Cost of Goods Sold

What it measures

COGS is the period sum of every direct cost attributable to delivering your product or service: hosting and infrastructure, third-party software baked into the product, customer support and onboarding, payment processing, and professional services. It excludes sales, marketing, R&D, and any corporate overhead not tied to serving the customer — those sit below the gross-margin line in OpEx.

Why it matters

You track COGS because it is the foundation of gross margin — the profitability of your core product before you spend a dollar on growth or overhead. Investors read it to answer one question: does the model scale? If COGS grows slower than revenue, every new customer adds more profit and the business compounds; if COGS grows in lockstep, you are running a services shop dressed as software. COGS also sets the floor for pricing and the ceiling on unit economics — a customer's gross margin, not their revenue, is what has to outrun CAC.

How to read it

Read COGS as your cost of production per period, and always read it two ways: in absolute dollars and as a percentage of revenue. Rising absolute COGS can be perfectly healthy when revenue is rising faster — that is margin expansion. The warning sign is COGS rising as a share of revenue, which means you are getting less efficient as you scale. Compare your COGS ratio to your own prior period and to plan, then decompose any move: hosting, support labor, payment fees, and third-party costs each tell a different story. A ratio creeping upward usually points to over-supporting low-margin customers, an infrastructure bill outpacing usage, or a third-party API price hike.

What good looks like

Watch-outs

• Misallocating overhead into COGS. Shared finance, HR, or general admin costs belong in OpEx, not COGS, unless they are fully dedicated to customer delivery — folding them in overstates product cost and understates true gross margin.
• Using cash spend instead of accrual. Prepay hosting annually and you must spread it across the months it covers; booking the lump sum in one period breaks the match between COGS and the revenue it supported.
• Ignoring the fixed-versus-variable split. If part of your hosting and support is fixed and part scales per customer, lumping them together hides your real variable cost to serve — and that variable cost is what determines whether margin holds as you grow.
• Leaving one-time spikes in run-rate COGS. A data migration, security audit, or infrastructure rebuild is not recurring cost of delivery; strip it out so the underlying COGS trend stays readable.

Worked example