:
Training session on use of CAD / Specification system for Ethospace.
Typical floor plan of Ethospace workstations.
The custom-built CAD symbol library - the heart of the improved process.
I founded Phase II thinking I would provide training services to the Pavilion dealers. Within a few months of getting started, I had pivoted from that original notion to creating a computer-based service bureau.
In my meetings with Pavilion leaders, I learned they were less interested in one-time training on their IT infrastructure compared to improving their overall business processes. Once I had completed an end-to-end investigation of how they delivered their services, I was able to envision, design, and deliver a computer-based service that replaced the dealer's 20% contingency with a 5% cost-of-service.
Further, by re-thinking how the CAD and specification software worked together, I was able to guarantee Phase II's error rate at less than 0.01%. On a $1M project, Phase II guaranteed no greater than $100 in errors. Compared to $20K, the costs of Phase II's services were a no-brainer.
Not only did the re-engineered method reduce errors, it simplified the use of the tools. As a result, Phase II was able to hire less experienced designers, improving our bottom line.
System diagram indicating how errors increase as data moves from one actor to another.
Typical furniture layout using standard symbols. Note the reliance on text.
Revised system diagram; errors can only be introduced at the beginning.
I began the effort in two ways: interviewing key stakeholders in the value stream and immersing myself in the existing technologies.
I interviewed Herman Miller staff, Pavilion dealership owners, finance offers, account managers, designers and project managers. I also interviewed furniture installers, who it turned out, became the key to improving the overall process.
Multiple products were required to deliver electronic specifications and CAD drawings: A CAD system (Cadvance or Autocad), a specification system (originally CAP, later purchased by McGraw Hill) and a database system, dBase. I flew to Michigan to be trained by CAP; I became an Autocad and Cadvance VAR and I dug into dBase.
Numerous minor "ahas" emerged throughout these two efforts. For example:
Installations were limited by the amount of packing material the installers could remove from a floor.
The CAD system could count areas of workstations to the installer's specifications.
The breakthrough came in an interview with a well-respected installer: from experience, he had come to distrust the specifications from the designers. During our interview he demonstrated highlighting the CAD drawings with markers in preparation for counting all of the pieces by hand!
He also told me the real crux of the problem: garbage. Literally garbage. His installation crew was limited to using the freight elevator, and only during off-hours. For large installations he was limited by how much packing material, garbage, would fit on the elevator ride down.
Here again, Phase II could solve his problem by giving him product counts based on areas in the floor he specified.
The same layout as above, but rendered with graphic symbols instead of text.
Panel symbol with height infill (62"), width symbol (60") and dashed line indicating power.
Panel detail showing embedded text used for automating panel counts.
As a result of that interview, I envisioned a solution to reducing errors: a custom designed symbol library. The foundation of the system, the symbol library, wasn't up to the tasks required by the different actors. I approached the symbol library as if it were an object-oriented database, crafting each component in the system using graphical markers to distinguish one piece from another:
Once a designer or installer learned the key, they could easily interpret any drawing. By reducing the amount of visible text, the process became far more efficient: the plotters moved lightning fast, and the people could quickly identify errors in the layout by simple inspection. As the process became more mature, Phase II authored CAD macros to audit the drawings to identify potential symbol duplication, further reducing our error rate.
With the exception of establishing Phase II as a consultancy (my business partner, who I later married, helped establish all of the operational elements of the business - payroll, invoicing, project management and the like) I was responsible for all of this effort.
I performed all of the research and analysis; I crafted the symbols, wrote macros and dBase scripts (as needed), hired and managed all employees, and delivered training to clients.