ProjekinAja. Founder's Playbook Stage 3 · Revenue Architecture Lesson 3.6
● Stage 3 · Lesson 3.6

Gross Margin — The Number That Controls Your Destiny

Gross margin is the single most consequential number in a software business — here is how to compute it honestly and what it tells you about whether the business can ever be great.

⏱ ~24 min read 🧮 Unit economics 💰 Venture fundability

🎯 By the end you'll understand

What gross margin actually measures and why it is the structural ceiling on everything downstream — company valuation, ability to reinvest, venture fundability, and long-run profitability. How to compute it honestly for a hardware-plus-software business, including every COGS component that most founders leave out: hosting, inference compute, third-party API fees, support labor, customer success attributed to service delivery, and hardware cost of goods. The gross margin targets by business type that determine how investors and acquirers value you. And critically: why a 40% gross margin business cannot raise venture capital at a software multiple — and why hardware-software founders typically discover this at Series A when it is nearly impossible to fix quickly.

KEY INSIGHTGross margin is not a finance metric — it is a destiny metric

A business with 40% gross margin can be a fine business. It cannot be a venture-backed software business — and no amount of growth, product-market fit, or NPS will change that. The multiple compresses, the investors leave, and you are left running a profitable small company that nobody wanted to fund at scale.

Every time a Forge sensor ships to a manufacturing plant, two numbers are born simultaneously: the revenue number and the gross margin number. The revenue number tells you how much you charged. The gross margin number tells you how much you actually kept — and that retained fraction is the only money available to fund sales, R&D, G&A, and eventually profit. A business with high gross margin has structural leverage: every additional dollar of revenue produces a large contribution to funding the rest of the company. A business with low gross margin runs a treadmill — more revenue, more cost, marginal improvement in the economics of the whole machine.

🔑 Why it controls everything downstream

Gross margin is the structural ceiling on your operating leverage. If your gross margin is 40%, you can spend at most 40 cents of every revenue dollar on sales, R&D, G&A, and profit combined. If it is 75%, you have 75 cents. That 35-cent difference, at $10M ARR, is $3.5M per year of additional investment capacity — the difference between three sales reps and nine, between one product team and three. Over a 5-year growth trajectory, that margin difference compounds into a fundamentally different company. Gross margin is not one of several important metrics. It is the single number that sets the upper bound on what the business can ever become.

DEFINITIONWhat gross margin actually is — and what it is not

Gross margin is simple in concept and routinely miscalculated in practice. The formula:

📐 Gross margin formula

Gross Profit = Revenue − Cost of Goods Sold (COGS)
Gross Margin % = Gross Profit ÷ Revenue × 100

The critical question is not the formula — it is what belongs in COGS. COGS captures every cost that is directly and causally linked to delivering the product or service to the customer. It does not include sales & marketing, R&D, or G&A. The precision of your gross margin number depends entirely on the completeness of your COGS definition.

The most common founder error is computing gross margin as: revenue minus hardware cost of goods. For a pure hardware business, that might be defensible. For a hardware-plus-SaaS business like Forge — where you are delivering ongoing software, compute, support, and success services — it dramatically overstates your true margin. The investors and acquirers who will eventually price your business know this. If your numbers do not reflect it, they will restate them during diligence and the conversation will become unpleasant.

Revenue $0 Revenue 100% Hardware Cloud/Inference 3rd-party APIs Support labor Implementation COGS layers Gross Profit ~60% COGS ~40% Gross margin Funds: Sales & Mktg R&D G&A Profit
Gross margin is the fraction of revenue that remains after every direct delivery cost. The remaining fraction — not the top-line revenue — is what funds the rest of your business. Illustrated at 61% gross margin (Forge's early-stage model).

HOW IT WORKSEvery cost that belongs in COGS — and why founders miss them

For a hardware-plus-SaaS business, COGS has more components than most founders initially include. Each category below belongs in COGS because each one is causally required to deliver the product to the customer. If removing one customer would eliminate or reduce that cost, it belongs in COGS.

1

Hardware cost of goods (BOM + manufacturing + fulfillment)

For Forge, this is the bill of materials for the edge vision sensor: camera module, edge compute chip, enclosure, power supply, cabling, PCB assembly. Add contract manufacturing fees and inbound logistics to the factory, plus outbound shipping to the customer site. In most hardware-first businesses this is the largest single COGS component and the one that creates the structural ceiling on gross margin. Physical goods have non-trivial material cost floors that software does not.

