How to Analyze like a Financial Analyst

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