Unit Cost Architecture

Your tech costs are on the P&L.
What they cost per unit isn't.

Q: What's a good gross margin for a SaaS company?

Most pure software businesses target 75% or higher. Anything embedding heavy AI, LLM, or infrastructure cost typically runs lower by design — see the benchmark table for ranges by business model. These are general starting points, not a verdict on your specific business.

Q: How do I price a feature built on Claude, GPT, or another AI model?

Start by isolating the per-request token cost for that feature specifically, then decide whether to price it as a flat add-on (simpler, but risks losing money on heavy users) or usage-based (protects margin, adds billing complexity).

Q: Should Customer Success headcount be in COGS or operating expense?

Generally, headcount that keeps the product running for existing customers — support, DevOps, SRE — belongs in COGS. Headcount focused on expansion, upsell, or new sales sits in operating expense.

Q: What is capitalized software amortization, and why does it matter?

Under standards like ASC 350-40, certain development costs get capitalized rather than expensed immediately, then amortized into COGS once the product is live. It's a non-cash line that's easy to miss.

Q: Is the Diagnostic's benchmark data based on real client data?

No — the ranges shown are general practitioner starting points, stated plainly rather than dressed up as proprietary research. A full teardown replaces these with numbers specific to your actual contracts, usage, and headcount.

Q: What happens after I submit my email in the Diagnostic?

You'll get your full sensitivity report and the benchmark guide. If your numbers suggest a real margin gap, we'll follow up about whether a full teardown makes sense — there's no obligation either way.

Margineer deconstructs your consolidated COGS lines bottoms-up—isolating cloud infrastructure, LLM token consumption, third-party APIs, and technical delivery payroll. We transform raw operations data into clear, defensible gross margins so you can optimize pricing ahead of rising costs.

Fig. 1 — Where You Stand

Gross Margin Benchmarks by Business Model

Business Model	Healthy	Watch	At Risk
Pure SaaS / Software	≥75%	60–75%	<60%
AI / LLM-Embedded Product	≥60%	45–60%	<45%
Infrastructure or Compute-Heavy Product	≥65%	50–65%	<50%
Services-Heavy / High-Touch Delivery	≥55%	40–55%	<40%

General starting points, not a verdict — find your row, then run your own numbers.

Built by operators who've run cloud cost and FinOps functions inside real technology organizations — not analysts who've only studied the reports.

Field Record

An eight-figure cost target, met — roughly 70% of it out of non-labor.

...

8-figure

Savings target set for engineering — found and delivered in-year

~70%

Came out of software, cloud, and hosting before headcount

~97%

Faster month-end close — a full day down to 10–15 minutes

The Problem

One line on the P&L. Six things blended into it.

Your COGS is already consolidated — that's what accounting is for. It's just too blended to answer the question that actually matters: does the price still cover what it costs to deliver?

Cloud & Hosting

AWS, GCP, Azure, and owned infrastructure that scales with usage — one part of a bigger, blended COGS number.

AI & LLM Consumption

Claude, OpenAI, and other model costs billed by the token — often the fastest-growing line nobody's isolated yet.

Embedded Software & APIs

Auth0, SendGrid, Twilio, and other vendor APIs baked into what it actually costs to deliver the product.

Merchant Processing Fees

Stripe, Adyen, and gateway fees that scale as a % of gross revenue — a different lever than usage-based software.

Production & Delivery Headcount

The fully loaded cost of DevOps, SREs, and technical support keeping the product live for customers.

Capitalized Product Amortization

Non-cash amortization of capitalized development costs — easy to omit until an audit finds it.

How It Works

From one blended COGS line to a price per unit.

Enter revenue, units, and your six COGS lines.
Cloud, AI usage, embedded software, merchant fees, delivery headcount, and capitalized amortization — broken out, not blended.

See your real gross margin and cost per unit.
The number a single COGS line on the P&L never shows you directly.

Get the price needed to hit your target margin — as costs rise.
Including what happens to the required price at 10% and 25% higher costs.

78% Gross Margin

This is the number a blended COGS line never shows you directly — and the one your pricing should be built around.

Not ready to run the full Diagnostic?

Get the one-page benchmark cheat sheet — the margin ranges by business model, sent straight to your inbox. No calculator required.

One email. Sent once. No spam.

"If you don't know your unit cost, you don't actually know your price — you're just guessing and hoping the gap doesn't close on you."

