Data Analysis Skills Training

Because starting in supply planning is brutal... This document contains the supply planner starter kit: 1️⃣ Get the Basics Right ↳ Understand core supply chain concepts: inventory, lead times, and capacity 2️⃣ Learn Excel Inside Out ↳ Practice working with formulas, PivotTables, and Power Query. Supply planner roles lean heavily on Excel for data analysis and reporting. 3️⃣ Study Demand Forecasting ↳ Familiarize yourself with simple forecasting methods (for example, moving averages) and how forecasts tie into production and inventory 4️⃣ Get Comfortable with Data ↳ Develop basic data management skills like cleaning, merging, and interpreting information from multiple sources 5️⃣ Seek Cross-Functional Exposure ↳ Connect with sales, marketing, and logistics to see how supply planning impacts each function 6️⃣ Understand the “Why” Behind Processes ↳ Ask questions about safety stock policies, lead times, and supplier performance metrics to identify opportunities 7️⃣ Build a Problem-Solving Mindset ↳ Look for root causes, not just quick fixes. Use tools like “5 Whys” or simple process mapping Any others to add?

If you interview at Lockheed Martin for a Procurement Analyst or Supply Chain role, you might talk about SAP Ariba, FAR/DFARS compliance, or supplier risk management. If you interview at Boeing, you might discuss Lean supply principles or cost avoidance strategies. If you interview at Raytheon, you might bring up defense sourcing compliance or contract lifecycle optimization. But after 5 years in procurement, I can tell you with certainty — in every interview and every organization, one skill always separates average buyers from strategic professionals: Cost Modeling. Cost modeling is the backbone of procurement, no matter the company, system, or spend category. If you know how to break down cost drivers, analyze supplier pricing, benchmark against market trends, and build negotiation levers — you already understand 70% of what matters in procurement. Don’t get distracted by every new sourcing tool or dashboard. Platforms evolve — but the fundamentals of cost modeling never do. The frameworks might vary — should-cost, total cost of ownership, or parametric modeling — but the core stays the same: – Understand how a product or service is built – Know the variables that influence price – Identify where leverage exists – Present data that drives decisions Master how cost flows through your supply chain, and you’ll understand how every other procurement function — sourcing, contracts, and supplier performance — connects. Learn cost modeling deeply. You’ll thank yourself later. ____ Follow Erika T. for more content like this♻️

