Video summary
How to Analyze like a Financial Analyst
Main summary
Key takeaways
Finance-focused summary (how to analyze like a financial analyst)
Core problem & context
- The presenter analyzes March 2026 financials after software subscription costs increased significantly in March.
- Goal: explain the variance vs budget and the resulting forecast FY26 P&L impact, supported by operational data.
Methodology / step-by-step framework (explicit process)
Step 1: High-level variance analysis (March 2026)
- Compute variance = revenue minus expense (as stated).
- Identify which line items are over budget.
Step 2: Investigate drivers of the variance
- Use transactional data filtered to cost of revenue items.
- Further filter to the specific item(s): software subscription and third-party tools.
- Break down by vendor supplier (e.g., via a pivot table) to identify the main cost driver.
- Drill down further to determine the cost model/usage driver using:
- engineering comments
- internal documentation linking usage to the financials
Step 3: Forecast the impact to annual targets
- Build the forecast directly into the P&L actual tab.
- Apply a rough growth rate for most items (generally aligned with budget).
- Treat known outliers separately—specifically Project Atlas.
- Compare FY26 forecast vs FY26 budget and quantify the net impact.
Key findings & numbers
Variance vs budget (March 2026)
- Software subscription and third-party tools: +$73,000 vs budget.
- A significant variance percentage is noted, but the exact percentage is not provided.
- Primary concern: subscription cost increase (revenue is comparatively less concerning).
Transaction-level driver analysis
- Vendor-level findings:
- Main driver identified as OpenAI, with Anthropic singled out as the primary driver.
- Anthropic cost increase (month-over-month):
- Anthropic increased by ~500% month over month.
- Subscription spend used in the explanation:
- Total subscription spend: ~$202,000 (from vendor breakdown summing to this amount)
- Anthropic March spend/cost: ~$87,500
Operational explanation (why Anthropic spiked)
- Engineering director context for Project Atlas:
- Runs from March through May at roughly the current run rate.
- In June, usage drops by ~40%.
- “Ship end of June” is mentioned as a timeline marker.
- Token/usage documentation ties usage to financials:
- Document 1: total Anthropic token usage in March totals $87,500.
- Document 2: Project-level department spend shows Project Atlas ~ $77,500 in March, stated as ~82% of total Anthropic spend (~$87K).
- Document 3: token usage by headcount, mapped to initiatives.
Forecasting assumptions & timeline (FY26 P&L impact)
General approach
- The forecast includes limited operational detail for items other than Anthropic, so the presenter uses a simplified method.
- Base assumption for most subscription cost items:
- a growth rate “maybe in line with budget” (no exact % provided).
- Budget baseline:
- $129,800 per month, based on a vendor spend profile from prior-quarter planning.
Project Atlas cost adjustment approach
- Expected run rate:
- March ~ $77,000
- June:
- apply 40% decrease
- July onward:
- shift to lower “production API” spend, about $18k–$20k per month
- “60% drop” referenced as the mechanism to reach ~$18k/month
- August–December:
- similar pattern to July
Quantified recommendation / outcome
FY26 forecast vs budget
- The presenter states the company is still expected to land “ahead of target” because subscription revenue upside offsets the subscription cost increase.
Incremental spend attributed to Project Atlas
- ~$333,000 incremental subscription cost expected due to Project Atlas.
Caution / limitation
- Other forecast items were modeled using a very general approach with no operational context, so forecast confidence is limited.
Instruments / entities / tickers mentioned
- No market tickers (stocks/ETFs/bonds) are mentioned.
- Vendors / AI services as cost drivers
- Anthropic
- OpenAI
- Project / internal initiative
- Project Atlas
Disclosures / disclaimers
- No explicit “not financial advice” disclaimer appears in the provided subtitles.
Presenters / sources
- Presenter: not explicitly named in the provided subtitles.
- Sources used within the case:
- an (unnamed) director of engineering
- internal documentation tabs (“documentation 1/2/3”) tied to token usage and departmental/headcount allocation