We'll assume here that bottlenecks are "bad" -- the flow restriction you don't want. The place on the production line (for instance) where product is not getting through, causing materials to pile up upstream and delaying sales to consumers and profits for the company. Generally speaking, the bottleneck is the first thing you have to look for if you're interested in improving the quality of a product or service.

Here's how to address the bottleneck, greatly simplified:

1. Find it. Usually, you find it by locating an accumulation of input materials, or by observing downtime either immediately upstream (all done) or immediately downstream (nothing to do yet).

   BUT, keep in mind that bottlenecks don't occur on production lines alone -- you can find them along any causal chain. That includes offices and social systems as well.

2. Fully exploit it. This means that if all the machines aren't on, you turn them on; if not everyone who could work at the bottleneck is, you move them there; etc.

   At this point you may see (if you haven't seen already) that there's a natural problem. One of the machines is down, or is due/overdue for maintenance and is not running efficiently. One of the workers is either not sufficiently trained, or is having health problems, or is short on motivation due to factors outside of work. One of the organizations you deal with has a different set of priorities and hasn't gotten to your work yet.

3. Increase the capacity of the bottleneck.

   Fix the machine. Or upgrade it. Train the worker. Or offer counseling. Or offer some family leave. Or move the worker elsewhere in the organization. Or (if all else fails) let the worker go. Rearrange the relationship with the other organization so the priorities match. If it's an organization in the same company, bring a common manager in to work it out. If it's outside the company, look hard at working with someone else.

Now, keep in mind that the only guarantee you have in terms of cause and effect is that if the cause is the three-step process above, the effect will be that you've increased the capacity of the bottleneck. For now. You haven't improved (or at least, you haven't shown you've improved) quality. That takes some more observations.

And there are problems involved:

• All that effort might only serve to reveal the next bottleneck. Then you start over.
• If you don't correct the bottleneck properly, you may have to do it again. "There's never time to do it right; there's always time to do it over." This can be especially embarrassing if the correction involved spending a lot of money, or firing people.
• You can't usually identify more than one at a time. Not unless the other bottleneck(s) are sufficiently separated on the causal chain that they don't feed each other directly.
• Everybody who is interested in the bottleneck must help to resolve it. Even if they actually work upstream or downstream or in management. And like anything else that distracts us from what we see as "our job," this means efforts to resolve the bottleneck will probably meet with opposition at least at first.
• Resolving the bottleneck, and in some cases even recognizing it, depends on your system knowledge. What quantities comprise the critical flow of your system? How are they to be measured? You may be unable to start until you know.

Rizzo offers an example of product development flow, based on information about customer needs as the critical quantity to be monitored:

In this case, information flow is represented by arrows: yellow for information based on direct contact between company and customers; white for information contributed from resellers. The flow goes (counterclockwise from top right) from marketing to design (with a flow limiter) to development to distribution. The flow limiter is thought to prevent a bottleneck that occurs when the marketing function is permitted to sell products without limit, even when the products can't be ready in time to satisfy customers.

The above problems in resolving bottlenecks are generally something you can overcome. Provided it's you in charge. You'd have to be the "puppet master" to make these all go away yourself. If you're not, all you can do is recommend. You can't assume things will be done, no matter how simple or inexpensive or logical they seem to you to be.

Fixing bottlenecks is still a worthy goal, and also (conveniently) an excellent source of rhetorical arguments! :-)

References

If you want to find out more about types of bottlenecks and how to overcome them via the Theory of Constraints, check out The Goldratt Institute. E. M. Goldratt has written novels that serve as "parables" of bottleneck repair:

The Goal. ISBN 0-884-27061-0
It's Not Luck. ISBN 0-884-27115-3
The Haystack Syndrome. ISBN 0-884-27089-0
Theory of Constraints. ISBN 0-884-27085-8

The PD Institute specializes in applications of the Theory of Constraints