错觉设计_做出以数据为依据的决策的错觉

错觉设计

We’ve all been there. Decision-makers sit around a large, well-made table with the typical projection and collaboration technologies we’ve come to expect. With their dashboards, charts, and interactive reports blazoned on the conference room walls, a lively discussion ensues between these heads of state. Once they’ve arrived at a collective warm-and-fuzzy moment regarding their conclusions, they brainstorm a to-do list of action items to be dispersed throughout the company — finally, a company making data-driven decisions.

我们都去过那里。 决策者坐在一张精心制作的大桌子旁，摆放着我们期望的典型投影和协作技术。 在会议室墙壁上竖立着仪表板，图表和交互式报告，这些国家元首之间进行了热烈的讨论。 一旦得出结论后，他们就集体进行了热烈的讨论，然后集体讨论将要散布在整个公司中的行动项目的待办事项清单，最后是一家由数据驱动的决策公司。

Wait! Everyone has just been duped. You’ve merely witnessed the illusion of making data-driven decisions.

等待！ 每个人都被骗了。 您只是亲眼目睹了制定数据驱动型决策的幻想。

场景 (The Scenario)

Everyone talks about using data to make decisions. Many companies claim to be making data-driven decisions. But their results are mixed. So they engage consultants like me to advise them, hoping to improve their insights. Unfortunately, these business leaders mistakenly attribute their poor performance to failure in the data science department. Ironically, any poor decision resulting from their use of the data is typically their own.

每个人都在谈论使用数据进行决策。 许多公司声称正在做出由数据驱动的决策。 但是他们的结果好坏参半。 因此，他们邀请像我这样的顾问来为他们提供建议，以期改善他们的见解。 不幸的是，这些业务领导者错误地将其糟糕的业绩归因于数据科学部门的失败。 具有讽刺意味的是，由于使用数据而导致的任何错误决定通常都是他们自己的。

These companies have copious amounts of data. They’ve created a small team of individuals capable of the necessary data science skills to curate the data into meaningful information. Data wranglers, data scientists, programmers, data engineers, graphic artists, UI/UX professionals, and subject matter experts are all represented in the effort. So what’s the problem?

这些公司拥有大量数据。 他们创建了一个由小团队组成的团队，他们具备必要的数据科学技能，可以将数据整理为有意义的信息。 数据争吵者，数据科学家，程序员，数据工程师，图形艺术家，UI / UX专业人员和主题专家都参与了这项工作。 所以有什么问题？

问题 (The Problem)

Data acts like a muse inspiring the business leaders rather than driving their decisions. Their method of using data assumes they can make unbiased decisions after being influenced by the massive amounts of information shared in their meetings. Genuine data-driven decisions come through an entirely different approach. To ensure data drives your business, you must establish the decision criteria before introducing data insights.

数据就像是缪斯女神在鼓舞业务领导者，而不是驱动他们的决策。 他们使用数据的方法假定，在受到会议共享的大量信息影响后，他们可以做出公正的决定。 真正的数据驱动决策是通过完全不同的方法实现的。 为确保数据驱动您的业务，您必须在引入数据见解之前建立决策标准。

Decision science is a discipline unto itself. If you’re serious about using decision science techniques to their fullest, you’ll need to invest much time and money. Since companies with small and large budgets need to make better decisions, let’s outline just a few fundamental decision science practices that all companies can afford and implement.

决策科学本身就是一门学科。 如果您想充分利用决策科学技术，则需要投入大量时间和金钱。 由于预算大小的公司都需要做出更好的决策，因此，我们仅概述一些所有公司都能负担和实施的基本决策科学实践。

实现真正的数据驱动决策的5个步骤 (5 Steps to Achieve Genuine Data-Driven Decisions)

＃1 —定义业务问题 (#1 — Define the Business Question)

Begin by establishing the question or questions you are trying to answer. If your company is shooting from the hip every time your leaders gather, its destination is undefined. Your data will either lead you off a cliff or resemble the path of an intoxicated sloth. Neither result is desirable.