If I were leveling up as a data analyst right now, I’d focus on these 5 areas (that are actually changing our field with AI) 1. 𝐀𝐈-𝐀𝐮𝐠𝐦𝐞𝐧𝐭𝐞𝐝 𝐃𝐚𝐭𝐚 𝐂𝐥𝐞𝐚𝐧𝐢𝐧𝐠 → Use AI tools to detect anomalies, missing values, and outliers faster → Learn prompt-based data profiling to speed up EDA → Automate data transformation scripts with LLMs 📘 Resource: Introducing AI-driven BigQuery data preparation 𝐋𝐢𝐧𝐤: https://lnkd.in/d2W7D_Qt 2. 𝐒𝐦𝐚𝐫𝐭 𝐕𝐢𝐬𝐮𝐚𝐥𝐢𝐳𝐚𝐭𝐢𝐨𝐧 & 𝐃𝐚𝐬𝐡𝐛𝐨𝐚𝐫𝐝𝐬 → Use AI to generate dynamic narratives and summaries alongside charts → Explore tools that auto-suggest the best chart for your data → Learn how to build “ask-your-data” interfaces using embedded LLMs 🎓 Resource: Building Python Dashboards with ChatGPT (DataCamp Code Along) 𝐋𝐢𝐧𝐤: https://lnkd.in/dZinchP9 3. 𝐏𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐯𝐞 𝐀𝐧𝐚𝐥𝐲𝐭𝐢𝐜𝐬 & 𝐅𝐨𝐫𝐞𝐜𝐚𝐬𝐭𝐢𝐧𝐠 → Go beyond trends — learn time series modeling with AI support → Combine traditional models with AI-powered forecasts → Use AI to simulate what-if scenarios from business questions 📘 Resource: Practical Time Series Analysis by Aileen Nielsen (Book) 𝐋𝐢𝐧𝐤: https://lnkd.in/dUVkx4Gx 4. 𝐐𝐮𝐞𝐫𝐲 𝐎𝐩𝐭𝐢𝐦𝐢𝐳𝐚𝐭𝐢𝐨𝐧 𝐰𝐢𝐭𝐡 𝐀𝐈 𝐇𝐞𝐥𝐩 → Use AI copilots for writing/debugging complex SQL → Learn how to validate and optimize joins, filters, and aggregations with AI → Automate SQL documentation and data lineage tracking 🎓 Resource: DB-GPT: AI Native Data App Development Framework 𝐋𝐢𝐧𝐤: https://lnkd.in/dc_SpmM6 5. 𝐁𝐮𝐬𝐢𝐧𝐞𝐬𝐬 𝐒𝐭𝐨𝐫𝐲𝐭𝐞𝐥𝐥𝐢𝐧𝐠 𝐰𝐢𝐭𝐡 𝐀𝐈 → Practice generating insights in plain English from data tables → Learn how to convert raw metrics into executive summaries using LLMs → Build dashboards with auto-generated explanations for decision-makers 📘 Resource: Storytelling with Data by Cole Nussbaumer Knaflic (Book) 𝐋𝐢𝐧𝐤: https://lnkd.in/dhD6ZDgJ AI won’t replace your thinking, it will amplify it. Use it to automate the repetitive, and double down on the business impact only you can create. ♻️ Save it for later or share it with someone who might find it helpful! 𝐏.𝐒. I share job search tips and insights on data analytics & data science in my free newsletter. Join 12,000+ readers here → https://lnkd.in/dUfe4Ac6

If you’re just starting out in Procurement This one’s for you… Your first Procurement role probably entails a good dose of analysis A foundation skill critical for your career And often a prerequisite for landing a meaty buyer role The problem? Analysis techniques aren’t always taught well So here are my top tips to become a competent analyst: 📶 1. Define the right questions → Get clear on what you’re trying to solve → Align with stakeholders before diving into data 📶 2. Gather data methodically → Map out what you need – no more, no less → Catalogue your data and version everything (Bonus tip: date every file yy/mm/dd. Eg 250409 then file name) 📶 3. Clean and check your data → Consistent formats, clear headers, no rogue formula → If it looks messy, people will trust it less 📶 4. Handle missing or patchy data → No dataset is perfect → Use assumptions (and clearly flag them) 📶 5. Deploy the right tools → Tableau is easier than Power BI → Visuals matter – make trends jump off the screen 📶 6. Extract meaningful insights → Stay curious. Ask “so what?” → Translate what you see into what it means 📶 7. Go deeper only when needed → Zoom in on key areas → Use what-if and scenario modelling 📶 8. Tell the story → Build a simple narrative from complex data → Your audience should get it - fast These aren’t complex techniques Yet we often come across sub-standard analysis Master the basics! Frost Procurement Adventurer ⬇️ What analysis techniques do you swear by? 🔔 Follow Simon Frost for more on procurement fundamentals ♻️ Repost to help others starting their procurement careers