Typical range for a $500–$2,000 BOM sensor: 25–45% of hardware unit revenue, depending on volume and whether the hardware is sold at cost, below cost (land-and-expand), or at a margin.

2

Cloud hosting and inference compute

Every customer whose sensor is active is generating inference workloads — images are being processed, either at the edge or in the cloud, or both. The compute cost per customer is a real, ongoing, per-customer expense that scales with usage. For vision AI specifically, GPU inference costs can be significant: running a batched inference model at 10 fps on an A100-class GPU in the cloud can cost $0.15–$0.40/hour per active line. At 10 lines per customer, that is $13–$35/day per customer, or $5K–$12K/year in compute cost against a $30K ACV SaaS fee. That is 17–40% of the SaaS revenue just in compute — before any other cost. Many founders discover this late when their cloud bill arrives.

3

Third-party API and data fees

If the product integrates with a customer's MES (manufacturing execution system), ERP, or quality management platform via a third-party connector or API, those licensing or call-volume fees belong in COGS. Same for mapping, geolocation, labeling services, or any per-transaction fee from an infrastructure provider. These are often small individually but accumulate: $200/month per customer across 50 customers is $120K/year in cost you may not be modeling.

4

Customer support labor (the fraction attributed to delivery)

Not all support labor belongs in COGS — only the portion that is direct delivery. The heuristic: support activities that are part of keeping the product operational and in the customer's hands belong in COGS; support activities that are purely about customer success, expansion, or renewal belong in S&M. For a hardware product with physical installation and firmware updates, this is typically higher than pure SaaS: expect 1–3 hours of support labor per customer per month in early stage. At a fully-loaded cost of $50–$75/hour, that is $600–$2,700/customer/year in support COGS.

5

Implementation and onboarding labor

If your customer success or engineering team spends time physically installing sensors, configuring the dashboard, training operators, and tuning models for a specific product line, that labor is a direct cost of delivering the product. At early stage, this is often the CEO and founding engineer doing 3–5 days of on-site work per new deployment. At a fully-loaded cost of $800–$1,200/day, a 4-day implementation costs $3,200–$4,800 against a hardware fee of $12K. That is 27–40% of the hardware fee before a single day of ongoing service. Implementation cost is the most commonly underestimated COGS component in enterprise hardware-software companies.

⚠ The "we're a software company" accounting error

Many hardware-software founders describe themselves as "primarily a software company" and compute gross margin accordingly — hardware at cost (pass-through), everything else treated as overhead. This is incorrect accounting and it produces a flattering but misleading number. GAAP does not care about your self-description. If a cost is directly attributable to delivering the product to the customer, it is COGS. The investors and acquirers who will eventually price your business will restate your margin correctly during diligence. It is far better to know your real number now.

COMPARISONGross margin benchmarks by business type

Gross margin benchmarks are not arbitrary preferences — they reflect the structural economics of different delivery mechanisms. Knowing where your model fits tells you which pool of capital is available to you and at what multiple your business will be valued.

Business type Typical gross margin Why the range exists Venture fundability at scale
Pure SaaS (no hardware, no heavy support) 70–85% Near-zero marginal cost per additional seat; hosting scales sub-linearly Strong — software multiples apply
Usage-based / inference-heavy SaaS 60–75% Compute costs are real and per-unit; managed carefully they still allow high margin Good — if margin trajectory is clear
Hardware + recurring SaaS (well-structured) 55–70% Hardware dilutes margin; software portion lifts it; mix and pricing discipline matter Moderate — with clear path to 65%+
Hardware + SaaS (poorly structured) 35–55% Hardware priced near cost, heavy implementation labor, no compute optimization Weak — hardware multiples, not software
Services-wrapped-in-software 25–50% Labor-intensive delivery; COGS is mostly people; does not scale without hiring Poor — services multiples apply (1–2× revenue)
Pure hardware (no recurring software) 25–45% Raw material costs dominate; some pricing power at premium end Very poor — hardware multiples below 2×
🔑 The Series A reckoning

The most common failure mode for hardware-software founders is discovering at Series A that their gross margin — once correctly computed including all COGS — is 40–50% rather than the 65–70% they were modeling. At 40–50%, investors apply hardware multiples (2–4× revenue) rather than software multiples (8–15× revenue). On $3M ARR, that difference is $6M–$12M in enterprise value. The company that believed it was raising a $20M Series A at 10× ARR discovers it is actually raising at a hardware multiple, and the conversation collapses. The structural problem is that gross margin cannot be fixed quickly: hardware costs require volume to improve, compute requires architectural rework, implementation labor requires product changes to reduce. Companies discovered this at Series A are typically 18–24 months from being able to fix it, and many do not survive the wait.