首先确定一个或多个您要回答的问题。 如果您的公司每次领导人聚集时都从头到尾射击，则其目的地是不确定的。 您的数据将使您脱颖而出，或者像醉了的树懒一样走。 两种结果都不可取。

Avoiding this problem is intellectually simplistic, even though taking the proper actions to fix it can be challenging. Before assessing any data, begin by identifying the business question.

避免此问题在理论上很简单，即使采取适当的措施来解决它也很困难。 在评估任何数据之前，首先要确定业务问题。

Business Question: Should we expand our sales efforts into the western region?

业务问题：我们是否应该将销售工作扩展到西部地区？

＃2-在查看所有数据之前定义您的决策标准 (#2 — Define Your Decision Criteria Before You See All the Data)

Next, define the criteria for making this decision. There’s a great scene from The Insider movie. Russel Crowe plays a whistleblowing insider from a tobacco company. Just as he is heading to the deposition, the proverbial point-of-no-return, he pauses. His attorney questions whether he’s ready or not. Crowe’s character replies, “I can’t find the criteria to decide.”

接下来，定义做出此决定的标准。 The Insider电影中有一个很棒的场景。 拉塞尔·克劳(Russel Crowe)是一家烟草公司的内部举报人。 就在他即将要撤离的消息之际，他停了下来。 他的律师质疑他是否准备好了。 克劳的角色回答：“我找不到决定的标准。”

The same problem faced the boardroom members in our example. A lack of predetermined standards for evaluation lays the foundation for cognitive bias. The insights from the dashboards and reports subconsciously influence their use. Subsequently, the decision is no longer objective. It’s not data-driven.

在我们的示例中，董事会成员面临相同的问题。 缺乏预定的评估标准为认知偏见奠定了基础。 仪表板和报告中的洞察力会潜意识地影响其使用。 随后，该决定不再是客观的。 它不是数据驱动的。

If you are in sales, you probably already know the requirements for deciding this.

如果您正在从事销售，那么您可能已经知道决定这一要求的条件。

Some example criteria might be:

一些示例标准可能是：

• Will sales in the new region impact sales in existing regions in any significant way?

  新地区的销售是否会对现有地区的销售产生重大影响？

• Does the market for our product in the new region meet the minimum potential sales required for the new market to be independently viable within three years?

  我们的产品在新地区的市场是否满足新市场在三年内独立生存所需的最低潜在销售额？

• Do we have or can we obtain the necessary sales resources for the new market?

  我们是否拥有或可以获取新市场所需的销售资源？

• Do we have or can we obtain the necessary marketing resource to support the new market?

  我们是否拥有或可以获得必要的营销资源来支持新市场？

• Do we have or can we obtain the necessary supply chain resources to support the new market?

  我们是否拥有或能够获得必要的供应链资源来支持新市场？

• Do we have or can we obtain the necessary supporting business services to support the new market?

  我们是否拥有或可以获得必要的支持业务服务来支持新市场？

• Do we have realistic targets to determine success with this new market in one, two, and three years?

  我们是否有现实的目标来确定这一新市场在一年，两年和三年中的成功？

• Do we have a realistic exit strategy for this market?

  我们是否有针对该市场的切合实际的退出策略？

• How much capital will be required to begin expanding?

  开始扩张需要多少资金？

• Do we have sufficient capital to fund the expansion until the new marketing is self-sustaining?

  在新的营销活动能够自我维持之前，我们是否有足够的资金来为扩张提供资金？

The easy part about this step is you don’t need a data scientist to do it. Now that you have the questions, see if you can answer them. You might need a data analyst to assist you, or you might have all of these answers already available to you. If you already have the answers, jump to step four. Otherwise, you’ll need to stop at step three along the way.

关于此步骤的简单部分是您不需要数据科学家来执行此操作。 现在您有问题了，看看是否可以回答。 您可能需要一位数据分析师来协助您，或者您可能已经拥有所有这些答案。 如果您已经有了答案，请跳至第四步。 否则，您将需要在过程的第三步停止。

＃3 —将数据一分为二 (#3 — Divide Your Data in Two)

If you use your entire dataset for exploratory purposes, you’ll create the same cognitive bias we saw in the board room. Instead, divide your data into two sets using roughly a 20/80 split. The smaller of the two datasets is for exploratory data analysis. Any department in your organization is welcome to use it. Share it with everyone asking you for data. Reserve the larger dataset for reporting. This last dataset drives your business decisions.

