The secrets to a good employee survey Likert or not Mar 19th 2026|4 min read 1. “This sentence is false” is an example of a logical inconsistency known as the liar paradox. If this sentence is true, then it is indeed false. But if this sentence is false, then it must be true. This is the kind of thing that makes philosophers go weak at the knees and gives normal people a headache. 2. A small echo of the liar paradox can be heard in a ritual of modern management: the annual employee survey. Imagine being asked to react to this statement: “This survey is a complete waste of time.” If enough people Strongly Agree with this proposition, then it’s probably true. But if a company is the kind of place where employees are prepared to give such honest feedback, then isn’t it likely to be false? 3. Employee surveys are a staple of corporate life. Knowing what workers are thinking is an important goal. High employee churn imposes financial and operational costs. There is lots of research to suggest that employee satisfaction leads to better financial outcomes. But set-piece surveys are really useful only if three conditions are met: they are properly designed, they are used in conjunction with other tools and they lead somewhere. 4. Among other things, proper design means grappling with the problem that employees are not necessarily incentivised to be honest. Faced with a Likert scale and the proposition that “My bosses have the communication skills of a banana,” you might Strongly Agree but still opt to Neither Agree Nor Disagree on your submitted form. Promises of confidentiality and anonymity can help, but only to a point. 5. Impression management, a fancy name for making yourself look good, can skew results on questions about things like job-safety practices. There are ways to mitigate this, however. A recent study by Emma Zaal of the University of Groningen and her co-authors found that using different survey formulations can have a big impact on responses. In a survey of Dutch adults, which asked questions like whether they had sent text messages while driving a car, the inclusion of face-saving options such as “occasionally” or “only when no other option” elicited a very different set of answers from binary “yes” or “no” options. 6. Employers can look at unvarnished feedback, most obviously on workplace-review sites such as GlassDoor. Artificial intelligence has the ability to build a coherent picture out of a mass of unstructured comments. In one recent paper, Tom Reader and Alex Gillespie of the London School of Economics looked for evidence of high-pressure cultures in employee reviews of European firms. Reviews that suggested very ambitious targets and expediency in reaching them were predictive of companies experiencing a future corporate scandal. 