SEE ITGross Margin Breakdown Builder — build Forge's margin waterfall

⚙ Interactive · Gross margin breakdown & expansion simulator

Set Forge's per-customer annual revenue, then dial in each cost component. The waterfall shows your gross margin by layer. The benchmark panel shows where you land versus model-appropriate targets. The scale simulator shows which cost lever produces the biggest margin improvement at 10× current customers.

Revenue inputs
$30,000
$12,000
COGS inputs (per customer, annualized)
$5,500
$4,000
$600
$1,800
$3,200
Gross margin waterfall
Total revenue (yr 1)
—
Total COGS
—
Gross margin %
—
Benchmark comparison — where you land
Pure SaaS target (70–85%)
77.5%
Hw+SaaS well-structured (55–70%)
62.5%
Hw+SaaS poorly structured (35–55%)
45%
Margin expansion at 10× customers — which lever matters most?

At 10× current customers, these are the realistic cost improvements and their margin impact. The lever with the biggest absolute improvement at scale is the one to optimize first.

Hardware COGS +? pts volume lowers BOM ~20–30%
Cloud / inference +? pts committed pricing lowers 30–40%
Implementation labor +? pts product-led onboarding halves cost
Support labor +? pts better tooling reduces hours/customer
Projected margin at 10× ?%
Your gross margin is in the hardware-or-services zone. At this level, venture investors will apply a hardware multiple (2–4× revenue), not a software multiple (8–15×). You need a clear, credible path to 65%+ before Series A.
Your gross margin is in the middle zone — acceptable for a hardware-plus-SaaS business but not yet strong. Investors will want to see the margin expansion trajectory before committing to software multiples. Show the 10× roadmap.
Your gross margin is in the well-structured hardware-plus-SaaS range. Software multiples are attainable — the key is maintaining this trajectory as you scale and not letting implementation or compute costs grow faster than revenue.

A few things to notice. First, notice how sensitive the overall margin is to the implementation labor slider — a single $4K implementation against a $42K ACV drags margin by roughly 9–10 percentage points. That is the most dangerous component because it is the most founder-invisible: at the beginning, the founder and CTO are doing it, it does not show up in cash burn, and it feels like "just helping customers succeed." Second, notice that cloud and inference costs can be 10–15% of revenue even at a reasonable price point — many founders with inference-heavy products discover this when their AWS bill arrives. Third, notice the scale simulator: at 10× customers, the implementation labor improvement from product-led onboarding is often the largest single margin improvement available — larger than hardware volume pricing and larger than compute commitments.

FORGEForge's gross margin model — built honestly

Forge has closed two pilots converted to annual contracts at $42K ACV each: $30K platform SaaS fee plus $12K one-time hardware fee. Here is what the gross margin looks like when computed correctly, using real figures from the stage-3 milestone.

Component Per-customer (year 1) % of total revenue Notes
Total revenue $42,000 100% $30K SaaS + $12K hardware
Hardware COGS (BOM + mfg) −$5,500 13.1% At 2-unit volume; BOM includes edge chip, camera, enclosure
Cloud / inference costs −$4,000 9.5% GPU inference + data storage + egress; ~$333/month per site
Third-party API fees −$600 1.4% MES connector licensing; $50/month
Support labor −$1,800 4.3% ~30 hrs/yr at $60/hr fully loaded; firmware + dashboard queries
Implementation labor −$4,800 11.4% 4 days on-site (CEO + CTO); installation, model tuning, training
Total COGS −$16,700 39.8%
Gross Profit $25,300 60.2% Approximately 60% (60.24%); the curriculum milestone targets ~61%, achievable with a slightly lower implementation cost. This lesson uses the detailed COGS stack which gives 60.2%.
🔑 The milestone finding: ~60% gross margin is real, not fabricated

Forge's stage-3 gross margin on first contracts is approximately 60% using this detailed COGS stack — it includes the implementation labor that most founders leave out. If implementation labor were excluded (the common error), the model would show ~72% and be misleading. The value of computing it correctly is that the 20-customer roadmap becomes actionable: you know exactly which components to reduce, and by how much, to hit 75%.

