Essential Data Analysis Skills for Operations: From Metrics Systems to Data-Driven Decisions

Operations Without Data Skills Is Just "Going by Gut Feeling"

I spent four years in data operations at Meituan, working across three business lines — food delivery, hotel & travel, and community group buying. One feeling has grown increasingly strong: the core competitive advantage of operations is shifting from "writing good copy and running events" to "reading data and making data-driven decisions."

It's not that copywriting and events don't matter — it's that in the 2026 internet industry, virtually every operational decision needs data support. You say you want to run a promotion? Your boss asks: what's the expected ROI? You say users are churning? Your boss asks: at which step? What's the profile of churned users?

In this article, I'll systematically cover the data analysis capabilities operations professionals need — from building metrics systems to analysis methodologies, tools, and real-world cases.

1. Why Must Operations Professionals Understand Data Analysis?

1.1 Industry Trends Demand It

From 2024 to 2026, the internet industry has undergone a transformation from "extensive growth" to "refined operations." After traffic dividends disappeared, companies focus more on efficiency and ROI. This means:

• Every budget dollar needs data support
• Every operational action must be quantifiable and trackable
• The era of "gut-feeling" decision-making is over

1.2 Job Requirements Have Evolved

Open any recruitment platform and search for "operations" positions — you'll find that over 80% of job descriptions mention "data analysis skills." Whether it's content operations, user operations, or event operations, data analysis has become a baseline requirement.

1.3 Career Development Ceiling

I've seen many operations colleagues with strong execution skills who run great events but hit a ceiling at mid-senior levels. The reason is simple: senior operations roles require "strategic capability," and the foundation of strategy is data insight.

2. Core Operations Metrics System

2.1 North Star Metric

The North Star Metric is the single most important metric for the entire business — all operational actions should revolve around it.

Different businesses have different North Star Metrics:

2.2 Process Metrics

The North Star Metric is the result; process metrics are the path to achieving that result. Taking e-commerce as an example:

GMV = Traffic × Conversion Rate × Average Order Value × Repeat Purchase Rate

Each factor can be further broken down:

• Traffic = Organic traffic + Paid traffic + Social viral traffic
• Conversion Rate = Browse→Add-to-cart rate × Add-to-cart→Order rate × Order→Payment rate
• Average Order Value = Average product price × Quantity
• Repeat Purchase Rate = Users who purchased again within 30 days / First-time purchasers

2.3 Health Metrics

Beyond business metrics, you also need to monitor "health metrics" to ensure growth is sustainable:

• User Quality: Next-day retention rate, 7-day retention rate for new users
• Cost Efficiency: CAC (Customer Acquisition Cost), LTV/CAC ratio
• User Satisfaction: NPS (Net Promoter Score), complaint rate

3. Core Data Analysis Methods

3.1 Funnel Analysis

Funnel analysis is the most commonly used analysis method in operations, used to identify key stages where users drop off.

Case: Purchase conversion funnel for an e-commerce app

The data shows that the "Homepage → Product Detail Page" conversion rate is only 45% — this is the biggest drop-off point. Operations should focus on optimizing homepage product recommendations and search experience.

3.2 Retention Analysis

Retention analysis measures users' long-term value and product stickiness.

Common Retention Metrics:

• Day-1 Retention: Reflects the product's first impression
• Day-7 Retention: Reflects the product's short-term value
• Day-30 Retention: Reflects the product's long-term value

Three Shapes of Retention Curves:

1. Smile Curve (drops then rises): Indicates the product has long-term value — users come back
2. Flat Curve (drops then stabilizes): Indicates the product has a stable core user base
3. Continuous Decline Curve: Indicates the product can't retain users — needs urgent attention

3.3 Attribution Analysis

Attribution analysis answers "what caused this result?"

Common Attribution Models:

• First-Touch Attribution: Credits the channel the user first interacted with
• Last-Touch Attribution: Credits the channel the user last interacted with before converting
• Linear Attribution: Distributes credit equally across all touchpoints
• Time-Decay Attribution: Touchpoints closer to conversion receive more credit

Practical Application: Suppose a user's conversion path is "Douyin ad → Xiaohongshu seeding → WeChat Official Account → Direct search purchase." Different attribution models will give different channel value assessments. Operations teams need to choose the appropriate attribution model based on business characteristics.

3.4 Comparative Analysis

Comparative analysis is the simplest yet most effective analysis method:

• Time Comparison: This week vs. last week, this month vs. last month, this year vs. last year
• Cohort Comparison: New users vs. existing users, paid users vs. free users
• Channel Comparison: ROI comparison across different acquisition channels
• A/B Testing: Experiment group vs. control group

3.5 RFM Model

The RFM model is a classic method for user segmentation, based on three dimensions:

• R (Recency): Time since last purchase
• F (Frequency): Purchase frequency
• M (Monetary): Purchase amount

Through RFM, users can be classified into categories:

4. Common Data Tools for Operations

4.1 Excel / Google Sheets

Use cases: Daily data organization, simple analysis, report creation.