The Diagnostic

Gross Margin & Unit Cost

Enter annual figures to see your gross margin and cost per unit instantly. Unlock the full sensitivity report and pricing guidance with your email.

Executive Snapshot

Five inputs. A 60-second read on what your cloud bill is really doing to your margin, efficiency, and (if it matters to you right now) your valuation.

Quick Read

Total Monthly Revenue

Total Monthly Cloud Bill (AWS/GCP/Azure)

Active Customers / Tenants

Estimated Untagged / Unallocated Spend

Actively Managing Reserved Instances / Savings Plans?

The Four Signals

Enter your monthly revenue and cloud bill above to populate these.

Cloud Waste Estimate

—

Idle and over-provisioned resources, estimated from tagging maturity.

Cloud Cost % of Revenue

—

Healthy SaaS range is typically 15–30% of revenue.

Unallocated / Untraceable Spend

—

Industry target is under 10% — above that, nobody owns the number.

Cloud Cost Per Customer

—

What it costs, in raw infrastructure alone, to serve one account.

These are illustrative estimates built from industry-standard FinOps heuristics, not a measured audit of your actual tagging or usage data — a real teardown replaces these with your real numbers.

Want the granular, bottoms-up version?
Continue to the Full Diagnostic ↓

Gross Margin & Unit Cost Calculator

Break your COGS down bottoms-up to see what it costs to serve one unit.

FY2026 · No. 002

Load a Starting Profile (Optional)

Illustrative profiles — not real client data. Your architecture doesn't look like our templates?

Annual Revenue

Units Sold / Active Customers (Annual)

Target Gross Margin

Which Best Describes Your Business?

Cost of Goods Sold (COGS) — bottoms-up

Total COGS

—

Enter your COGS lines above to see the breakdown.

These six lines are what's already blended into your COGS on the P&L. Breaking them out bottoms-up — AI usage, merchant fees, and capitalized amortization included — is what makes the margin and unit cost below meaningful.

—Margin

Cost Per Unit

—

Fill in revenue, units, and all six COGS lines to see your gross margin and cost per unit.

If COGS Rises	Cost / Unit	Price Needed
Current	—	—
+10%	—	—
+25%	—	—

The Playbook

A working reference for
pricing ahead of rising costs

Not a framework to memorize — a set of tools to reach for when you're building a budget, setting a price, or trying to figure out why margin slipped.

The COGS Breakdown Margin Benchmarks The Pricing Math Where Margin Quietly Erodes Glossary Review Cadence FAQ

01 — Framework

The COGS breakdown

This is the detail sitting inside the single COGS line on your income statement. Naming each part separately is what makes gross margin actionable instead of just reportable.

Cloud & Hosting

What scales with usage

AWS, GCP, Azure, and owned or colocated infrastructure billed by consumption. The most usage-sensitive line — it moves with every customer you add, whether or not your price does.

AI & LLM Consumption

What scales with every request

Claude, OpenAI, and other model or inference costs — priced by token or request, moving with usage in a way flat subscriptions rarely account for. Increasingly one of the fastest-growing lines in the stack, and one of the least isolated.

Embedded Software & APIs

What's built into the product

Auth0, SendGrid, Twilio, and other vendor APIs embedded in product delivery — distinct from internal tools, which are overhead, not COGS.

Merchant Processing Fees

What scales with revenue, not usage

Stripe, Adyen, PayPal, and other gateway fees — typically a fixed percentage of gross revenue plus a per-transaction charge. Blended into "software," it hides a lever that moves in lockstep with growth rather than efficiency.

Production & Delivery Headcount

What people cost to keep it live

The fully loaded cost of DevOps, SREs, and Tier-1/2 technical support directly required to keep the product running — not account managers or R&D headcount building what's next.

Capitalized Product Amortization

What GAAP requires you to see

Under standards like ASC 350-40, development costs capitalized during the application development stage get amortized into COGS once the product goes live — a non-cash line non-finance founders routinely miss.

Restricted — Full Playbook

Request the Full Playbook

The margin benchmark tables, the pricing math, the nine places margin quietly erodes, the glossary, the review cadence, and the FAQ are held back. Tell us who you are and we'll open the full reference and send it to you directly.

Unlocked — and on its way.

The full playbook is now open below, and we've emailed you the reference material.

Don't see it in a few minutes? Check your spam or promotions folder.

02 — Reference

Gross margin benchmarks

General reference ranges by business model — not a verdict, and not a substitute for your own peer data. How you classify a cost between COGS and operating expense changes this picture as much as the cost itself does.