At 20 customers, three things change materially. Hardware BOM drops as Forge reaches a volume tier with its contract manufacturer — expected reduction of 25–30% on the $5,500 BOM. Cloud costs drop as Forge migrates from on-demand GPU instances to reserved capacity and optimizes the inference pipeline for edge-first execution — expected 35–40% reduction. And implementation labor drops as Forge builds a product-led onboarding workflow with remote configuration tools — expected reduction from 4 days to 1.5 days per site. Together, those three improvements at 20 customers produce a gross margin of approximately 74–76%, putting Forge squarely in the well-structured hardware-plus-SaaS tier and making a software-multiple Series A defensible.

IN PRACTICEThe path from 61% to 75%+ — what actually moves the needle

Margin expansion is not passive — it requires deliberate product and operational decisions made now, when they are cheap to make. The choices that improve gross margin at scale must be built into the architecture of the product and the delivery model today. The founders who wait until Series A to start optimizing gross margin have typically locked in structural costs that take 18–24 months to unwind.

1

Push inference to the edge — reduce cloud cost floor

For Forge specifically, every inference workload that runs on the edge sensor rather than in the cloud is a compute cost that does not exist. At year 1, Forge is running hybrid inference (edge for latency-critical detection, cloud for model updates and batch analytics). Migrating the primary detection pipeline fully to edge by year 2 reduces cloud costs from $4,000/customer/year to an estimated $1,200–$1,500/year — a $2,500–$2,800 reduction per customer per year, which translates to roughly 6–7 margin points at the current revenue level. This requires a more capable edge chip (increasing BOM by $300–$500) but the margin math is clearly positive.

2

Build product-led onboarding — eliminate implementation labor

Four days of on-site implementation labor is not a scalable delivery model. The target at 20 customers is: 1.5 days on-site (physical sensor install only) plus remote configuration, remote model tuning via a labeled dataset upload workflow, and a self-serve operator training module. This requires product investment now (remote configuration dashboard, dataset ingestion pipeline, training content) but reduces implementation cost from $4,800 to approximately $1,800 — a $3,000 reduction per customer, worth roughly 7 margin points. Implementation labor is the highest-ROI margin expansion investment for Forge in the next 12 months.

3

Negotiate volume pricing on BOM as you scale

Hardware volume pricing is non-linear: the jump from 2 units to 20 units typically produces a 20–30% BOM reduction with contract manufacturers. At 20 customers, Forge can negotiate a dedicated production run rather than spot orders, which also reduces lead time variability. A 25% reduction on the $5,500 BOM produces $1,375 in cost reduction per customer — roughly 3 margin points. This does not require product changes; it requires a commercial relationship with the CM and volume commitments that Forge can make once the pipeline is visible.

Now 60% ~6 mo 66% +6 pts (edge inference) ~12 mo 73% +7 pts (onboarding) 20 cust. 76% +3 pts (BOM vol.) 75%+ target zone
Forge's margin expansion roadmap from ~60% to 76%. Each improvement is sequenced by ROI — edge inference first (fastest to implement), product-led onboarding second (most margin impact), BOM volume pricing third (requires customer pipeline to unlock). The compound effect of all three at 20 customers reaches the target zone.

WEAKNESSWhere this bites

✗ Excluding founder labor from COGS until it becomes unaffordable

In the first 2–4 customers, the founders do the implementation. No cash changes hands — it is just their time. So it does not appear in the P&L, and the computed gross margin looks great. Then you hire a customer success engineer at $120K/year to do it. Suddenly the margin drops 15 points and feels like a crisis. It is not a crisis — it is the correct number finally appearing. The right discipline is to cost your own implementation time at market rate from day one, even if no cash actually leaves the company. This forces you to build the product-led onboarding path before it becomes an emergency, rather than after.