It took me 2 years to find the right workflow for using LLMs for data work, that I am satisfied with. I'll teach it to you in 5 minutes: I provide the following context (one point at a time): get LLM's outputs; ask LLM to review its output and iterate until I am satisfied with the results. 1. Define the problem The clearer your problem statement, the more targeted your solution will be. * Articulate why this problem matters * Quantify benefits for stakeholders * Keep scope tight (e.g., "improving DQ check effectiveness" vs "data quality") * Document assumptions A well-defined problem is already half-solved. 2. Understand your constraints * Map existing architecture/tools * Assess user proficiency realistically * Document hard timeline requirements * Identify technical debt implications The best solutions acknowledge constraints rather than fighting against them. 3. Validate that the problem is worth your time. Not all problems deserve to be solved, especially when resources are finite. * Calculate business impact * Consider opportunity cost * Distinguish between real and perceived benefits * Get stakeholder confirmation "Is this the highest-leverage problem I could be solving right now?" 4. Look for elegant workarounds before building something new. The best solution might be avoiding the problem entirely. * Refine existing systems * Eliminate problematic components (e.g., troublesome data models) * Leverage parallel work from other teams Ask if simpler approaches exist. 5. Design your solution with both present needs and future flexibility. * Match solution to problem + constraints + usage patterns * Evaluate alternatives with clear pros/cons * Plan for iterative delivery * Document key decisions 6. Write code that others will thank you for maintaining. Quality code follows patterns appropriate to the problem domain. * Use OOP for configs, FP for data pipelines * Implement type hints and static checks * Keep pipelines simple to re-run end-to-end * Avoid scattered conditional logic 7. Test & review * Create comprehensive test coverage * Add DQ checks for data models * Use LLMs for code review before team PRs * Inspect implementation thoroughly Aider + Sonnet has been my go-to tool to complement my LLM workflow. How do you use LLMs in your data work? Let me know in the comments below. --- Also, follow me for more actionable data insights. #data #dataengineering #LLM #AI

Most of us think we have a clear ICP. But when you look at the pipeline? It’s a wild mix of company sizes, industries, and personas — all getting the same campaigns & pitch. 3. Some deals move fast. Others stall for months. 2. Some channels print money. Others burn cash. 1. Some personas love the product. Others ghost after a demo. This isn’t a sales problem. It’s a segmentation problem. If we don’t know who our best-fit customers are, we’re running blind. Here’s how I segment 👇 Side note: Get the spreadsheet template along with step-by-step guide from my newsletter. Click the link in my profile to get a copy. 📌 Step 1: Pull Closed-Won Deals Your best customers leave clues — follow them. - Pull closed-won deals from the last 6-12 months. - Grab key data: Job titles, company size, industry, ACV, deal cycle. - Clean up your CRM (because it’s always messy). Why? Real data > gut feelings. Sell to who’s already buying. 🔍 Step 2: Enrich Your Data CRM data alone won’t cut it. Use Clay to enrich contacts (seniority, decision-making power). Pro Tip: Integrate Keyplay to your CRM have accurate industry tags added to your account. Add growth signals (hiring, funding, ad spend). Think of it as turning an old map into GPS with live traffic. 📊 Step 3: Find Your Winning Segments Look for patterns in your best deals: - Which industries & company sizes close the fastest? - What roles drive decisions? - Which channels bring in high-ACV deals? Example: Demos from Marketing VPs at Mid-market Dental SaaS = High ACV & 2x faster close rate. When they come from Paid Channel, the sales cycles are longer compared to when they come organically. Once you see the patterns, targeting becomes easy. ❌ Step 4: Learn from Closed-Lost Deals Your losses reveal what’s broken. - Pull & enrich closed-lost deals. - Identify why deals fell through — wrong fit? Wrong persona? Budget? - Which channels did these closed lost deals come from? - Compare all of these with your closed won patterns. Red flags to watch: - High demo volume, low conversion → Fix qualification/messaging. - Some industries never close → Stop targeting them. - Prospects ghost post-demo → Value prop isn’t landing. 📈 Step 5: Prioritize, Cut, Scale Put your segments into a 2x2 matrix: - High demo volume, high conversion → Scale this segment fast. - High demo volume, low conversion → Fix qualification/messaging. - Low demo volume, high conversion → See if it makes sense to prioritize based on if you have enough time, money, and people. - Low demo volume, low conversion → Stop wasting effort. Why? More focus = more predictable pipeline 🚀 👆Link to the template along with the full guide in my latest newsletter. Grab it by clicking on the link in my profile.