Business Model	Healthy	Watch	At Risk
Pure SaaS / Software	≥75%	60–75%	<60%
AI / LLM-Embedded Product	≥60%	45–60%	<45%
Infrastructure or Compute-Heavy Product	≥65%	50–65%	<50%
Services-Heavy / High-Touch Delivery	≥55%	40–55%	<40%

A product that embeds heavy AI, LLM, or third-party data costs will run lower than pure software by design — that's a model difference, not automatically a problem. Run your own numbers in the calculator.

03 — Method

The pricing math, plainly

Three numbers, one formula. This is the entire logic behind the calculator.

Step 1

Find cost per unit

Total COGS ÷ units sold or active customers. This is what it actually costs to deliver one unit of your product, fully loaded.

Step 2

Set a target margin

The gross margin you need to fund R&D, sales, and overhead below the line — informed by the benchmarks above, not copied from them.

Step 3

Solve for price

Price = Cost Per Unit ÷ (1 − Target Margin). As cost per unit rises, the price required to hold the same margin rises with it — automatically, not by guesswork.

04 — Patterns

Where margin quietly erodes

None of these look like a bad decision in the moment. They look like reasonable calls that nobody re-priced against.

Pricing set once, never revisited

The price was right when COGS was lower. Nobody rechecked it as third-party or cloud costs rose underneath it.

Usage-based costs sold under a flat price

COGS that scales with usage — API calls, storage, compute — sold under a flat subscription. Margin erodes with every heavy user, invisibly.

Discounting without checking unit cost

A deal gets discounted below what it costs to serve, because nobody had the unit cost number in the room when the discount was approved.

A healthy blended margin hiding a losing segment

The overall number looks fine. One product line or customer segment underneath it is being subsidized by the rest.

Support cost that scales with complexity, not price

Enterprise accounts with heavy onboarding and support cost priced the same as self-serve customers who need almost none.

Infrastructure misclassified as overhead

Costs that directly support delivery get booked as general operating expense instead of COGS — margin looks better than it is, and nobody catches it until pricing breaks.

AI costs treated as a fixed line

Claude, OpenAI, or other model usage is billed by the token, but sold to customers as a flat feature or flat subscription tier. Heavier users quietly cost more to serve, with nothing in the price reflecting it.

Merchant fees blended into "software"

Payment processing scales as a percentage of gross revenue, not usage — a fundamentally different lever than a software subscription. Blending it in hides how much growth itself is costing in transaction fees.

Unamortized capitalized software

Development costs capitalized under standards like ASC 350-40 and never amortized into COGS make gross margin look better than it is — until an audit or acquirer's diligence team finds it.

05 — Glossary

Terms worth knowing

Vocabulary that shows up in margin and pricing conversations, defined plainly.

Gross Margin: Revenue minus COGS, divided by revenue. The percentage of every dollar that's left after covering the direct cost of delivering the product — before R&D, sales, and overhead.
Inference Cost: The per-request cost of running a prompt through an AI model like Claude or GPT, typically billed by input and output tokens. Behaves more like a cloud cost than a flat software subscription — it scales directly with usage.
Merchant Discount Rate (MDR): The percentage plus fixed fee a payment processor like Stripe or Adyen charges per transaction — typically around 2.9% + $0.30. Scales with gross revenue, not usage, which makes it a different lever than the rest of COGS.
Capitalized Software (ASC 350-40): An accounting standard requiring certain internal-use software development costs to be capitalized rather than expensed immediately, then amortized into COGS once the product goes live. Easy for non-finance founders to miss until an audit or acquirer's diligence process surfaces it.
COGS (Cost of Goods Sold): The direct costs of delivering your product — hosting, third-party software embedded in the product, infrastructure, and delivery headcount. Distinct from R&D or general overhead, which sit below the gross margin line.
Unit Economics: What it costs to deliver one unit of your product or service, set against what that unit brings in — the building block for understanding whether growth makes the business more or less efficient.
Cost-Plus Pricing: Setting price by starting from unit cost and adding a target margin on top — the logic behind the calculator's "price needed" figure.
Contribution Margin: Revenue minus variable costs for a specific product, deal, or customer — useful for spotting the segment hiding underneath a healthy blended gross margin.
Fully Loaded Cost (Headcount): An employee's total cost to the business — salary plus benefits, payroll taxes, equity, and a share of overhead — not just the number on their offer letter.
FinOps: The discipline of bringing finance, engineering, and business teams together to make informed, collaborative decisions about cloud spend — treating cost as something engineering actively manages, not something finance discovers after the fact.
Shelfware: Software that's paid for but sitting unused — licenses bought for a plan that changed, seats provisioned for people who've since left.