✗ Treating the hardware as a pass-through or land-and-expand asset at the cost of the margin model

Some hardware-software founders deliberately price hardware at or below cost to reduce the friction of initial deployment, planning to recover the margin through SaaS fees over time. This is a legitimate strategy — but it requires modeling the gross margin correctly over the full customer lifetime, not just at any single point. If hardware is priced at $6K with a $5,500 BOM (near zero hardware margin) and the SaaS fee is $20K with $8K in compute and support costs (60% software margin), the blended year-1 margin is 44%. That is a hardware-multiple business until year 3 or 4. Founders who run this model need to be clear-eyed about the investor conversation they are actually going to have.

✗ Confusing gross margin with operating margin when reporting to investors

Gross margin and operating margin are different numbers and should never be conflated in investor conversations. Gross margin is revenue minus COGS only. Operating margin is revenue minus COGS minus operating expenses (S&M, R&D, G&A). At early stage, a company might have 65% gross margin and −80% operating margin because it is investing heavily in sales and product. That is normal and healthy. Presenting operating margin when an investor asks for gross margin, or vice versa, signals that you do not have command of your own unit economics — a significant credibility problem at any fundraise stage.

KEY INSIGHTHow this connects

  • 3.1 · Business Model Archetypes — What You Are and What You're Not — the archetype you chose in 3.1 determines your gross margin range and ceiling. A services-wrapped-in-software model has a structural gross margin ceiling of 40–50% regardless of how efficiently you operate; a pure SaaS model has a ceiling of 80%+. Knowing your archetype tells you whether your gross margin problem is fixable or structural.
  • 3.5 · Contract Structure — ARR, NRR, and the Terms That Define Your Business — contract structure directly affects how gross margin is computed and reported. Whether hardware revenue is recognized upfront or amortized, whether implementation is a separate line item or bundled, whether support is included in the platform fee or billed separately — all of these structure the P&L differently and change the visible margin number. The contract terms you write in 3.5 are the same terms that produce the margin you show investors.
  • 3.7 · Services Revenue — The Trap and the Bridge → — services revenue (professional services, training, custom integration) typically carries 20–40% gross margin, far below the SaaS component. How you account for and eventually separate services revenue from software revenue determines whether investors see your true software-business margin or a blended number that hides it. Lesson 3.7 shows how to use services as a bridge without letting it obscure the core margin story.

Gross margin is where all of Stage 3 converges. Your business model archetype (3.1) sets the ceiling. Your pricing (3.2–3.4) determines the revenue numerator. Your contract structure (3.5) defines how costs are allocated and recognized. And your services strategy (3.7) determines whether you are building leverage or labor. Getting gross margin right is not an accounting exercise — it is the test of whether the revenue architecture you have built in Stage 3 actually produces a scalable business.

GLOSSARYKey terms introduced

Gross margin
Revenue minus cost of goods sold, expressed as a percentage of revenue. The fraction of each revenue dollar that remains after paying for direct delivery costs. The structural ceiling on operating leverage, reinvestment capacity, and business valuation at a software multiple.
Cost of goods sold (COGS)
Every cost directly and causally required to deliver the product or service to the customer. For a hardware-plus-SaaS business: hardware BOM, manufacturing, cloud hosting, inference compute, third-party API fees, support labor (delivery fraction), and implementation labor. Excludes sales & marketing, R&D, and G&A.
Blended gross margin
The overall gross margin of a business that combines multiple revenue streams with different margin profiles (e.g., hardware + SaaS + services). Blended margin is lower than any single component's margin. The danger: reporting blended margin when an investor is pricing you on your software component's margin.
Software multiple vs. hardware multiple
Software companies are valued at 8–15× ARR (or higher) because gross margin is 70–80%+ and incremental revenue costs little to deliver. Hardware companies are valued at 1–4× revenue because margin is lower and growth requires more capital. A company straddling both must earn the software multiple by demonstrating that the software component — not the hardware — drives the economics.
Gross profit
Revenue minus COGS in dollar terms (not percentage). $25K gross profit on $42K revenue is 60% gross margin. Gross profit is the cash the business has available before operating expenses. Higher gross profit means more internal funding capacity for growth without additional capital raises.
Margin expansion
The deliberate improvement of gross margin over time through reduced COGS: volume pricing on hardware, compute optimization, product-led delivery replacing labor, API renegotiation. Margin expansion is a strategic priority, not a passive outcome. Investors fund margin expansion trajectories; they do not fund current margin alone.
Operating margin
Revenue minus COGS minus operating expenses (S&M, R&D, G&A). A different, and typically more negative, number than gross margin at early stage. Confusing the two in investor conversations signals weak financial literacy. A company can have strong gross margin (65%+) and deeply negative operating margin (−80%) — both can be correct simultaneously.
Inference compute cost
The per-customer, per-period cloud or edge compute cost of running machine learning model inference. For vision AI products like Forge, this is a non-trivial COGS component — GPU inference at scale is expensive and must be explicitly modeled, optimized, and reported in the gross margin calculation.

