• The industrial revolution provided mass-produced products: everybody got the same (most famously, you could get a Ford T in any color you liked, as long as you liked black).
  • The Internet revolution will drive custom-manufactured products, where each customer gets exactly the unique product they want.

The change to customer-designed products is based on the ability to connect the user interface to the manufacturing backend through a computer. As the product is manufactured, the computer knows what customer it is intended for, what that customer specified, and how to track the product through the manufacturing process so that it can be shipped directly to the desired destination. No inventory (one of the business benefits of custom manufacturing). Examples:

  • Computers are the most classic example of a customer-designed product: almost all PC manufacturers allow users to specify exactly what they want in the box.
  • Compact disks that are burned on-demand with a set of songs chosen by the individual user from a vast database.
    • Buy for yourself to get the tracks you like.
    • Give to others as a personal gift that reflects both your own taste and your insight into the recipient's taste.
    Because you designed the gift, it becomes a much stronger givable item than anything you picked from a store shelf. Also potentially more emotionally significant: create a CD with exactly the right love songs.
  • Insurance plans; your own micro-mutual fund. Non-tangible products are even stronger candidates to be designed by the individual customer than anything that has to be manufactured and shipped.

Configurators

The current standard user interface for user-driven product design is the configurator. Usually in the form of a long list of pull-down menus, each specifying one variable.

These configurators are miserable:

  • clumsy to operate (drop-down menus are an evil UI widget)
  • do not show alternative choices clearly and make it hard to compare alternative designs
  • do not explain the options or teach users to understand the domain of the design problem (contextual help is only a small part of what's needed in most cases)
  • often confusing and overwhelming to novice users; seem to go on and on with more and more obscure choices you need to make

Despite these downsides, configurators have the true benefit of being well-known and following a Web-wide convention. When something is a standard, you should follow it unless you can come up with a significantly better design. Don't just deviate for the sake of being clever: users will be even more confused if you give them a non-standard way of specifying what they want.

Because a configurator is likely to be difficult to use, I strongly recommend subjecting it to intense usability scrutiny and making room in the project plan for several iterations of the design.

It is normal for e-commerce sites to increase sales by 100% or more as a result of usability, but configurator-driven sites can probably increase sales by at least 500% by emphasizing usability. More important, they can probably avoid 9 of 10 returns by eliminating most mis-designed items (a 1,000% improvement of the error rate metric).

(The expected increase in e-commerce sales from usability is based on a good deal of data. The best recorded case so far of increasing sales as a result of following my principles is 2,500%, but that is far from typical: a few hundred percent is more normal. Unfortunately, I don't have credible data for the sales impact of improved configurator usability, so the numbers stated above are my rough projections based on how much one can usually improve the approachability and error rate of complex designs.)

Update added December 2002:
Our studies of Web-based applications and tools in Flash found several good configurators designed in Flash (as well as many usability problems with advanced configurators).

Edit; Don't Design

A blank piece of paper in the typewriter creates writer's block.

Similarly, having to design a product from scratch is tough work, an intimidating challenge, error-prone, and something most users cannot do. There is a reason they are not professional product designers.

It is much easier to make changes to an existing template than to create something from the bottom up.

Limit yourself to showing users a small number of pre-designed templates.

  • span the design space
  • demonstrate the range of things that are possible
  • provide starting points for the users' own design

If you provide too many designs, the initial choice process will be overwhelming and become a usability problem in itself.

In the design interface, emphasize the most common change criteria and relegate secondary choices to an "expert" or "more features" area. If everything is given equal weight, then the novice designer will not know where to begin.

Expert Systems and Design Advisors

In the future, customer-driven design may well be supported by expert systems. Despite much enthusiasm in the mid-1980s, artificial intelligence has acquired a reputation of being unworkable and useless. Mostly, this reputation is well-deserved: it will be a long time before we have perfect natural language understanding or a translation system that produces pleasant text.

Expert systems may be the one form of AI that could be immediately useful on the Web. One of the first examples of a commercial expert system was Digital Equipment's Xcon system for configuring VAX computers. Similar ideas could be used to:

  • advise users on their designs
  • guide users to fruitful next steps
  • caution users against potential problems with their designs

Newer Research

See also our usability research on customization and configuration user interfaces.