06 — Practice

A simple review cadence

Margin doesn't erode because nobody cares. It erodes because nobody's checking the price against the cost on a rhythm.

Monthly

Review cloud and usage-based COGS against forecast. Flag any category that moved without a known cause.

Quarterly

Recalculate cost per unit and check it against current pricing — not just against last quarter's number.

Annually

Run the full gross margin breakdown, reset target margin, and reprice ahead of the cost increases you can already see coming.

07 — FAQ

Common questions

The things people usually ask before they run the numbers.

What's a good gross margin for a SaaS company?

Most pure software businesses target 75% or higher. Anything embedding heavy AI, LLM, or infrastructure cost typically runs lower by design — see the benchmark table above for ranges by business model. These are general starting points, not a verdict on your specific business.

How do I price a feature built on Claude, GPT, or another AI model?

Start by isolating the per-request token cost for that feature specifically, then decide whether to price it as a flat add-on (simpler, but risks losing money on heavy users) or usage-based (protects margin, adds billing complexity). The Diagnostic's sensitivity table shows what happens to required price as that cost rises.

Should Customer Success headcount be in COGS or operating expense?

Generally, headcount that keeps the product running for existing customers — support, DevOps, SRE — belongs in COGS. Headcount focused on expansion, upsell, or new sales sits in operating expense. This is exactly the kind of classification question a full teardown resolves line by line.

What is capitalized software amortization, and why does it matter?

Under standards like ASC 350-40, certain development costs get capitalized rather than expensed immediately, then amortized into COGS once the product is live. It's a non-cash line that's easy to miss — and a common surprise during an audit or an acquirer's due diligence.

Is the Diagnostic's benchmark data based on real client data?

No — the ranges shown are general practitioner starting points, stated plainly rather than dressed up as proprietary research. A full teardown replaces these general ranges with numbers specific to your actual contracts, usage, and headcount.

What happens after I submit my email in the Diagnostic?

You'll get your full sensitivity report and the benchmark guide. If your numbers suggest a real margin gap, we'll follow up about whether a full teardown makes sense — there's no obligation either way.

The Approach

Built by people who've
priced a product against
rising cost, not just studied it

Margineer is a free tool from Marginwell Advisory — a fractional VP of Finance, Technology & FinOps practice. The calculator is the first look; the teardown is where the real work happens.

Why this exists

Most cost tools stop at spend

Cloud cost tools optimize cloud. Procurement tools optimize contracts. AI usage — Claude, OpenAI, and the rest — usually gets buried inside whichever of those it was easiest to bill to. Almost nothing connects the resulting COGS back to gross margin and price.

What the calculator is

A first read, not a full audit

Six COGS lines and a unit count won't catch everything — a mispriced enterprise contract, an unamortized software asset, or usage-based cost hiding inside a flat category. What it will do is tell you, honestly, whether your price still covers your cost. That's the point.

What the benchmarks are

Starting points, stated plainly

The margin ranges behind every stamp are general practitioner guidance, not a proprietary dataset claiming false precision. They're visible in the Playbook because a benchmark you can't see isn't one you can trust — or challenge.

What happens after

A teardown, line by line

A "Watch" or "At Risk" margin means cost is outrunning price somewhere specific — not where, or by how much. That's the work a full engagement does: COGS line by line, against your actual contracts and pricing.

Marginwell Advisory

Fractional VP of Finance, Technology & FinOps

Rising COGS changes how you have to price. That's a specific problem, and a generalist fractional CFO usually isn't the right fit for it. This is a small, capped practice that works only on margin and cost questions, for technology companies from growth-stage SaaS to established enterprise.

Case Study — Technology Cost Governance

A nine-figure technology
expense base nobody
could attribute

Governance built across public, private, and PE-backed technology organizations. An eight-figure engineering savings mandate was set for the year. It was met in full — and roughly seven of every ten dollars came out of software, cloud, and hosting before compensation was ever sized. Client details, geographies, and ownership structures are withheld; figures and mechanics are representative.

Case Study // Engagement Profile

Ref. MGN-CS-001

01 / Relationship

Engagement

Multi-year embedded engagement inside the engineering and finance organizations. Not an outside quarterly review — hands-on governance with the teams that own the spend.

