Skip to content
7-Step Data Analysis SOP: Standardize Workflows with AI & BI Integration
Big Data 23 min read

7-Step Data Analysis SOP: Standardize Workflows with AI & BI Integration

This article presents a seven-step standardized operating procedure for data analysis, combining AI-assisted requirement breakdown, data preparation, indicator design, and analysis with BI-powered integration, visualization, and tracking to transform fragmented analytical work into a reusable, end-to-end workflow.

Data Integration and Governance
Data Integration and Governance
Data Integration and Governance

Overview

The author describes a complete data analysis Standard Operating Procedure (SOP) built by combining an AI assistant (WorkBuddy) for analytical thinking and code generation with a BI platform (FineBI) for data integration, indicator management, and dashboarding. The goal is to replace scattered, repetitive manual work with a structured, reusable workflow that covers requirement clarification, data preparation, indicator design, analysis, visualization, reporting, and continuous tracking.

Step 1: Clarify the Analysis Requirement

Many analysts start by pulling data immediately. The article argues the first step is to convert vague business requests into concrete analytical questions. Four elements must be defined: analysis goal (what problem to solve), analysis object (company, department, product, customer, project), analysis period (day, week, month, quarter, year), and analysis dimensions (time, region, product, customer, channel).

Example: A boss asks: "Company profit is poor, analyze it." This expands into: Is revenue down or cost up? Is overall margin falling or product mix shifting? Are sales or admin expenses out of control? Which region/product/customer drags profit? What operational actions should follow?

WorkBuddy prompt: Provide the analysis topic, object, period, and purpose; ask WorkBuddy to decompose goals, list core business questions, design an analysis framework, list required data fields, and suggest dimensions and metrics.

Step 2: Data Collection and Preparation

Enterprise data lives in multiple systems (ERP, finance, CRM, HR, business systems) with inconsistent field names, periods, and calculation rules. Preparation involves four tasks:

Data collection: Identify source system and table for each field.

Data cleaning: Handle duplicates, missing values, outliers, and errors.

Data transformation: Unify date formats, product/customer codes, and data types.

Data linking: Join datasets on key fields such as customer, product, order.

Example: Product profitability analysis requires joining revenue from finance with procurement cost, production cost, and product master data.

WorkBuddy usage: Assist writing SQL/Python for cleaning and transformation, and document cleaning rules. FineBI role: Connect disparate sources once, build a unified data foundation so analysts don't repeatedly export and stitch tables.

Step 3: Build a Layered Indicator System

Dumping all metrics into a report creates noise. The article recommends a hierarchical indicator system aligned with business goals. For enterprise operations analysis, six top-level dimensions are suggested:

Revenue analysis: revenue, growth rate, target attainment.

Profitability analysis: gross margin, net margin, ROE, ROA.

Cost & expense: unit cost, sales expense ratio, admin expense ratio.

Customer analysis: new customers, repurchase rate, customer profit contribution.

Operations analysis: inventory turnover, accounts receivable days.

Cash flow analysis: operating cash flow, collection rate, cash conversion cycle.

Each indicator must have a definition, formula, data source, statistical period, statistical caliber, and owning department. Example: Gross margin = (Revenue - COGS) / Revenue * 100%. If departments use different revenue/COGS definitions, margins will disagree.

WorkBuddy prompt: Given a topic, output a full indicator system with level-1 dimensions, level-2 metrics, definitions, formulas, sources, business meaning, and suggested charts. FineBI role: Freeze these indicators into dashboards so all departments share the same caliber.

Step 4: Perform Data Analysis – Six Core Methods

Basic YoY/MoM comparisons are not enough. The article details six methods with concrete illustrations:

Comparison analysis: Actual vs. budget, prior year, prior month. Example: Budget revenue 10M, actual 9M, attainment 90%.

Trend analysis: Continuous time-series to judge improvement or deterioration.