Bad account prioritization is killing your pipeline. The biggest mistake companies make? Using flimsy ICP criteria that’s:  ↳ too broad ↳ vague ↳ based on a “gut feeling.” At first, it seems like casting a wide net will bring in more meetings: "We closed some deals in this industry. They're all a good fit." But guess what? It doesn’t. When your ICP isn’t ridiculously clear, SDRs: ↳ Waste time working accounts that will never buy. ↳ Burn out from working harder, not smarter. ↳ Miss out on accounts that do want to buy. And because of this, your team spins its wheels: ↳ More activity, less pipeline. ↳ More frustration, fewer wins. Instead, here’s how to pinpoint the right accounts: 1). Analyze closed-won deals. What patterns do you see?  ↳ Industries and company size  ↳ Decision-maker personas ↳ Buying triggers ↳ Deal velocity. 2). Focus on intent signals. Who’s actively searching for solutions in your category? Use tools that track website visits, content downloads, or competitor research to prioritize the accounts already in-market. 3). Use engagement data. Which accounts are interacting with your brand? Track email opens, clicks, LinkedIn engagement, to find interest. 4). Prioritize by revenue potential. Which accounts have the need, budget, and urgency to buy? Align outreach to the accounts that offer the biggest ROI. 5). Combine firmographics with behavior. It’s not just about industry or company size. Layer in real-time activity to prioritize the accounts that are showing actual buying intent. Double down on those patterns. 6). Refine over time. Your ICP isn’t static. Revisit to reflect market changes and deal insights. Everyone knows the saying: "Try to sell to everyone, ultimately close no one." If you're involved with identifying ICP in your company, don't be lazy. Do the work up front. It sucks really, really bad. But, if you focus, you can have it done in under a week. You owe it to your top-of-funnel teams to give them a crystal-clear north star. Don't wing it. A perfect list gets you a minimum of 50% of the way there. The rest is process & messaging. Happy Selling! — Enjoyed this? Consider resharing ♻️ for others. Hit the 🔔 to FOLLOW Keith M. Laughner for more.

Stop treating QBRs like relationship check-ins. They're expansion goldmines hiding in plain sight. Week 2 of building my QBR Intelligence Agents, and I'm starting with the one that unlocks your Land & Expand revenue lever: The Pattern Spotter. 𝗧𝗵𝗲 𝗣𝗮𝘁𝘁𝗲𝗿𝗻 𝗦𝗽𝗼𝘁𝘁𝗲𝗿: 𝗬𝗼𝘂𝗿 𝗘𝘅𝗽𝗮𝗻𝘀𝗶𝗼𝗻 𝗢𝗽𝗽𝗼𝗿𝘁𝘂𝗻𝗶𝘁𝘆 𝗗𝗲𝘁𝗲𝗰𝘁𝗼𝗿 Here's what sellers miss: expansion signals are everywhere. The Pattern Spotter finds them in <30 seconds. 𝗣𝗮𝘁𝘁𝗲𝗿𝗻𝘀 𝗧𝗵𝗮𝘁 𝗦𝗰𝗿𝗲𝗮𝗺 "𝗘𝘅𝗽𝗮𝗻𝗱 𝗧𝗵𝗶𝘀 𝗔𝗰𝗰𝗼𝘂𝗻𝘁": 𝗨𝘀𝗮𝗴𝗲 𝗦𝗽𝗶𝗸𝗲𝘀 • Login frequency jumps 40%+ = new initiative launching • Feature adoption spreads to new departments = organic growth • Power users hitting system limits = ready for enterprise tier 𝗕𝗲𝗵𝗮𝘃𝗶𝗼𝗿 𝗦𝗵𝗶𝗳𝘁𝘀 • Exporting data weekly → daily = embedding in workflows • Support tickets shift from "how to" → "can we" = maturity signal • Multiple stakeholders joining QBRs = strategic importance rising 𝗘𝘅𝘁𝗲𝗿𝗻𝗮𝗹 𝗧𝗿𝗶𝗴𝗴𝗲𝗿𝘀 • LinkedIn shows 20%+ headcount growth • New VP hired from company using your premium features • Funding announcement or acquisition = budget unlock But here's the kicker—these patterns vary by segment. Enterprise accounts expand differently than SMBs. SaaS patterns differ from services. Your patterns are unique to YOUR business. 𝗕𝘂𝗶𝗹𝗱 𝗬𝗼𝘂𝗿 𝗢𝘄𝗻 𝗣𝗮𝘁𝘁𝗲𝗿𝗻 𝗦𝗽𝗼𝘁𝘁𝗲𝗿 (𝗦𝘁𝗮𝗿𝘁 𝗧𝗼𝗱𝗮𝘆): 1. Pick your top 3 expanded accounts from last quarter 2. List every data source you checked during expansion 3. Ask: "What pattern predicted the expansion?" 4. Document it. That's your Pattern Spotter's job description. Example from last week: "Customer went from 50 to 200 licenses after their power user became head of operations." Pattern identified: Power user promotion = expansion opportunity. 𝗙𝗼𝗿 𝗦𝗮𝗹𝗲𝘀 𝗧𝗲𝗮𝗺𝘀 𝗙𝗼𝗰𝘂𝘀𝗲𝗱 𝗼𝗻 𝗟𝗮𝗻𝗱 & 𝗘𝘅𝗽𝗮𝗻𝗱: Your QBRs should surface: • Which accounts are primed for growth • What specific expansion play to run • Who to engage and when What expansion pattern do you see repeatedly that you wish was automated?