Embedded Multi-year

02 / Surface Area

Scope

Nine-figure annual technology expense base
Eight-figure hyperscaler cloud spend, data center leases, and capital portfolio
Multiple engineering leaders, several hundred technical FTEs

03 / Objective

Mandate

Eight-figure engineering savings mandate, single fiscal year, with no single owner and no controls in place.

Primary Objective

"Deliver the eight-figure mandate in-year while building the attribution controls that make the number repeatable."

Savings Mandate8-figure

Restricted — Full Record

Request the Full Case Study

The situation, the diagnosis, the eight controls that were built, the attribution model, and the final result are held back. Tell us who you are and we'll send the complete framework and follow up directly.

Unlocked — and on its way.

The full record is now open below, and we've emailed you the framework. Someone from our team will follow up directly.

Don't see it in a few minutes? Check your spam or promotions folder.

The Situation

The spend wasn't overspent. It was unattributed.

Nobody was certain what the technology expenses were, what made up the software line, or why it kept growing. Headcount was the only line legible enough to cut against — so headcount was the only lever anyone reached for.

Spend resolved to the reseller

Software bought through a cloud marketplace booked with the marketplace as the vendor. There was no source vendor field, so no contract-level detail existed beneath the software line at all.

Purchase orders with no term

Contracts entered the system with no start or end date recorded — which meant auto-renewals nobody was tracking, and cancelled software still being approved for payment.

Purchase orders with no description

No record of what the software did, what the hosting was for, or what service was actually being bought.

Assets not mapped to location

Data center equipment across a global estate couldn't be tied to a physical site, so shutdown decisions couldn't be priced against leases or capital spend.

Consumption booked to the host

All AI cost sat in engineering regardless of which department consumed it. Engineering looked expensive because engineering was the ledger of record.

Close ran on manual review

Month-end took a full day of line-by-line vendor review, because actuals never landed where the plan said they would.

The Diagnosis

The company believed it had a cost problem. It had a governance and attribution problem.

You cannot cancel, renegotiate, accrue, or shut down against a contract you cannot see, a term that was never recorded, or an asset you cannot locate. The work started with a line-by-line P&L review, software walkthroughs with each engineering leader individually, a cloud environment review, and a look at what was actually entering the system at purchase-order stage.

Exposure found	Share of mandate
Software — canceled contracts & approval-amount errors	Largest single component
Cloud — identified against annual hyperscaler spend	Second component
Post-acquisition duplicate vendors and duplicate roles	Third component

Composition as identified during the engagement; client, vendor names, and absolute figures withheld.

The Execution

Fix the system of record before touching the budget.

None of these controls existed beforehand. Each one turned an illegible line into something that could be governed.

Control 01

System of record migration

Escalated until the organization adopted an enterprise financial system, in-year — the precondition for every other control, since attribution cannot exist inside a ledger that cannot hold it.

Control 02

Source vendor field

Specified and implemented in the new procurement and ERP system, so spend resolves to the actual software vendor rather than the marketplace reseller. This is what made contract-level detail visible for the first time — and what exposed the over-accruals and the software no longer needed.

Control 03

Approval gate at position one

First approval on every invoice and purchase order — ahead of the leader, ahead of accounting. Source vendor correct, agreement terms present, record clean. Nothing moves downstream until it is.

Control 04

Purchase order data standards

Start and end dates mandatory on every PO. Service description mandatory — what the software is, what the hosting is, what the service provides.

Control 05

Asset location mapping

Every piece of data center equipment mapped to its physical site across a global estate, so shutdown savings became priceable against leases and capital spend.

Control 06

Bottoms-up budget files

One per engineering leader. Every vendor, every contract term, expected renewal uplift sourced from procurement, geographic raise assumptions modeled, and cloud growth projected forward — so the board sees the trajectory in advance instead of discovering it at variance.

Control 07

Product-level margin reporting

Built so leadership and the board could hold the margin conversation on product economics rather than a blended cost line.

Control 08

Consumption chargeback

AI and cloud cost moved out of engineering and onto the department that actually consumes it — each team budgets its own seats, owns its own overages, and sees its own consumption.

Attribution

Cost sits with whoever can control it.

A single month's AI consumption came in at roughly 175% of forecast once access opened up. Rather than absorb the variance, it was escalated with invoice evidence — and AI cost was moved out of engineering and onto the consuming department. Each team budgets its own seats, owns its own overages, and sees its own consumption.

