Reading an S-1

Snowflake's IPO materials (2020 Form 424B4 / S-1 family) are a clean teaching example of a usage-oriented data platform with SaaS-like metrics. Numbers below are from those IPO disclosures — illustrative of how to read, not an update of current results.

You are training your eye, not picking a stock.

The big idea

A prospectus is a guided tour. Management highlights the metrics that make the engine legible. Your job is to take the tour, then ask what would make those metrics less impressive — consumption that does not arrive, COGS that rise with usage, a cloud competitor, concentration.

The guided-tour analogy breaks if you memorize a single cell ("RPO was $426M") and call it analysis. The point is the trajectory and the definition.

Walking through it

Business model. Customers consume compute, storage, and data transfer. Revenue is recognized as capacity is consumed, not as a flat seat subscription. That makes product revenue and RPO central: RPO shows contracted future consumption; product revenue shows actual usage.

Key metrics the filing highlighted (selected periods in the prospectus):

• Product revenue scaling from on the order of ~$100M toward ~$250M+ across the compared snapshots — growth trajectory and definition, not one memorized cell.
• RPO growing from roughly $128M → $426M in one comparison set and $221M → $688M in another — contracted backlog expanding fast.
• Net revenue retention reported as high as ~158–180% in the periods shown — existing customers were expanding consumption dramatically.
• Customers with TTM product revenue > $1M rising from the teens into the ~40–50+ range — landing expanding enterprise accounts.

Suppose a cohort produced $10M of product revenue in year 1 and $16M in year 2 with no logos fully gone (churned customers contribute $0 in year 2 but stay in the denominator per Snowflake's described method). NRR ≈ 16/10 = 160%. Expansion is the growth engine. Pair it with gross margin: if consumption COGS rise as fast as revenue, the story is weaker than NRR alone suggests.

How to read it as an engineer. High NRR and rising large-customer counts imply the product embeds in workflows (switching costs), usage expands with the customer's data footprint, and reliability work has direct revenue leverage. Risk factors in such filings typically call out consumption variability, competition from cloud providers, and the cost of serving analytical workloads — all engineering-relevant.

When you pick a company to study today (Datadog, Cloudflare, Snowflake, and others), copy the checklist: revenue definition, RPO/cRPO, NRR, customer tiers, gross margin, S&M %, SBC %, cash from operations.

Common mistakes

• Memorizing 2020 cells as if they were current. The method transfers. The numbers age.
• NRR without gross margin. Usage businesses can expand revenue and COGS together.
• Ignoring risk factors that name your team's work — reliability, cost to serve, cloud competition.
• Treating RPO as recognized growth. It is a commitment. Consumption still has to happen.

Check yourself

Year-1 product revenue $10M, year-2 $16M, no logos fully gone. NRR?

About 160%, if churned customers stay in the denominator as described.

What should you pair with that 160%?

Gross margin. If COGS rose as fast as revenue, expansion is less impressive.

Key takeaways

• An S-1 is a guided tour of the engine at IPO. Take it, then stress-test it.
• Snowflake-style usage businesses highlight product revenue, RPO, NRR, and large-customer counts.
• High NRR plus rising $1M customers is a quantitative shadow of switching costs.
• Numbers here are 2020 teaching figures, not a live model.

Further learning

Snowflake IPO prospectus (424B4) on EDGARDocumentsec.gov SEC EDGAR full-text searchSEC.gov | EDGAR Full Text Searchsec.gov

Back to the map: reading company financials.