What is margin management? Defending gross margin on consumption and AI products
Written by Arnon Shimoni
✓ Expert
Last updated on:
Margin management is the deliberate work of protecting and improving the profit a business keeps from what it sells. Where pricing decides what you charge, margin management decides what you keep after the cost of delivering it.
For two decades most software companies did not need it in any serious sense. Gross margins sat around eighty to ninety percent and marginal cost was close to zero, so revenue growth was the only variable worth managing. AI products removed that assumption, and margin management moved from a finance reporting exercise to an operational discipline.
Field | Detail |
|---|---|
What it is | Deliberately protecting profit per unit, per account and per meter |
Core metric | Gross margin, at the account and meter level rather than company level |
Traditional SaaS | 80-90% gross margin, marginal cost near zero, low urgency |
AI and infrastructure products | Often 40-70%, marginal cost real and variable, high urgency |
Main levers | Price, packaging, meter selection, cost of goods, concession control |
Fails as | A quarterly finance review that reports margin without being able to change it. See margin leakage |
What to actually measure
Company-level gross margin is a reporting number. It is nearly useless for management, because it averages together customers whose economics have nothing in common.
Level | Question it answers | Why it matters |
|---|---|---|
Per meter | Which billable units are profitable | Reveals which meters you can afford to discount |
Per account | Which customers make money | Usually a small number of accounts are unprofitable and invisible |
Per cohort | Whether newer deals are better or worse | Detects pricing decay before it reaches the P&L |
Per workload type | Which use cases carry cost | One heavy use case can dominate infrastructure spend |
Per model or provider | What routing decisions cost | For AI products this often moves faster than pricing does |
The distribution matters more than the average. A company reporting sixty percent blended gross margin may be running at eighty percent on most accounts and losing money on three, and the three are usually the largest and most reference-worthy.
Why AI products changed the discipline
On a classic SaaS product the cost of serving one more user is negligible, so almost any deal at almost any price adds contribution. That single property is why aggressive discounting and unlimited plans were survivable strategies.
On an AI product, every unit of usage carries a real and often volatile cost. Three consequences follow.
Unlimited plans become dangerous. Flat rate pricing on a variable-cost product transfers all volume risk to you, and the heaviest users are the ones who stay.
Discounting has a floor that is not zero. A rate below cost of goods loses money on every unit, and it loses more as the account grows.
Cost moves independently of price. Model pricing, context window growth and routing changes shift your cost basis without any change to your contracts.
The practical implication is that margin has to be visible at the point where commercial decisions get made, not reviewed a quarter later.
The levers, in order of effectiveness
Meter selection. Billing on a unit that correlates with your cost is the single most durable protection. If cost scales with tokens and you bill per seat, no amount of discipline downstream saves you.
Concession control. Enforced expiry on every discount and credit grant. See discount management.
Commitment structures. Minimum commits convert variable revenue into a floor and improve the predictability of the cost you plan against.
Packaging. Separating a high-cost capability into its own meter or tier stops it being subsidised by everything else.
Cost engineering. Caching, routing to cheaper models, batching. Real, and slower to move than the commercial levers.
Price increases. Effective and expensive in trust, which is why grandfathering exists.
Where margin management usually breaks
It becomes a reporting function rather than an operating one. Finance produces a margin analysis, sales never sees it, and the deal desk continues approving concessions on headline discount percentage with no view of the underlying cost.
The fix is unglamorous. Put gross margin, at account and meter level, in front of the people approving deals, at the moment they approve them. A deal desk that can see the modelled margin on a proposed structure makes different decisions from one that can only see the discount depth.
Where Solvimon fits
Solvimon holds meters, rate cards, commitments and concessions in one system, so effective rate and consumption per account are directly observable rather than assembled from exports. Cost inputs can be attributed against the meters that generate them, which makes gross margin per account and per meter a live number.
That is what allows margin to be a deal-time input rather than a quarterly report.
Frequently Asked Questions
What is margin management?
The deliberate practice of protecting and improving the profit retained after the cost of delivering what you sell, managed at the level of meters, accounts and cohorts rather than as a company-wide average.
What is a good gross margin for software?
Traditional SaaS typically runs eighty to ninety percent. AI and infrastructure products commonly run between forty and seventy percent because inference and compute are real marginal costs. Comparing an AI product against SaaS benchmarks leads to bad decisions.
Why is per-account margin more useful than company margin?
Because the average hides the distribution. Most companies have a small number of accounts that are unprofitable, and they tend to be the largest ones with the deepest negotiated rates.
How does meter selection affect margin?
It is the most durable lever available. If your cost scales with tokens or compute and you bill per seat, margin erodes automatically as customers use the product more, and no downstream discipline corrects it.
Is flat rate pricing viable for AI products?
Rarely at scale. Flat pricing on a variable-cost product transfers all volume risk to the vendor, and the heaviest users have the strongest reason to stay, which worsens the mix over time.
Why does margin management fail in practice?
Because it lives in finance reporting rather than in the deal approval path. A deal desk that can only see discount depth cannot protect margin. It needs modelled gross margin on the proposed structure.
Related
Margin leakage. How margin disappears when nobody is managing it.
Marginal cost pricing. Pricing relative to the cost of one more unit.
Discount management. Controlling the concessions that erode margin.
Usage-based pricing. Aligning what you charge with what you spend.
FX volatility and AI margins. The currency exposure most cost models ignore.
Ready for billing v2?
Solvimon is monetization infrastructure for companies that have outgrown billing v1. One system, entire lifecycle, built by the team that did this at Adyen.