如果您将整个数据集用于探索目的，则将创建与董事会会议室相同的认知偏见。 而是使用大约20/80的划分将数据分为两组。 两个数据集中的较小者用于探索性数据分析。 欢迎您组织中的任何部门使用它。 与要求您提供数据的所有人共享。 保留较大的数据集以进行报告。 最后一个数据集可驱动您的业务决策。

After you’ve split your dataset, you may ask a data analyst to help you establish the decision metrics corresponding to your pre-defined criteria.

分割数据集后，您可以要求数据分析师帮助您建立与预定义标准相对应的决策指标。

＃4 —通过回答标准问题来建立指标 (#4 — Establish Your Metrics by Answering Your Criteria Questions)

The answers you supply to these questions will become your threshold for making your final decision. It’s essential to answer these before you’re sitting in the conference room looking at the latest reports. Whatever metrics you establish will protect you from using data as a muse and propel it into the driver seat. Take your time on this step, and be sure to use reasonable and valid metrics.

您对这些问题提供的答案将成为您做出最终决定的门槛。 坐在会议室查看最新报告之前，必须回答这些问题。 无论您建立什么度量标准，都可以防止您将数据用作缪斯数据并将其推入驾驶员座椅。 请花一些时间在此步骤上，并确保使用合理有效的指标。

Although we dream of perfect data, it doesn’t exist. You will find spots within your data that simply cannot offer an exact measurement. Don’t panic. Don’t throw in the towel.

尽管我们梦perfect以求的是完美的数据，但是它不存在。 您会在数据中发现根本无法提供精确测量值的斑点。 不要惊慌 不要扔毛巾。

When you encounter insufficient data, use your human intuition and your buddies in data science to help you establish an acceptable threshold. You now have your decision metric. Imperfect data is nothing to fear; it is still valuable and can even be part of the decision making process.

当您遇到数据不足时，请运用您的直觉和数据科学领域的伙伴来帮助您建立可接受的阈值。 现在，您有了决策指标。 不完美的数据没什么可担心的。 它仍然很有价值，甚至可以作为决策过程的一部分。

＃5 —做出以数据为依据的决策 (#5 — Make Data-Driven Decisions)

If you’ve used the first four steps, the fifth step is easy. Let your data science team create dashboards from your reporting data. Measure the outcome against your decision criteria and metrics. Act accordingly. When metrics meet a specific action’s threshold, take action. If not, wait. Do not act. Isn’t making decisions simple?

如果您使用了前四个步骤，则第五个步骤很容易。 让您的数据科学团队根据您的报告数据创建仪表板。 根据您的决策标准和指标衡量结果。 按指示行动。 当指标达到特定操作的阈值时，请采取措施。 如果没有，请稍候。 不要行动。 决策不简单吗？

结论 (Conclusion)

As your company grows, your decision-making process should become more involved by fully leveraging decision science practices. But these five steps are a strong start. They will dispel the illusion you’ve been practicing and help you truly make decisions driven by data.

随着公司的发展，通过充分利用决策科学实践，您的决策过程将更加参与。 但是这五个步骤是一个好的开始。 它们将消除您一直在练习的幻想，并帮助您真正做出由数据驱动的决策。

Rod Castor helps companies Get Analytics Right! He helps international organizations and small businesses improve their efforts in data analytics, data science, tech strategy, and tech leadership. In addition to consulting, Rod also enjoys public speaking, teaching, and writing. You can discover more about Rod and his work at rodcastor.com and through his mailing list.

Rod Castor 帮助公司正确完成分析！ 他帮助国际组织和小型企业改善在数据分析，数据科学，技术战略和技术领导力方面的工作。 除了提供咨询服务外，Rod还喜欢公开演讲，教学和写作。 您可以在 rodcastor.com 和他的 邮件列表中 找到有关Rod及其工作的更多信息 。

翻译自: https://towardsdatascience.com/the-illusion-of-making-data-driven-decisions-bf54a2e594c4

错觉设计