为帮你建立“永动机”，要有一套清晰的规则、一套清晰的衡量指标，来追踪人们按照这些规则行事的情况，也要有基于那些指标结果由公式决算出的预设的对应后果。
规则越清晰，关于某人是否做错了的争执就越少。
例如，我们规定了员工该如何管理自己的个人投资，从而不与我们为客户的投资相冲突。由于这些规则清晰明了， 一旦有人违规，就没什么争执的余地。
设立衡量指标，使所有人都能看到其他任何人的过往记录，这会使评价更客观公正。
人们会做能使他们加分的事，并会减少争辩。当然，由于多数人要做的事具有各种不同的重要性，这就需要采用不同的衡量指标及适当的权重。你搜集的数据越多，反馈就越快、越精确。这也是我创造集点器（提供了大量即时反馈）的一个原因。人们常常利用在会议期间实时获得的反馈，在会议中就及时做出修正。
有了衡量指标，你就可以把它们跟某种算法联系起来，得出相应结果。它们可以简单到比如你每次做了X，就可以赚到Y数量的钱（或奖励分数），也可以更复杂些（如，把指标得分进行加权后，与各种算法相结合，得出报酬或奖励分数的估计值）。
这种过程永远不会精确无误，但它就算是最粗糙的形式也很管用，假以时日，它将变得很棒。即便有瑕疵，公式的计算结果也有助于你做出判断，提供更精确的绩效评价和薪酬建议。时间长了，它将成为一部很有用的机器，会比你自己进行管理要好得多。
To help you build your perpetual motion machine, have a clear set of rules and a clear set of metrics (i.e. KPIs) to track how people are performing against those rules—and predetermined consequences that are determined formulaically based on the output of those metrics.
The more clear-cut the rules are, the less arguing there will be about whether someone did something wrong.
For example, we have rules about how employees can manage their own investments in a way that doesn’t conflict with how we manage money for clients. Because these rules are clear-cut, there’s no room for argument when a breach occurs.
Having metrics that allow everyone to see everyone else’s track record will make evaluation more objective and fair.
People will do the things that will get them higher grades and will argue less about them. Of course, since most people have a number of things to do that are of different importance, different metrics have to be used and weighted appropriately. The more data you collect, the more immediate and precise the feedback will be. That is one of the reasons I created the Dot Collector tool to work as it does (providing lots of immediate feedback); people often use the feedback that they get during a meeting to course-correct in the meeting in real time.
Once you have your metrics, you can tie them to an algorithm that spits out consequences. They can be as simple as saying that for every time you do X you will earn Y amount of money (or bonus points), or it can be more complex (for example, tying the weighted mix of metrics grades to various algorithms that provide the estimated compensation or bonus points).
While this process will never be exact, it will still be good in even its crudest form, and over time it will evolve to be terrific. Even when flawed, the formulaic output can be used with discretion to provide a more precise evaluation and compensation; over time it will evolve into a wonderful machine that will do much of your managing better than you could do it on your own.