Structural analysis: Break down totals by product, customer, region, department. Example: Identify which products truly drive profit.

Attribution analysis: Decompose result changes into contributing factors. Example: Profit decline split into price, volume, mix, unit cost, period expenses.

Correlation analysis: Examine relationships between metrics. Example: Did sales expense growth bring new customers? Does inventory rise accompany sales drop?

Predictive analysis: Combine history, open orders, business plans, and key drivers to forecast revenue, profit, cash flow.

Key principle: Analysis does not end at metric calculation; it must use metric relationships to uncover root business drivers.

WorkBuddy prompt: Provide masked data samples, field descriptions, and analysis goal; ask for analysis code, step design, and statistical interpretation. Caution: AI cannot determine root cause from description alone; conclusions must be validated on real data and business verification.

FineBI role: Build interactive dashboards with filtering, linkage, and drill-down so analysts can continuously explore around business questions.

Step 5: Data Visualization – Charts Serving Analysis Goals

Tables suit precise verification; charts reveal trends, structure, anomalies quickly. The article maps chart types to scenarios:

Line chart: revenue, profit trends.

Bar chart: department/product performance comparison.

Stacked chart: revenue/cost structure changes.

Donut chart: expense/product revenue share.

Funnel chart: customer/sales conversion.

Scatter plot: relationship between two metrics.

Waterfall chart: profit change decomposition.

Map: regional sales/distribution.

Heatmap: anomaly distribution, performance gaps.

Design rule: Charts must serve the analytical objective, not aesthetics. A financial dashboard should show revenue/cost/profit trends, margin changes, expense structure, budget vs. actual, AR and cash flow. A manufacturing dashboard adds output, OEE, yield, inventory turnover, order fulfillment.

FineBI capability: Assemble charts into themed dashboards with filters, linkage, drill-down so managers move from KPIs to operational detail.

Step 6: Write the Analysis Report – From Data to Decisions

Stakeholders need conclusions, not chart galleries. A complete report should contain six sections:

Analysis background: purpose, object, period.

Core conclusions: key findings in a few sentences.

Metric performance: core KPIs vs. targets.

Problem diagnosis: abnormal changes and influencing factors.

Improvement recommendations: specific actions tied to problems.

Follow-up plan: owner, deadline, acceptance metric.

WorkBuddy prompt: Feed verified results, key charts, and context; ask for a management report with the above structure, professional language, conclusion-first, data-accurate, no fabricated data or unverified causes.

Reminder: Analyst must still verify every number, conclusion, and recommendation in the final output.

Step 7: Establish a Tracking Mechanism – Close the Loop

Without tracking, problems recur. The article proposes an issue tracking table with columns: Issue, Improvement Action, Responsible Department, Deadline, Acceptance Metric. Example rows:

Gross margin decline → Optimize product mix → Sales Dept → Month-end → Blended gross margin.

Inventory pile-up → Adjust production plan → Production Dept → 2 weeks → Inventory turnover.

Slow collection → Key-account collection plan → Finance + Sales → Weekly tracking → Overdue AR.

Expense over budget → Assess low-ROI spend → Business units → Next month → Budget variance rate.

FineBI role: Build an improvement tracking dashboard showing key metrics, progress, and target attainment for regular management review and operational adjustment.

Appendix: Six Reusable WorkBuddy Prompt Templates

The article provides ready-to-use prompts for each phase:

Requirement decomposition: Role, topic, object, period, purpose → goals, questions, framework, fields, dimensions/metrics.

Data cleaning & processing: Source, fields, issues → structure check, duplicate/anomaly detection, cleaning rules, format unification, executable SQL/Python.

Indicator system design: Topic → level-1 dimensions, level-2 metrics, definitions, formulas, sources, meaning, suggested charts.

Analysis & diagnosis: Real data → YoY/MoM/budget comparison, trend, structure, anomaly detection, factor decomposition, root-cause hypotheses (all data-backed, distinguish fact/inference/hypothesis).