ACTIVE RECALLCheck your understanding

Answer in your own words first, then reveal.

1A hardware-plus-SaaS founder tells you their gross margin is 72%. What are the three most important questions you would ask to verify whether that number is computed correctly?›

Ask: (1) Is implementation and onboarding labor included in COGS? This is the most commonly omitted component, especially when founders are doing it themselves. If a 4-day implementation is not costed, the margin can be inflated by 8–12 points. (2) Is cloud and inference compute included in COGS? For an AI product, inference costs are real and ongoing. If they are buried in "engineering costs" or treated as R&D rather than COGS, the margin is overstated. (3) Is support labor allocated between COGS and S&M correctly? Support that is directly required for product delivery belongs in COGS; support that is account management or expansion belongs in S&M. If all support is classified as S&M, margin is inflated. Any "yes" to these questions means the real margin is lower than reported — possibly materially.

2Why does a hardware-software company with 40% gross margin face a specific fundraising problem at Series A that it would not face at Seed?›

At Seed, investors are primarily pricing on the potential of the idea and team — they apply qualitative judgment and early traction signals, and gross margin is often not scrutinized with the same rigor as it will be later. At Series A, institutional investors run a financial model and apply a multiple to the business. A 40% gross margin business cannot support a software multiple (8–15× ARR) because the structural economics do not justify it — the incremental cost of each new customer is too high relative to the revenue they generate. Investors apply a hardware or services multiple (1–4× revenue) instead. On $3M ARR, this is the difference between a $24–$45M valuation and a $3–$12M valuation. The company that built its seed narrative around being a software business suddenly discovers it cannot price as one. And critically, gross margin cannot be fixed quickly — hardware costs require volume, compute requires re-architecture, implementation requires product investment. The fix is 18–24 months away when the Series A conversation is happening today.

3Forge's stage-3 milestone states gross margin of 61% with a path to 75%+ at 20 customers. What are the three specific cost levers that produce this improvement, and in what order should they be prioritized?›

The three levers are: (1) Product-led onboarding replacing implementation labor — worth approximately 7 margin points, this is the highest-ROI improvement because it requires product investment rather than volume, can be started immediately, and also improves the customer experience. (2) Edge inference migration reducing cloud costs — worth approximately 6 margin points, this requires architectural work on the inference pipeline but produces a permanent per-customer cost reduction that scales. (3) Hardware BOM volume pricing at 20 customers — worth approximately 3 margin points, this is passive (it unlocks as the customer pipeline grows) and therefore the lowest priority to actively manage. Priority order: (1) first because it requires product decisions made now; (2) second because it is an architectural change that compounds over time; (3) third because it comes automatically with scale. Together they compound to reach 76–77% at 20 customers.

4What is the difference between gross margin and operating margin, and why does confusing them in an investor conversation signal a credibility problem?›

Gross margin is revenue minus COGS only. Operating margin is revenue minus COGS minus all operating expenses (S&M, R&D, G&A). At early stage, a company with 65% gross margin might have −80% operating margin because it is investing heavily in sales and product. Both numbers are correct — they measure different things. When an investor asks about gross margin, they are asking about the structural economics of delivery: how much does it cost to serve one more customer? When they ask about operating margin, they are asking about the overall profitability of the business including all overhead. Confusing them — especially presenting the gross margin number when asked about profitability, or presenting operating margin when asked about unit economics — signals that the founder does not have rigorous command of their financial model. In a fundraise, this is a significant credibility problem: it implies that the financial projections, the unit economics model, and the investor materials may all be built on shaky foundations.

Next: navigating the services revenue trap — when it helps and when it destroys your margin story →