SG&A vs COGS Classification for Finance Teams

The line between SG&A and COGS is not paperwork. It decides where a dollar lands on the income statement, and that placement changes what gross profit and operating income actually tell you about the business. Revenue minus COGS gives you gross profit. Gross profit minus SG&A gives you operating income. Those two subtractions answer two different questions: how much does it cost to deliver what you sell, and how much does it cost to run the operation around it. Investors and lenders read them separately for a reason. A guide from Ramp, published in July 2026, points out that high COGS usually flags a production or pricing problem, while high SG&A usually flags operational inefficiency, the kind a company can fix with cost controls rather than a product overhaul. Mix the two lines up, and you're not just filing costs in the wrong drawer. You're changing the diagnosis.

The distinction also runs on timing. SG&A is an indirect cost that can't be pinned to one unit of output, so it gets expensed as soon as it's incurred. COGS is a direct cost, tied to a specific sale, and it gets expensed only when that good or service actually moves. SG&A itself splits into three buckets, and each behaves differently on the P&L. Selling expenses include commissions, advertising, content production, events, CRM tools, and sales travel, all aimed at generating revenue. General expenses cover the fixed cost of simply existing as a company: rent, utilities, insurance, office infrastructure. Administrative expenses cover the cost of managing the organization itself, things like executive salaries, legal, accounting, HR, and IT support. None of this is decorative categorization. It's the scaffolding every classification argument in this piece stands on.

The primary decision rule that resolves most classification disputes

Most of the disputes finance teams have over a given cost line resolve the moment they apply one test consistently: does the cost disappear when production stops, or does it stick around regardless? If it persists when production stops, it's SG&A. If it disappears, it's COGS. ExpensePoint's SG&A guide, published in September 2026, puts it in almost identical terms: if the cost would vanish when production stopped, call it COGS; if it would show up on the books whether or not anything sold that month, call it G&A. A parallel version of the same test asks it from the other direction: stop selling tomorrow but keep producing, and the costs you'd still be paying are likely COGS; stop producing but keep the lights on, and the costs you'd still be paying are SG&A.

A second test works alongside the first one rather than against it, and it's built specifically for service delivery rather than manufacturing: does the cost rise when you deliver more client work? If yes, treat it as Cost of Service, the services-business equivalent of COGS. If the cost would exist even during a month with zero delivery, it's SG&A. Both tests agree on the obvious cases. Raw materials, direct labor, and manufacturing overhead land in COGS. Sales salaries, advertising, executive compensation, and rent are SG&A. Once the classification call is made, the arithmetic behind it is trivial. The hard part was always the judgment call about which bucket the cost belongs in to begin with, not the math.

Why SaaS and Services Businesses Face Harder Calls

Both tests above were built for a factory floor. Raw materials, factory labor, and manufacturing overhead are separable, countable, and obviously tied to output. SaaS breaks that model at the foundation. A software company's marginal cost of producing one more unit is close to zero, and its biggest spending categories, sales teams, marketing, customer success, all sit nominally under SG&A. Meanwhile the costs that actually keep the product running and in customers' hands, hosting, DevOps personnel, customer support, retention-focused customer success, and embedded third-party software, are the real COGS. The rule still applies. It's just harder to see which side of it a given SaaS cost sits on, because the product isn't a physical thing coming off a line.

GAAP doesn't help here. There's no standard, prescribed definition of COGS for SaaS companies. The whole exercise depends on consistent internal logic measured against what the rest of the market does. That consistency is what makes gross margin numbers comparable from one company to the next. Beancount's guide notes that software and SaaS companies typically carry SG&A ratios far higher than manufacturers do, a structural fact about how costs distribute differently across a software P&L than a factory one. Investors lean on specific gross margin bands to size up whether a SaaS business looks services-heavy, functional, acceptable, or premium. Those bands are signals investors read to make a judgment about the business model, and misclassifying even one cost line can shift a company from one band to another while the underlying business stays the same. That's the actual danger: not a "failing grade," but a signal that no longer says what it's supposed to say.

Classifying the Costs That Appear in Every SaaS COGS Debate

Six cost categories generate most of the COGS/SG&A arguments that come up in SaaS finance, and each one has a defensible default that can be written down and applied the same way every time. Internal IT, company-wide software subscriptions, and office infrastructure are SG&A, since they serve the whole business rather than any specific act of delivery. And sales commissions, CRM tools, advertising, content creation, and demand-generation spend stay in SG&A no matter how tightly they correlate with revenue growth, because they're selling expenses by function, not delivery expenses. Expensify's guide, from June 2026, backs this up directly, listing marketing campaigns and content production as selling expenses inside SG&A. Each one has a clear default, and the exceptions (support that splits by purpose, mainly) are narrow enough to write down as a rule rather than argue about every quarter.

Customer success: the cost line that legitimately straddles the boundary

Customer success is the cost line most likely to get misclassified, and it gets misclassified in both directions, because one CS team often performs work that belongs on opposite sides of the boundary. CS activity tied to onboarding, adoption, and retention is operationally required to deliver what the customer already paid for, which puts it in COGS. CS activity aimed at expansion and upsell is a revenue-generation function, not a delivery obligation, which puts it in SG&A. Plenty of companies never make this split. They classify the whole CS org as one thing or the other because splitting it feels like more trouble than it's worth, and because CS reps themselves often don't track their own time by function. That avoidance doesn't produce a cleaner number. It produces a wrong one, since it blends a delivery cost and a sales cost into a single line and then reports the blend as if it measured one thing.

