Academy
All courses / Worth Backing? — SQL and Python for Startup Data and Investment Decisions

Worth Backing? — SQL and Python for Startup Data and Investment Decisions

Intermediate Python

What this course is

An accelerator's investment committee meets to decide its spring round: 80 applications are waiting, the round holds $2,500,000 and room for at most ten cheques, and four years of records are there to learn from, 920 decided applications and 91 funded startups followed month by month. The chair asks the analyst five questions, from "What did four years of decisions get us?" to "Which ten of the eighty should we back?". This is the Academy edition of the LucidPlexus field book Worth Backing?: seventeen lessons, one per chapter, with the book's figures and every SQL and Python example runnable on the page. It teaches SQL and Python together, the way analysts work now: SQL (DuckDB, inside Python) shapes the data, pandas and matplotlib analyze and draw it, and the analyst turns the result into an insight and a decision.

You start with both languages from their first lines, then filter, total, join and rank the raw tables in SQL, meeting the faults planted in them as they get in the way, and clean all thirty in a layer of views with a record of every decision. Then you move the clean data into pandas and ask what a single number cannot show: how money is spread, whether a comparison between groups is fair, what selection hides, and how the funded startups grew, lost customers and spent their cash. The last part chooses charts that answer questions, builds a market map and a one-page dashboard, and turns the evidence into a scorecard, ten picks within the budget and a memo, followed by two checklists to keep beside the keyboard. Each lesson has a quiz and two Python practice challenges with automatic marking.

What this course proves

The certificate shows you can take raw tables to a decision: shape and clean data in SQL with a record anyone can rerun, analyze it in pandas with fair comparisons and honest uncertainty, draw charts that do not mislead, and write a recommendation that states what would change it. It is awarded when all lessons are complete, every lesson quiz is passed and the final project is approved.

Data

Every lesson, the practice and the final project use six tables about Lanternfield, a fictional accelerator and seed fund, and the 1,000 startups that applied to it from January 2022 to March 2026: invented data built to behave like an accelerator's records, with 30 faults planted on purpose. Nothing in the course is investment advice. The code reads each table as data/<name>.csv, for example read_csv('data/funding_outcomes.csv') in DuckDB. All six come from one dataset page: https://lucidplexus.com/community/dataset.php?slug=digitalfinance.

The final project

The spring-round investment memo, end to end, in the career-project IDE in your browser: meet the six tables and the funnel, rebuild the clean layer and reconcile its log, measure the signals four years of outcomes support, score the pending applications, choose ten picks within the budget and test them, draw the memo's chart and write the one-page memo. It needs no Community membership and is reviewed by a person.

Community members can also take the track Worth Backing?, which does the same work as one notebook for the whole book, with an automatic check for each chapter.

Start this course

What you'll learn

  • Write SQL and Python from their first lines, together in one session: DuckDB queries inside Python, results as pandas DataFrames
  • Filter, total, group and rank tables in SQL, with the right denominators, NULL handled on purpose and medians for skewed money
  • Join tables without losing or multiplying rows, and place each application in its market
  • Clean data in a layer of SQL views with a cleaning log, never editing the raw files
  • Analyze distributions, outliers and fair comparisons in pandas, with expected rates and bootstrap intervals
  • Measure growth, churn, cohorts, runway and follow-on rounds from monthly data
  • Draw honest charts, a market map and a one-page dashboard with matplotlib
  • Build a scorecard with stated weights, test its sensitivity, choose within a budget and write an investment memo

17 lessons · 23.8 h

  1. A Thousand Applications, Ten Investments Quiz 50 min
  2. SQL and Python in One Session Quiz 115 min
  3. Meet the Tables Quiz 105 min
  4. Filters, Sorting and the First Shortlist Quiz 100 min
  5. Totals, Averages and Rates Quiz 100 min
  6. Joining Plan, Pitch and Revenue Quiz 90 min
  7. Ranks, Shares and Growth Quiz 100 min
  8. Cleaning in SQL, with a Record Quiz 135 min
  9. From Query to DataFrame Quiz 100 min
  10. Distributions, Outliers and the Median Quiz 75 min
  11. Comparing Segments Fairly Quiz 80 min
  12. Time, Cohorts and Growth Quiz 90 min
  13. Choosing the Chart Quiz 65 min
  14. The Market Map and the Dashboard Quiz 55 min
  15. Writing the Insight: the Investment Memo Quiz 75 min
  16. Ten Checks Before an Insight Reaches the Committee Quiz 45 min
  17. Ten Ways a Startup Chart Misleads Quiz 45 min

How you earn the certificate

  1. Complete at least 100% of the lessons.
  2. Pass each of the 17 lesson quizzes with 70% or more. You can retake them.
  3. Build the final project, Worth Backing? — the spring-round investment memo, and have it approved by a reviewer.

Every certificate has a public link employers can use to check it, and you can add it to LinkedIn.

Who it's for

  • Analysts and students who want the whole workflow, from question to decision
  • People in or near venture capital, accelerators, consulting, corporate strategy or business intelligence
  • Founders who want to see their numbers as an investor does

Before you start

  • None: the course teaches SQL and Python from their first lines, faster than a first course
  • Green on Paper, the first course of the path, is a gentler start if Python is new to you
  • A browser: everything runs in it, nothing to install