7. Frequency is another aspect of good design. A lot can change in the space of a year; an annual survey is a long time to wait for an update on employee sentiment. Retrospective evaluations are also subject to biases like the peak-end rule, which describes how people overweight the most extreme and the closing moments of an experience when they recall it. In one famous experiment, Daniel Kahneman and others put volunteers through two unpleasant tasks: the first involved holding their hands in icy water for a minute, and the second for 90 seconds, though for the final 30 seconds the water’s temperature rose by a little. The first experience was objectively less painful but, given a choice, the second was the one people chose to repeat. 8. Shorter “pulse” surveys cannot eradicate these problems, but are a way to gather more timely data. HappyOrNot, a Finnish company that makes those smiley-face feedback terminals you see in airports and elsewhere, also installs its machines inside companies as a way of keeping track of employee sentiment on a daily or weekly basis. 9. Good design and multiple sources of information contribute to a successful employee survey. But nothing matters more than being seen to act on feedback. If you say that your bosses have the communication skills of a banana and then hear nothing back, you have the faint satisfaction of knowing you are right but not much else. Surveys that prompt no follow-up action deepen cynicism rather than enthusiasm. 10. All of which leads to another paradox. Surveys are most useful in organisations that care about what their employees think. But organisations that care about what their employees think often have less need for surveys.■ 員工滿意度調查成功的秘訣 不論你喜不喜歡（Likert scale 刻度調查的雙關語） 2026年3月19日 | 4 分鐘閱讀 「這句話是假的」是一個被稱為「說謊者悖論」的邏輯矛盾案例。如果這句話是真的，那麼它確實是假的；但如果這句話是假的，那麼它就必須是真的。這種事情會讓哲學家雙膝發軟，並讓一般人感到頭痛。 在現代管理的一項儀式中，可以聽到說謊者悖論的微弱迴聲：那就是年度員工調查。想像一下，當你被要求對這句話做出反應：「這次調查完全是浪費時間。」如果有足夠多的人「強烈同意」這個命題，那麼它很可能是真的。但如果一家公司是那種員工願意提供如此誠實反饋的地方，那麼這句話難道不更有可能是假的嗎？ 員工調查是企業生活的家常便飯。了解員工在想什麼是一個重要的目標。高員工離職率會帶來財務和營運上的成本。有大量的研究表明，員工滿意度能帶來更好的財務成果。但是，這種固定形式的調查只有在滿足三個條件時才真正有用：它們經過妥善設計、與其他工具配合使用，並且能帶來實際的改變。 其中，妥善的設計意味著必須解決一個問題：員工不一定有動機保持誠實。面對李克特量表（Likert scale）以及「我的老闆具備香蕉般的溝通技巧」這個命題時，你內心可能「強烈同意」，但最終在提交的表單上還是會選擇「既不同意也不反對」。保護隱私和匿名的承諾會有所幫助，但也只能起到一定程度的作用。 印象管理（這是一個讓自己看起來面子好看的精緻名稱）可能會在有關工作安全規範等問題上扭曲調查結果。然而，有一些方法可以減輕這種情況。格羅寧根大學的艾瑪·扎爾（Emma Zaal）及其合著者最近發表的一項研究發現，使用不同的問卷設計表述會對回答產生重大影響。在一項針對荷蘭成年人的調查中（詢問他們是否曾在開車時發送簡訊），加入「偶爾」或「僅在別無選擇時」等顧及面子的選項，所引出的答案與二分法的「是」或「否」選項截然不同。 雇主可以參考未經修飾的反饋，最顯而易見的是在 GlassDoor 等職場評價網站上。人工智慧有能力從大量非結構化的評論中，拼湊出一個清晰完整的輪廓。在最近的一篇論文中，倫敦政治經濟學院的湯姆·里德（Tom Reader）和亞歷克斯·吉萊斯皮（Alex Gillespie）在歐洲企業的員工評論中，尋找是否存在高壓文化的證據。那些暗示目標非常宏大、且在達成目標時不擇手段的評論，預示著這些公司未來將會經歷企業醜聞。 頻率是妥善設計的另一個面向。在長達一年的時間裡，很多事情都會發生變化；對於獲取員工情緒的最新動態而言，年度調查是一段漫長的等待。回溯性的評估也容易受到像是「峰終定律」（peak-end rule）等偏誤的影響，該定律描述了人們在回憶一項經歷時，會過度重視最極端以及結束時的時刻。在一個著名的實驗中，丹尼爾·康納曼（Daniel Kahneman）等人讓志願者經歷兩項令人不快的任務：第一項涉及將手放在冰水中一分鐘；第二項則是放 90 秒，不過在最後的 30 秒內，水溫稍微上升了一點。第一種體驗客觀上痛苦較少，但如果可以選擇，人們卻選擇重複第二種。 更短期的「脈搏」（pulse）調查雖然無法完全消除這些問題，但卻是收集更即時數據的一種方法。HappyOrNot 是一家芬蘭公司，負責製造你在機場和其他地方看到的那些笑臉反饋終端機，該公司也將這些機器安裝在企業內部，作為每日或每週追蹤員工情緒的一種方式。 好的設計和多元的信息來源有助於員工調查的成功。但沒有任何事情比「被看見對反饋採取行動」更重要。如果你說你的老闆具備香蕉般的溝通技巧，隨後卻沒聽到任何回應，你只會得到一種知道自己是對的、卻於事無補的微弱滿足感。沒有引發任何後續行動的調查，非但不能深化熱情，反而會加深諷刺與犬儒主義。 這一切引出了另一個悖論：調查在那些關心員工想法的組織中最有用。然而，關心員工想法的組織，往往較不需要進行調查。■ --Hosting provided by SoundOn