贵公司是否准备好进行人力资源分析 Is Your Company Ready for HR Analytics?
尽管许多公司一直在大数据和分析方面进行大量投资，但将分析应用于人力资源的成功案例却很少。但这可能即将改变。
作者：Bart Baesens是比利时鲁汶的KU Leuven教授，也是英国南安普顿南安普顿大学管理学院的讲师
大数据和分析在当今的商业环境中无处不在。更重要的是，诸如物联网，不断扩展的在线社交图以及开放的公共数据的出现等新技术只会增加对深层分析知识和技能的需求。许多公司已经投入大数据和分析，以更好地了解客户行为。事实上，由于引入了各种监管指南，一些最成熟的分析应用程序可以在以客户为中心的保险，风险管理和财务欺诈检测领域找到。
但是，如何利用大数据和分析来深入了解贵公司的另一组关键利益相关者：您的员工？虽然我们看到许多公司加大了对人力资源分析的投入，但我们还没有看到该领域的许多成功案例。由于人力资源分析是业务分析应用程序中的“新手”，我们相信其从业者可以从将分析应用于以客户为中心的领域中获得的经验教训中大大受益 - 从而避免了许多新手错误和昂贵的初学者陷阱。
基于我们的研究和我们在以客户为中心的分析方面的咨询经验，我们提供了四个关于如何成功利用人力资源分析来支持您的战略性劳动力决策的课程。更具体地说，我们将客户分析中的一些最新研究和行业见解与人力资源分析并列，并强调四个重要的溢出效应。
第1课：建模，衡量和管理员工的网络动态。在我们自己的研究中，我们发现客户之间的关系（例如社会关系，与同一商家进行的信用卡交易，或公司之间的董事会成员关系）在解释和预测集体行为（如客户流失，客户响应）方面非常有意义。营销外展或欺诈。我们相信，这些原则可以很容易地用于在人力资源分析中收获一些悬而未决的成果。特别是，可以构建一个网络 - 员工作为节点，并根据诸如（匿名）电子邮件交换，联合项目，主机托管和人才相似性等因素与他们之间的链接进行构建，并且可能对最近这样的连接的加权进行加权。然后可以利用该网络来了解新员工融入您的员工网络的顺利程度;
出于同样的原因，在解雇或解雇员工时，了解员工的社会影响和影响非常重要，以防止病毒影响或人才流失发生在您的网络或公司中。在制定解雇决策时，应仔细联系在组织网络中充当社交影响者或社区连接器的员工，以避免在功能上断开网络的基本部分。
第2课：大数据和分析并不神奇。与任何新技术一样，从一开始就设定适当的期望非常重要。虽然它们可以成为有价值的工具，但分析技术并不是解决公司所有关键任务和困难人力资源决策的灵丹妙药。毕竟，几乎只要分析人力资源模型投入生产，它就会变得过时，因为它的生态系统（包括但不限于公司战略，员工组合和宏观经济环境）经常会发生变化。因此，人力资源最终用户使用他或她的商业智慧，经验以及对问题和组织的了解来批判性地解释，反映，调整和操纵分析模型的结果，这一点至关重要。例如，如果您的分析模型告诉您，您的招聘和解雇政策完全没有 - 或甚至是歧视性的，该怎么办？你使用错误的选择标准或正在寻找不可能的？最近客户流失可以追溯到特定员工的离职？任何意外但有效的分析结果都应该以认真和深思熟虑的方式进行。显然，这需要人力资源经理具有既知情又开放的心态。
第3课：分析人力资源模型应该做的不仅仅是提供统计绩效 - 他们应该提供商业见解。在任何业务环境中部署分析模型时，典型的新手错误是对统计性能（如拟合，相关，R平方等）和过于复杂的分析模型的盲目痴迷。统计绩效很重要，但分析性人力资源模型应该做得更多。另外两个重要的绩效标准是模型可解释性和合规性。
可解释性意味着任何基于分析的人力资源决策都应该得到适当的激励，并且可以简单地向所有涉及的利益相关者解释。这种对简单性的追求阻碍了使用过于复杂的分析模型，这些模型更多地关注统计性能而不是正确的业务洞察力。
另一个关键性能标准涉及模型合规性 保护法规，隐私和道德责任对于成功部署HR分析至关重要。这在人力资源应用中尤为重要。应始终谨慎解释分析模型，在选择构建分析HR模型的数据时，应尊重性别平等和多样性。
第4课：回溯测试分析人员模型的影响。在客户分析中，模型的平均寿命为两到三年，我们没有理由相信这在人力资源分析中会有所不同。然而，考虑到人力资源决策对组织和个人的影响，重要的是通过将预测与现实进行对比来不断地对人力资源中的分析模型进行反向测试，以便可以立即注意到任何性能下降并采取行动。例如，从招聘的角度来看，应该不断评估招聘前的有效性（哪些招聘渠道给我们的候选人提供正确的资料？）和招聘后的有效性（招聘渠道给我们最好的候选人？）。
我们相信现在是时候增加您对人力资源分析的投资了。一旦您的人力资源分析工作成熟，我们就会期待组织的下一个变革步骤。我们认为，当组织将人力资源分析的结果与客户分析的结果汇总在一起时，我们就会发生这种情况。然后，公司可以更全面地了解他们的两个关键人力资产组合之间的关系：员工和客户。
关于作者
Bart Baesens是比利时鲁汶的KU Leuven教授，也是英国南安普顿南安普顿大学管理学院的讲师。他还是“ 大数据世界中的分析：数据科学及其应用基本指南”一书的作者（John Wiley＆Sons，2014）。Sophie De Winne是KU Leuven的副教授。Luc Sels是KU Leuven的经济学和商业学院教授和院长。
Is Your Company Ready for HR Analytics?
