善用数据，为企业决策服务
|
几个月前我收到了一份便函，要求我所在的埃森哲(Accenture)办公室的员工们必须保持室内整洁，接受定期检查。碰巧我是一个喜欢整洁的人，但我希望知道是否有数据证明整洁的办公室就能促进生产效率的提高。 毫不奇怪，我的提问没有得到很好的回答，答复是“基肖尔，整洁的办公室会给到访的客户留下更好的印象。” 这听起来有几分道理，因此我继续问，是否有数据支持这种观点，即在拜访过我们整洁的办公室后，客户更可能购买我们的服务或对我们有更正面的看法。至此，我似乎是在本应显而易见的事情上浪费人们的时间了，有几位同事甚至建议我别再纠缠这个问题了。 在当今高度竞争的全球商务环境中，你应该如何运用数据来支持你的大小决定，正是企业应该探讨的话题。而且随着商业分析理论的完善，没有理由不基于充分的信息作决定，而且很多时候支持性数据完全可以随手拈来。 如今，你的企业可以轻松获得关于客户购买模式、自身供应链内商品动向等多年的数据。而且，你的雇员、你的客户、你的竞争对手以及你竞争对手的雇员和客户也都在谈论，包括在博客和微博上提供对你的企业可能有用的信息。当今的一些技术——如数据/文字挖掘和机器学习——能帮助你对所有这些数据进行分析，而云计算也将信息研究规模提升到了可能几年前还不可想象的水平。 大多数企业领导人现在都要求重要决定必须要有经验数据的支持。随着分析理论的进步，现在我们已差不多到了这样的地步：即便是最日常的决定，各个层面的管理人士都必须问这样一个问题，“我们认为是这样，还是我们知道是这样？” 越来越多的公司都在朝着这个方向转变。他们必须要了解运用数据为其决策和行动提供指导的潜在机会和挑战： 1. 谨防数据误用 分析理论是个强大的工具，借用蜘蛛侠的话，“能力越大，责任越大”(with great power comes great responsibility)。企业应谨防三类常见的数据误用。 首先，拥有实时数据并不意味着你能够或应该做出实时决定。不同种类的数据有不同的时间尺度：例如，收银机反映的是当时的销售额，但供应链数据只能反映上次下单或上次订单的运输派车。必须所有数据在手，才能做出好的决定，因此你的决策速度只能取决于最慢的因素。 第二，分析理论能帮助你优化企业流程，将冗余和低效降至最低。但企业流程不能过度优化，否则可能导致犯错余地为零。高度优化的流程——如零库存或保持极低的库存，根据需求随时补充——是非常脆弱的，因为可能出现你无法控制的局面，而你的犯错余地为零。 最后，不要做无谓的决定。有好的数据，并不意味着你总要据此做点什么决策。 2. 做好准备，随时应对瞬息万变的信息世界 一个基于数据采取行动的公司能做出非常具体、精确的决定。事实上，你的决定可能基于一些细微之处，如“周日晚上在那些近期表现不错的主场足球队所在地区多备些啤酒”。但这样的决定随时可能调整，随球队的命运而快速变化。 3. 解读海量数据 当今企业拥有的信息已超过了他们所能利用或能采取行动的范围，因为很多不同的信息往往都是孤立的。未来的企业将需要花大量的时间和精力来整合它们拥有的有用信息。 以医药公司为例，传统上依赖临床试验数据确立新药的功效和副作用。如果临床试验没有问题，他们就能宣称对药物的不良反应不承担法律或道德责任。但随着互联网和社交媒体的出现，如今他们必须监控公共信息源，将这些信息与临床数据结合。当一家公司出现问题时，我们将更多地听到公司回应以“我本该知道”，而不是“我不知道”或“我不可能早就知道”。 4. 不要迷失于信息汪洋 如此多的数据可能很容易就会让未来的企业经理们误入“拖延决策，直到完成所有数据分析”的陷阱，但完成所有数据分析可能是无法完成的任务。你应该警惕陷入分析迷局的三个警示信号。 首先，警惕管理层的“过拟合”倾向——统计学词汇“过拟合”指的是一旦模式已经发现，搜集更多数据的价值趋于下降。数据搜集是有代价的。不行动也是有代价的。一个具有数据头脑的公司必须知道过拟合成本。 第二，不要苦等不存在的数据。具有数据头脑的公司知道信息差的存在，知道如何通过实验打破此类僵局。 最后，要知道你的企业在行动时愿意承受何种水平风险。如果员工因为行动失败所受处罚多于不行动，大多数员工都会宁愿不行动，也不愿将事情搞得一团糟。针对行动失败和根本不行动建立健全的惩罚机制，能提供帮助。 5. 发挥直觉 依赖数据并不意味着不需要直觉。是的，科学确实是以经验为根据，是理性的。但科学家们不是。大多数受人尊敬的科学家们都是在保持客观性的同时，发挥创造力、直觉和冒险精神。这为企业提供了一个良好的参照。 未来基于分析决策的企业将明显不同于今日的企业。回到文章开始我描绘的那些干净的办公桌、效率、客户以及是否有数据支持这样一个日常性决定。就此案而言，无数据提供。但为防万一，我还是将自己的办公桌弄得比以前更整洁了一些。 本文作者基肖尔•斯瓦米纳坦(Kishore S. Swaminathan)是埃森哲的首席科学家，以及埃森哲技术实验室(Accenture Technology Labs)的系统集成研究全球总监。 |
A few months ago, I received a memo saying that employees in my facility at Accenture must keep their offices clean, subject to regular inspections. As it happens, I am fairly tidy, but I wanted to understand if there was any data to show that clean offices lead to higher productivity. Not surprisingly, my request was sidestepped, and I was told, "Kishore, clean offices leave better impressions with visiting customers." That sounded reasonable, so I asked if there was any data to show that our customers are more likely to buy our services or view us more favorably after visiting our clean offices. Now I was wasting people's time on what should be obvious, and a few colleagues even suggested that I move on. In today's highly competitive global business environment, how you should use data to support your decisions -- large and small -- is exactly the kind of conversation that organizations should be having. And with advances in business analytics, there is every reason to make well-informed decisions since supporting data is, in many cases, readily available at your fingertips. Your company now can easily gain access to several years of data about your customer's buying patterns and the movement of goods through your supply chain. And your employees, your customers, your competitors, as well as the employees and customers of your competitors are all talking, blogging and tweeting, providing potentially useful information for your business. Today's technologies -- such as data and text mining and machine learning -- allow you to analyze all this data, and cloud computing allows you to examine this information at a scale that was not possible just a few years ago. Most business leaders now demand empirical data to support important decisions. With advances in analytics, we are nearing the point where every executive at every level will have to subject even the most mundane business decision to the following question: "Do we think this is true, or do we know this is true?" As more organizations move in this direction, though, they ought to be aware of the potential opportunities and challenges that go along with using data to guide more of their decisions and actions: 1. Avoiding the misuse of data Analytics places tremendous power in the hands of its users, and to borrow from