Must-have skills:

• Pivot Tables
• VLOOKUP / INDEX-MATCH
• Conditional formatting and data validation
• Basic chart creation

4.2 SQL

Use cases: Extracting and analyzing data from databases.

Operations professionals don't need to write complex SQL, but should at least master:

-- Basic query
SELECT user_id, order_amount, order_date
FROM orders
WHERE order_date >= '2026-01-01'
ORDER BY order_amount DESC;

-- Aggregation analysis
SELECT DATE(order_date) as date, COUNT(*) as order_count, SUM(order_amount) as total_gmv
FROM orders
GROUP BY

4.3 BI Tools

Common BI tools: Tableau, Power BI, Metabase, Sensors Data, GrowingIO.

The value of BI tools lies in:

• Visualizing data so non-technical people can understand it
• Automating reports to reduce repetitive work
• Real-time monitoring of key metrics

4.4 Data Tracking (Event Tracking)

Operations professionals need to understand basic data tracking concepts and communicate tracking requirements with technical teams:

• Page View Events: Which pages users visited
• Click Events: Which buttons users clicked
• Business Events: Key behaviors like ordering, payment, sharing
• User Properties: Registration time, acquisition channel, device type

5. Data-Driven Decision-Making Case Studies

Case 1: Optimizing a Promotion with Data

Background: An e-commerce platform planned a 618 promotion with a 5 million yuan budget.

Data Analysis Process:

1. Historical Data Review: Analyzed last year's 618 data and found that discount coupons had an ROI of 1:8, while direct price cuts only had 1:3
2. User Segmentation: Used the RFM model to segment users and found that "growth users" (high spending but low frequency) had the most room for improvement
3. Budget Allocation: 60% for discount coupons, 30% for targeted outreach to "growth users," 10% for new user acquisition
4. Real-Time Monitoring: Checked data hourly during the event, discovered an abnormally low conversion rate in one category, and promptly adjusted the recommendation strategy

Result: GMV grew 35% year-over-year, ROI improved from 1:5 last year to 1:7.

Case 2: Diagnosing User Churn with Data

Background: A content app's 30-day retention rate dropped from 25% to 18%.

Analysis Process:

1. By Channel: Found that organic traffic user retention was stable, but users from a specific feed ad channel had extremely low retention
2. By Behavior: Churned users shared a common trait — they didn't follow any creators within 3 days of registration
3. Root Cause: The ad channel's landing page promised "free novel reading," but the app's core content was short videos — user expectations didn't match
4. Solution: Adjusted ad creatives, optimized new user onboarding flow, added an "interest selection" step

Result: 30-day retention rate recovered to 23% within two months.

6. Data Analysis Questions in Interviews

6.1 Common Question Types

• "A product's DAU dropped by 10% — how would you analyze this?"
• "How would you evaluate the effectiveness of an operations campaign?"
• "Given this dataset, what conclusions can you draw?"

6.2 Answer Framework

Use the "Define Problem → Break Down Metrics → Form Hypotheses → Validate Hypotheses → Propose Solutions" framework:

1. Define the Problem: Clarify the metric definition, time range, and comparison baseline
2. Break Down Metrics: Decompose the big metric into smaller ones, find the factor with the biggest change
3. Form Hypotheses: Based on the breakdown, propose 2-3 possible causes
4. Validate Hypotheses: Explain how to use data to validate each hypothesis
5. Propose Solutions: Based on validated causes, offer specific solutions

6.3 Example Answer

Question: "An e-commerce app's order volume dropped 15% this week — how would you analyze this?"

Answer:

"First, I'd confirm some prerequisites: Is this 15% week-over-week or year-over-year? Is it an overall decline or specific to certain categories? Are there external factors (like a competitor's major promotion)?

Then I'd break down the order volume formula: Orders = DAU × Order Conversion Rate. First check if DAU changed — if DAU is stable, the issue is in conversion rate.

Next, I'd look at it by dimensions: by channel (which channel's users ordered less), by category (which category declined most), by user type (new vs. existing users).

If I find that existing users' order conversion rate dropped, I'd dig deeper: was there a recent feature change? Is a competitor running a major promotion? Has shopping cart data changed?

Finally, based on the identified cause, I'd propose targeted solutions."

Final Thoughts

Data analysis isn't a "fancy" skill — it's a fundamental competency for operations. You don't need to become a data scientist, but you need to:

1. Build data awareness: Before making any decision, ask yourself "what's the data support?"
2. Master basic tools: Excel and SQL are the minimum requirements
3. Learn to tell data stories: Transform data analysis results into actionable operations strategies

Data doesn't lie, but data doesn't speak for itself either. The value of operations lies in making data "speak" to drive business growth.