Although many companies have been investing heavily in big data and analytics, there have been few success stories in applying analytics to human resources. But that may be about to change.
Big data and analytics are omnipresent in today’s business environment. What’s more, new technologies such as the internet of things, the ever-expanding online social graph, and the emergence of open, public data only increase the need for deep analytical knowledge and skills. Many companies have already invested in big data and analytics to gain a better understanding of customer behavior. In fact, due to the introduction of various regulatory guidelines, some of the most mature analytical applications can be found in customer-focused areas in insurance, risk management, and financial fraud detection.
But what about leveraging big data and analytics to gain insights into another group of your company’s key stakeholders: your employees? Although we see many companies ramping up investments in HR analytics, we haven’t seen many success stories in that area yet. Because HR analytics is “the new kid on the block” in business analytics applications, we believe its practitioners can substantially benefit from lessons learned in applying analytics to customer-focused areas — and thus avoid many rookie mistakes and expensive beginner traps.
Based upon our research and our consulting experience with customer-focused analytics, we offer four lessons about how to successfully leverage HR analytics to support your strategic workforce decisions. More specifically, we will juxtapose some of our recent research and industry insights from customer analytics against HR analytics and highlight four important spillovers.
Lesson 1: Model, measure, and manage your employee network dynamics. In our own research, we have found that ties between customers (such as social ties, credit card transactions made with the same merchants, or board membership ties between companies) are very meaningful in explaining and predicting collective behavior such as customer churn, customer response to marketing outreach, or fraud. It is our belief that these principles can be easily used to harvest some low-hanging fruit in HR analytics. In particular, a network can be constructed — with employees as the nodes and with the links between them based upon factors such as (anonymized) email exchanges, joint projects, colocation, and talent similarity, and possibly weighted for how recent such connections were. This network can then be leveraged to understand how smoothly new hires will blend into your workforce network; it also can be used to quantify the optimal mix, from a performance perspective, between behaviors that bring cohesiveness to the employee network and those that bring diversity.
By the same token, when laying off or firing employees, it is important to understand the social influence and impact of an employee in order to prevent viral effects or talent drain from happening to your network or company. Employees who serve as social influencers or community connectors within your organization’s network should be carefully approached when making firing decisions to avoid functionally disconnecting essential parts of your network.
Lesson 2: Big data and analytics are not magic. As with any new technology, it is important to set appropriate expectations from the outset. While they can be valuable tools, analytics techniques are not a panacea for all of your company’s mission-critical and difficult HR decisions. After all, almost as soon as an analytical HR model is put into production, it becomes outdated, since its ecosystem (including but not limited to company strategy, the employee portfolio, and the macroeconomic environment) is constantly subject to change. Hence it is of key importance that the HR end user critically interprets, reflects, adjusts, and steers the outcomes of the analytical models using his or her business acumen, experience, and knowledge of the problem and organization. For example, what if your analytical model tells you that your hiring and firing policy is not at all sound — or is even discriminatory? That you are using the wrong selection criteria or are searching for the impossible? That the recent loss of customers can be traced back to the departure of a specific employee? Any unexpected yet valid analytical findings should be approached in a careful and thoughtful way. Obviously, this requires HR managers with a mindset that is both informed and open.
Lesson 3: Analytical HR models should do more than provide statistical performance — they should provide business insights. A typical rookie mistake when deploying analytical models in any business context is a blind obsession with statistical performance (such as fit, correlation, R-squared, etc.) and overly complex analytical models. Statistical performance is important, but analytical HR models should do more. Two other important performance criteria are model interpretability and compliance.
Interpretability means that any HR decision based upon analytics should be properly motivated and can be simply explained to all stakeholders involved. This quest for simplicity discourages the use of overly complex analytical models that focus more on statistical performance than on proper business insight.
Another key performance criterion concerns model compliance. Safeguarding regulations, privacy, and ethical responsibilities is crucial to successfully deploying HR analytics. This is especially important in HR applications. Analytical models should always be interpreted with caution, and gender equality and diversity should be respected when selecting the data to build your analytical HR models.
Lesson 4: Backtest the impact of your analytical workforce models. In customer analytics, the average lifespan of a model is two to three years, and we have no reason to believe that this will be different in HR analytics. However, given the impact of HR decisions on the organization and on individuals, it is important that analytical models in HR are constantly backtested by contrasting the predictions against reality, so that any degradation in performance can be immediately noticed and acted upon. For example, from a hiring perspective, both the pre-hire effectiveness (which recruitment channels give us the candidates with the right profile?) and post-hire effectiveness (which recruitment channels gave us the best candidates?) should be constantly evaluated.
We believe the time is right to boost your investments in HR analytics. And once your HR analytics efforts have matured, we look forward to the next transformative step for organizations. That, we think, will take place when organizations can bring together findings from HR analytics with those from customer analytics. Then companies can more fully understand the relationships between their two key sets of human assets: employees and customers.
ABOUT THE AUTHORS
Bart Baesens is a professor at KU Leuven in Leuven, Belgium, and a lecturer at the University of Southampton School of Management in Southampton, U.K.; he is also the author of the book Analytics in a Big Data World: The Essential Guide to Data Science and its Applications (John Wiley & Sons, 2014). Sophie De Winne is an associate professor at KU Leuven. Luc Sels is a professor and dean of the faculty of economics and business at KU Leuven.
Bart Baesens
扫一扫 加微信
hrtechchina