It doesn't save a dollar on day one, and it is still correct: it puts the number in front of the only person who can change it.

From → To

From	To
Software booked to a reseller	Resolved to source vendor
No source vendor field	Built and enforced at approval
POs with no term	Start and end dates mandatory
POs with no description	Service description mandatory
Assets unmapped	Every asset tied to a physical site
AI cost pooled in engineering	Charged to the consuming department

The Result

An eight-figure mandate, met in full. Roughly seven of every ten dollars out of non-compensation.

Most cost programs invert that ratio, because non-labor is illegible and headcount is not. The controls are what made that mix possible.

...

Target

8-figure met in-year

Delivered against a named mandate, not an open-ended efficiency exercise.

Mix

~70% non-labor

Software, cloud, and hosting exhausted first; compensation sized against what remained.

1 day → 10–15 min

Roughly a 97% reduction on 200 lines. It's faster because the data is right upstream, not because less is reviewed.

"You cannot govern spend you cannot attribute — and nobody is accountable for a number sitting on someone else's P&L."

Get The Insights

The margin research,
before anyone else sees it.

New benchmark data, pricing frameworks, and margin research as they're published — straight to your inbox. No sales calls, no screening, just the insights.

Stay in the Loop

We're building out the full Margineer research library. Drop your email to get new benchmarks, frameworks, and playbook updates as they go live.

You're on the list.

We'll send new margin research and benchmark data as it's published — no sales calls attached.

Don't see the confirmation email in a few minutes? Check your spam or promotions folder.

Get The Full Framework

Request the Blueprint

Tell us about your business and we'll send you the complete framework with all the phases, tactics, and implementation details.

The Blueprint is on its way.

Check your inbox for the complete framework — phases, tactics, and implementation details included.

Don't see it in a few minutes? Check your spam or promotions folder.

Your tech costs are on the P&L.What they cost per unit isn't.

Gross Margin Benchmarks by Business Model

An eight-figure cost target, met — roughly 70% of it out of non-labor.

One line on the P&L. Six things blended into it.

Cloud & Hosting

AI & LLM Consumption

Embedded Software & APIs

Merchant Processing Fees

Production & Delivery Headcount

Capitalized Product Amortization

From one blended COGS line to a price per unit.

Not ready to run the full Diagnostic?

Gross Margin & Unit Cost

Executive Snapshot

Gross Margin & Unit Cost Calculator

A working reference forpricing ahead of rising costs

The COGS breakdown

What scales with usage

What scales with every request

What's built into the product

What scales with revenue, not usage

What people cost to keep it live

What GAAP requires you to see

Request the Full Playbook

Unlocked — and on its way.

Gross margin benchmarks

The pricing math, plainly

Find cost per unit

Set a target margin

Solve for price

Where margin quietly erodes

Pricing set once, never revisited

Usage-based costs sold under a flat price

Discounting without checking unit cost

A healthy blended margin hiding a losing segment

Support cost that scales with complexity, not price

Infrastructure misclassified as overhead

AI costs treated as a fixed line

Merchant fees blended into "software"

Unamortized capitalized software

Terms worth knowing

A simple review cadence

Monthly

Quarterly

Annually

Common questions

Built by people who'vepriced a product againstrising cost, not just studied it

Most cost tools stop at spend

A first read, not a full audit

Starting points, stated plainly

A teardown, line by line

Fractional VP of Finance, Technology & FinOps

A nine-figure technologyexpense base nobodycould attribute

Engagement

Scope

Mandate

Request the Full Case Study

Unlocked — and on its way.

The spend wasn't overspent. It was unattributed.

Spend resolved to the reseller

Purchase orders with no term

Purchase orders with no description

Assets not mapped to location

Consumption booked to the host

Close ran on manual review

The company believed it had a cost problem. It had a governance and attribution problem.

Fix the system of record before touching the budget.

System of record migration

Source vendor field

Approval gate at position one

Purchase order data standards

Asset location mapping

Bottoms-up budget files

Product-level margin reporting

Consumption chargeback

Cost sits with whoever can control it.

An eight-figure mandate, met in full. Roughly seven of every ten dollars out of non-compensation.

8-figure met in-year

~70% non-labor

1 day → 10–15 min

Your tech costs are on the P&L.
What they cost per unit isn't.

A working reference for
pricing ahead of rising costs

Built by people who've
priced a product against
rising cost, not just studied it

A nine-figure technology
expense base nobody
could attribute

The margin research,
before anyone else sees it.