想進 Indeed 的同學注意了，TA 家的數據崗位 OA 不是那種只刷一刷 SQL 就能搞定的類型，而是綜合考察你在資料庫操作、機器學習理論、以及概率推理上的理解深度。 我們這次的同學就遇到了三大類題型，每一類都有難點，好在有我們即時助攻，最後順利通過進入面試階段！ 這篇 blog 給大家復盤一下她當時的 OA 題目內容，以及我們在答題過程中的知識點提示。
Part 1： SQL 模組
Q1. Compare Department Average Salary to Company Average
Given two tables (Employees, Departments), write a SQL query to compare the average salary of each department to the company’s overallaverage salary. The output should include department name and the comparison result as “higher”, “lower”, or “same”.
Q2. Quiet Students in All Exams
Write a query to report the students (student_id, student_name) who are “quiet” in ALL exams. A quiet student is one whoparticipated in at least one exam but never got either the highest or the lowest score in any exam.
Q3. 3+ Consecutive Days with 100+ Visitors
Given a table of daily visitor counts, write a query to find all periods of at least 3 consecutive days where each day had 100 or more visitors.
Q4. Count Bank Visitors by Number of Transactions
Write an SQL query to count how many bank users visited the bank and did 0 transactions, 1 transaction, and so on.
總結：题型非常贴近真实业务，考查你如何用 SQL 分析行为数据或指标归因。涉及窗口函数、group by、having，以及逻辑判断等。
第 2 部分：機器學習模塊
Q1. Detecting EV Households from Hourly Electricity Usage
How would you detect households that own electric vehicles using hourly electricity consumption data?
Q2. Predicting Out-of-Stock Inventory
How would you predict which items are likely to go out-of-stock?
Q3. Regression & Classification Metrics
List various evaluation metrics for regression and classification tasks.
Q4. Pros & Cons of Mean Squared Error (MSE)
Explain the advantages and disadvantages of using Mean Squared Error as a loss function.
Q5. What Is Learning Rate in Gradient Boosting?
Define learning rate and describe its role in gradient boosting algorithms.
Q6. More Features than Rows – What’s the Impact?
What happens when the number of features exceeds the number of observations in a dataset? How can regularization methods like Lasso or Ridge help?
Q7. Effect of Multicollinearity on XGBoost Feature Importance
What’s the impact of high correlation between features on their importance ranking in XGBoost?
總結：机器学习题不会让你写代码，但会考你是否具备工程落地视角。答题时可以结合业务例子 + 模型策略，programhelp 学员一般会有标准模板提前准备好，考场上快速套用即可。
第 3 部分：概率模組
Q1. Explain Probability Distribution
What is a probability distribution? Describe its basic types and usage.
Q2. Generate Random 1–7 Using a Die
How would you generate a uniform random number between 1 and 7 using a single standard 6-faced die?
Q3. Normal to Uniform Sampling
If you draw from a normal distribution with known parameters, how do you generate samples from a uniform distribution?
Q4. Probability That Two Customers Are in the Same Partition
If 75 customers are randomly assigned to 3 equal partitions, what is the probability that Bob and Ben are in the same one?
Q5. Explain Bayesian Probability
What is the Bayesian approach to probability? Give a real-life application.
總結：此類題目主要看你能否用直覺 + 推導解釋概率事件，建議：多舉例 + 分步推理，有時比公式更重要。
FAQ：關於 Indeed 數據崗 OA 的常見問題
Q1. 這個 OA 是開放刷題形式還是即時提交？
A：是 HackerRank 平臺，限時答題，不支援多次提交或跳題。
Q2. SQL 部分用什麼語言？ 支援視窗函數嗎？
A：用的是標準 SQL，支援視窗函數、CTE、多層嵌套查詢等。
Q3. ML 部分有沒有 coding？
A：沒有 coding，都是選擇題或簡答題，考察你對模型理解。
Q4. 幾道題？ 時間多久？
A：一共大概 10 道题，SQL 3–4 道，ML 4–5 道，概率 2–3 道，时间大概 90 分钟。
Programhelp 學員實錄分享
這位同學是在收到 OA 的當天聯繫到我們，時間比較緊，我們幫她做了一個快速的 SQL 高頻題精講 + ML 理論講解梳理。
答題當天我們通過無痕連線方式遠程輔助，像 Q2、Q3 這種容易踩坑的題，都是我們提前梳理過的套路題型。 最後她成功通過 OA，順利拿到 VO 機會！
正在準備 Indeed 或其他 FAANG 數據崗 OA？
你可能也遇到這些問題：
不確定題型分佈，複習效率低？
SQL 卡頓、機器學習答題空泛沒思路？
一到考場就緊張出錯，明明會但寫不出來？
Programhelp幫你一次解決！
我們提供：
✅ 高頻真題預測 + 拆解講義
✅ 模擬實戰訓練 + 答題節奏指導
✅ 無痕遠端連線協助
已經有多位學員在我們輔助下順利通過 Indeed、Meta、Amazon 等 OA 環節，拿下心儀 offer！
想瞭解更多 OA 代寫方式？ 後台私信「OA」或直接留言，我們來幫你定製上岸方案！