🚨 Want to automate your entire data workflow using LLMs? With LangChain, you can build agents that: -Write SQL -Clean data -Create plots -Generate reports -And even query APIs Here are the Top 10 LangChain Tools every data scientist should know: 🧵 1. create_react_agent This is your go-to method for creating a ReAct-style agent — one that reasons and acts with tools. Use it to build intelligent assistants that: -Clean data -Generate SQL - Explain insights 2. Tool Wrap any Python function as a Tool so your LLM can call it. Example: a plot_chart() function becomes callable via prompt. Critical for transforming LLMs from passive text generators into active agents. 3. AgentExecutor The command center for your agent. It selects tools, handles memory, and coordinates execution. Useful for multi-step workflows like: - Query → clean → analyze - "What happened in Q4?" → full insight pipeline 4. PythonREPLTool Gives your agent access to a live Python shell. Enables: - Chart creation - Math calculations - Custom logic Tip: Combine this with Tool wrappers to control behavior. 5. Pandas DataFrame Agent Upload a DataFrame, ask questions, get analysis. Examples: - “Drop missing rows” - “Plot a bar chart” - “Give top 5 outliers” Ideal for LLM-powered Exploratory Data Analysis (EDA). 6. SQLDatabaseChain Let your agent speak SQL. Input: “What was revenue by month last year?” → Generates SQL → Queries your database → Returns results Try SQLDatabaseToolkit to go even further. 7. create_history_aware_retriever + create_retrieval_chain The new & modular way to do contextual document Q&A. - create_history_aware_retriever: Adds memory - create_retrieval_chain: Handles document grounding 8. AI Data Science Team (Pandas Analyst Agent, SQL Analyst Agent) AI DS Teram = curated agents for fast prototyping. Great for: - Structured queries - Data transformations - Visual summaries Plug & play for faster agent dev. 9. Memory Want continuity in a conversation? Use ConversationBufferMemory or CombinedMemory to help your agents remember: - Previous questions - Chat history - Past tool calls This is key for multi-turn workflows. 10. LangGraph (Advanced) Build multi-agent, stateful, logic-aware workflows. LangGraph = LangChain + DAGs - Perfect for orchestrating pipelines - Use conditional routing - Build AI agents with memory + context === WANT TO BECOME A GENERATIVE AI DATA SCIENTIST IN 2025? On Wednesday, June 11th, I'm sharing one of my best AI Projects: How I built an AI Time Series Forecasting Agent with Python Register here (740+ registered): https://lnkd.in/gGKsiqKi