The fix is splitting CS headcount by time allocation or role definition, so the onboarding specialist and the account-growth specialist get counted separately even if they sit on the same team and report to the same manager. The test to apply is simple to state, even if it takes real internal digging to answer: does this CS activity exist because the contract requires it, or because the company is trying to grow the account? The answer sorts the cost. This is the least comfortable classification decision in the whole article, mostly because it demands actual data about how people spend their time rather than a policy statement about how a department is supposed to work. It's also the one that matters most, because gross margin and the sales-efficiency metrics calculated below it both depend on getting it right.

Content and Editorial Spend

Most content spend belongs in SG&A, and that's the correct default. But a growing slice of in-product content has a real COGS argument, one finance teams should weigh on its merits rather than dismiss out of habit. Blog posts, SEO articles, top-of-funnel editorial, and demand-generation content are marketing costs. They exist to build awareness and win customers, and that stays true no matter how directly they correlate with revenue, so they are SG&A. In-product tutorials, onboarding guides, and knowledge base articles work differently when a customer has to access them to get what they contracted for. That kind of content has a legitimate claim to COGS. Atlassian's setup illustrates the line well: in-product educational content that helps a user reach value faster or discover advanced features functions as part of service delivery, while anything published externally for editorial reach does not. The test to apply is whether the company would still have to produce the content if it stopped marketing. If yes, it may belong in COGS. If the content exists to attract or persuade, it belongs in SG&A.

One shift is making this call more consequential than it used to be. As AI-generated answers become a common way people discover products, spending aimed at brand visibility inside those answers, what's known as Generative Engine Optimization, is still SG&A marketing spend by function. But it behaves more like infrastructure than advertising, building compounding visibility rather than a return that disappears the day you stop paying for it. That affects how finance teams should think about content budget cuts: reducing SG&A-classified content spend that builds durable AI-answer visibility has different long-run cost implications than cutting a paid media campaign that stops working the moment it stops running. The classification stays the same either way. What changes is how a finance team should think about the tradeoff before cutting it.

What consistent misclassification actually does to the numbers finance teams rely on

A single misclassified cost line doesn't stay a single mistake. Applied the same wrong way quarter after quarter, it compounds into a gross margin figure that misleads everyone reading it, inside the company and out. Move sales and marketing costs into COGS and gross margin compresses artificially, making the company look more expensive to deliver than it actually is. That suppresses valuation and steers resource-allocation decisions about delivery capacity in the wrong direction. Move actual COGS costs into SG&A and the opposite happens: the margin line looks flattering, right up until an auditor unwinds it, and in the meantime it overstates how efficiently the company delivers and understates the true cost of serving each customer. A blended gross margin can look perfectly healthy against peer benchmarks while masking a delivery cost problem, because the classification behind it hasn't been consistent. The number is measuring something other than what it claims to measure.

Real-world roles rarely sit neatly on one side of the line, and inconsistency creeps in exactly at those blended roles. Expensify's June 2026 guide points to cases like a salesperson who also manages implementation, or a warehouse that serves both production and distribution, and resolves them the same way every time: tie the cost directly to producing a unit of product or delivering a service, and classify accordingly. For early-stage SaaS companies, a high SG&A ratio relative to revenue is normal during growth and doesn't need defending on its own. It's a ratio that's measured inconsistently from period to period, which blocks the company from ever proving that it's actually improving as it scales. Investors and auditors expect one thing above all else: that costs get classified the same way every period. Shifting expenses between COGS, R&D, and operating expenses without a documented reason is the specific pattern that leads to restatements.

Building a classification policy that holds up across periods and across reviewers

The biggest risk in SG&A/COGS classification is the lack of a written policy, which forces every new hire, every new vendor, and every new cost line to relitigate decisions the company already made months earlier. The fix starts with writing the rules down as policy rather than leaving them as something a few finance people remember: the persistence test, the delivery test, and the specific rulings already reached on every recurring gray-area cost, from CS headcount to content spend.

Ownership matters as much as documentation. The CS headcount split, the content cost review, and any role that blends functions need a named reviewer on a set schedule so the decision gets revisited rather than made once at setup. Automation should implement that policy, not replace it. Expense categorization tools and ERP rules are useful for applying a decision that's already been made, but a system that auto-categorizes new cost lines without a human checking them will eventually recreate the exact inconsistency the policy was built to prevent. When something genuinely new appears, an in-product AI-generated content line, a CS role that mixes delivery and expansion work, the classification call needs to happen before the first invoice gets coded, at the time the cost first occurs. And none of this works as a finance-only exercise. Teams that classify costs without input from content, CS, and product leadership will eventually get a call wrong simply because they don't know what the function actually does day to day. The policy conversation doubles as the information-gathering conversation.

None of this is about finding the classification that makes a given quarter's margin look best. It's about applying the same rule the same way, every period, regardless of what that does to the number. The SG&A/COGS boundary is a set of decision rules finance teams have to apply on purpose, not a bright line waiting to be found. It's a set of decision rules that finance teams have to apply on purpose, and a written policy is what turns that judgment into something repeatable instead of something that depends on who happens to be making the call that day.