Visualization dashboard design: Topic, audience, core metrics → layout, KPI cards, chart-to-metric mapping, modules, filters/linkage, drill paths.

Report generation: Verified results → background, core conclusions, metric performance, problems, factor analysis, recommendations, action plan (no fabrication).

Prompt 1: Requirement Decomposition

你是一名资深数据分析顾问。
我现在需要完成一项数据分析任务。
分析主题:【填写分析主题】
分析对象:【填写企业/部门/产品/客户】
分析周期:【填写时间范围】
分析目的:【填写需要解决的问题】
请帮我完成:
1. 拆解本次分析目标;
2. 明确需要回答的核心业务问题;
3. 设计完整分析框架;
4. 列出需要采集的数据字段;
5. 给出建议分析维度和指标。
要求:分析逻辑清晰,贴合实际业务,不要堆砌无关指标。

Prompt 2: Data Cleaning & Processing

请协助我完成以下数据处理任务。
数据来源:【填写数据来源】
数据字段:【填写字段】
数据问题:【填写重复值、缺失值、异常值等】
请完成:
1. 检查数据结构;
2. 识别重复和异常记录;
3. 制定数据清洗规则;
4. 统一字段格式和统计口径;
5. 给出可执行的SQL或Python处理方案。
要求:不得擅自删除有效数据,所有数据处理规则需要说明。

Prompt 3: Indicator System Design

请根据【分析主题】,建立一套完整的业务指标体系。
请按照以下结构输出:
1. 一级分析维度;
2. 二级指标;
3. 指标定义;
4. 计算公式;
5. 数据来源;
6. 指标分析意义;
7. 建议使用的图表。
要求:指标之间逻辑清晰,避免重复,兼顾财务指标和业务指标。

Prompt 4: Data Analysis & Diagnosis

请根据我提供的真实数据,开展经营分析。
重点完成:
1. 同比、环比和预算对比;
2. 指标趋势分析;
3. 业务结构分析;
4. 异常指标识别;
5. 关键影响因素拆解;
6. 提出需要进一步验证的根因假设。
要求:所有结论必须有数据支撑;
区分数据事实、分析推断和待验证假设;
不得编造数据和原因。

Prompt 5: Visualization Dashboard Design

请根据以下分析目标,设计一套BI可视化分析看板。
分析主题:【填写主题】
使用对象:【管理层/财务部/销售部/生产部】
核心指标:【填写指标】
请输出:
1. 看板整体布局;
2. 核心KPI卡片;
3. 图表类型及对应指标;
4. 页面模块划分;
5. 筛选和联动设计;
6. 数据下钻路径。
要求:突出核心指标,减少无效图表,符合实际业务分析场景。

Prompt 6: Report Generation

请根据我提供的真实分析结果,撰写一份经营分析报告。
报告结构:
1. 分析背景;
2. 核心经营结论;
3. 关键指标表现;
4. 主要经营问题;
5. 问题影响因素;
6. 改善建议;
7. 后续行动计划。
要求:
语言专业、凝练;先结论后分析;
数据准确、逻辑完整;不夸大分析结果;
不编造未经验证的业务原因。

Closing Thoughts

The author emphasizes: AI (WorkBuddy) accelerates analytical efficiency; BI (FineBI) makes data continuously valuable; but real business judgment and decisions remain human. The end-to-end flow is: Clarify Needs → Prepare Data → Design Indicators → Analyze → Visualize → Report → Track & Review. Integrating both tools into the full workflow ensures every analysis has method, data, conclusion, and follow-up action.

Original Source

Signed-in readers can open the original source through BestHub's protected redirect.

Sign in to view source
Republication Notice

This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactadmin@besthub.devand we will review it promptly.

business intelligencedata analysisindicator systemSOPdata cleaninganalysis methodsdata visualizationreportingFineBIWorkBuddy
Reader feedback

How this landed with the community

Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.