Spiderman, "with great power comes great responsibility." Organizations should watch for three common misuses of data. First, just because you have access to real-time data doesn't mean you can or should make real-time decisions. Different types of data have different time scales: for example, your cash register reflects your sales the moment they happen, but your supply chain data can only reflect the last time an order was placed or a truck carrying your order was dispatched. Best decisions are made with all the data at hand, so you can only make decisions as fast as your slowest moving event. Second, analytics enables you to optimize your business processes to minimize redundancies and inefficiencies. However, be careful not to overly optimize your business processes to the point that there is no room for error. Highly optimized processes -- just-in-time inventory or keeping a very small inventory and constantly replenishing it based on demand being an example -- are very fragile because circumstances beyond your control could arise, and there is little room for error. Finally, watch out for making decisions where none are needed. Having good data does not mean you always need to act on it. 2. Preparing for a rapidly changing information world A company that bases its actions on data can make very specific, fine-tuned decisions. In fact, your decisions can be based on subtleties such as "stock more beer on Sunday nights in locations where the home football team is on a winning streak." But these kinds of decisions are highly sensitive and can change as rapidly as the fortunes of a football team. 3. Making sense of a ton of data Today's enterprises have more information than they can use or act on because many difference pieces of information are often isolated from each other. The enterprise of the future will need to devote a lot of time and energy toward integrating the useful information it has. Pharmaceutical companies, for example, have traditionally relied on clinical trials data to establish the efficacy and side effects of drugs. If a problem didn't come up in clinical trials, they could claim legal or ethical immunity from adverse effects of their drugs. But with the advent of the Internet and social media, they must now monitor public sources and integrate that information with their clinical data. "I should have known" will be the new normal, replacing the "I did not know" or "I could not have known" response to a company's unexpected problems. 4. Avoiding paralysis by information overload With access to so much data, the business manager of the future could easily fall into a trap of putting off decisions until everything has been analyzed, which may never happen. Look out for three warning signs of analysis-paralysis. First, beware the managerial tendency to "over-fit the curve" -- a statistical term that refers to the diminishing value of gathering additional data once you find a pattern. Data collection has a price. Not taking action also has comes at a price. And a data savvy organization must understand the cost of over-fitting. Second, do not fall into the trap of waiting for data that just does not exist. Data savvy organizations understand information gaps and how experimentation can break these kinds of logjams. Finally, know what level of risk your organization is willing to tolerate when they take action. If you penalize employees more for failed action than for inaction, most employees will prefer to not take action rather than mess up. Having solid guidelines for how to treat failure versus not acting at all can help. 5. Intuition isn't dead Relying on data does not mean that there is no room for intuition. Yes, it is true that science is empirical and dispassionate. But scientists are not. Most respected scientists blend objectivity with creativity, instinct and risk taking. It's a good model for organizations. The enterprise of the future, based on analytical decision making, will be considerably different from today's enterprise. All of this goes back to that original scenario I painted about clean desks, efficiency, clients and whether there was any data to support a rather mundane policy decision. In this case, none was provided. But I keep my desk a littler cleaner just in case. Kishore S. Swaminathan is Accenture's chief scientist and the global director of Accenture Technology Labs' systems integration research. |
