Per-seat vs usage-based SaaS pricing: when each model is cheaper for your team

What each model actually means

Per seat vs usage based pricing is the first fork in any SaaS conversation, and most procurement teams pick the wrong side because they pick on instinct rather than math.

Per-seat pricing charges a fixed amount per named user per month. Slack Business+ is $15 per user per month. HubSpot Sales Hub Professional is $100 per user per month. Salesforce Sales Cloud Enterprise is $165 per user per month. The math is predictable: multiply headcount by the unit price and you get the bill. Finance loves this model because it forecasts cleanly into the budget spreadsheet.

Usage-based pricing charges per unit of consumption: per API call at Twilio, per gigabyte ingested at Datadog, per monthly tracked user at Mixpanel, per token at OpenAI. The unit is small, the volume is large, and the bill moves every month with traffic. Finance struggles with this shape because variance is the enemy of a clean forecast.

Hybrid pricing combines both axes: you pay per seat for access and per unit for consumption inside that seat. The hybrid model compounds when you misread it.

The crossover math, by category

Five categories, five named vendors, and a snapshot of where each shape breaks.

The crossover point is the volume at which one model becomes cheaper than the other. For communication tools the crossover sits high, because per-seat is cheap until headcount balloons. For observability and AI the crossover sits low, because usage scales faster than headcount. Run the math on the actual unit volume from your last 90 days before you sign anything, because the vendor sales rep will quote whichever model produces the larger contract value for them.

How to forecast a usage-based bill

Forecasting a usage-based bill is the single most useful procurement skill on the modern SaaS stack, because usage-based contracts are where finance loses control of the budget.

Pull the last 90 days of usage from the vendor billing dashboard. Every usage-based vendor exposes a daily or hourly metric. Datadog has a Usage and Cost page that exports CSV. Stripe Billing exposes per-meter line items via the API, documented in the Stripe Billing docs. OpenAI exposes per-model token counts. Pull the raw daily numbers, not the summary.

Calculate month-over-month growth from that data. The typical SaaS account grows usage 3 to 10 percent month over month, per the OpenView Partners annual SaaS benchmarks. Apply the trend forward: if the last three months ran 8, 9, and 11 percent, assume 10 percent for the projection rather than the average.

Project 12 months at the trend rate by compounding the monthly growth across the full term. A $10,000 monthly bill growing 8 percent compounds to roughly $25,000 by month 12 and a $210,000 annual total rather than the $120,000 a naive linear projection suggests. Add a 20 percent buffer for incident spikes, because retry storms and viral launches routinely produce 5x to 10x normal volume for days at a time.

Negotiate a committed-use discount at 70 percent of the projected total. Most usage-based vendors offer 15 to 30 percent off list for an annual commit, and the commit should sit below the projection so any overage bills at the discounted rate rather than at retail.

Walk a Datadog example through. Last 90 days averaged $12,000 a month on 200 hosts and 400 million log events. Growth was 7 percent month over month. Project 12 months at 7 percent and the total compounds to roughly $216,000. Commit at 70 percent of $216,000, which is $151,000. The most expensive Datadog bill I ever saw was a $437,000 single-month invoice triggered by a misconfigured custom metrics agent that emitted 14 billion data points in 72 hours, with no committed-use discount in place, so every metric billed at retail. A $151,000 commit would have capped the damage at roughly $90,000.

When to switch from per-seat to usage-based

Three signals say a per-seat contract is the wrong shape and a usage-based alternative will be cheaper. Apply them at renewal, because most vendors refuse mid-term changes.

The first signal is 30 percent or more inactive seats you cannot remove. Sales prospecting tools are the classic case. A team licenses 200 seats of Outreach, and 60 of those seats log in less than once a week because the named user is a prospect contact, a part-time SDR, or a churned headcount the rep has not surrendered. At $100 per seat per month, that is $72,000 a year of dead air. A usage-based alternative that bills per sent email collapses the dead seats into zero billed units.

The second signal is seasonal spikes. Tax software, retail commerce, and recruiting tools all see 3x to 10x usage in their peak quarter and near-zero usage in the trough. Per-seat pricing flat-lines the bill and overcharges in the off-season. Usage-based pricing flexes with the cycle and saves 40 to 60 percent of total annual spend on a seasonal workload.

The third signal is a growing team that hates per-seat lockstep. Engineering teams expanding from 50 to 150 in 12 months produce a per-seat budget conversation every quarter and a true-up invoice the rep refuses to discount. Usage-based licensing for the developer tools in that stack replaces the headcount conversation with a consumption conversation that finance can model.

When to switch from usage-based to per-seat

The reverse switch happens less often, but it happens. Three signals say the usage-based bill has matured into something a per-seat or flat-rate commit will beat.

The first signal is usage stabilizes. A product that grew 15 percent month over month for two years flattens to 2 percent in year three. The volatility that justified the usage-based contract is gone, and a flat-rate commit at the stabilized volume locks in the rate.

The second signal is finance wants predictability. A usage-based bill that swings 30 percent month to month is a board-meeting topic that no operator wants. A flat-rate contract with a per-seat or per-tier cap converts the bill into a single line item on the forecast.

The third signal is the vendor offers a flat-rate commit that beats the variable rate. Tomasz Tunguz of Theory Ventures has noted that vendors signing large committed-use deals routinely discount 25 to 40 percent below variable list price to lock in the revenue, and the discount usually shows up at renewal if the rep believes a competitor is in the room. Run the spreadsheet on both shapes before you re-sign.

The hybrid trap: both per-seat AND per-usage

Hybrid pricing is where the modern stack hides its sharpest cost surprises. Intercom Fin charges $0.99 per AI resolution on top of the per-seat Intercom subscription that ranges from $39 to $139 per seat per month. Salesforce Einstein layers per-conversation AI fees on top of the $165 per seat per month Sales Cloud Enterprise contract. ChatGPT Team is $30 per user per month and meters certain enterprise connectors separately. GitHub Copilot Business is $19 per user per month and prices premium model requests in a separate per-request tier.

Hybrid contracts compound the fastest because both axes scale at the same time. A 100-seat ChatGPT Team contract is $36,000 a year on seats alone, and per-request connector fees at typical enterprise volume push the bill past $60,000. A 50-seat Intercom Fin contract is $69,000 on seats and another $54,000 on AI resolutions at 5,000 resolutions a month. The vendor sales motion always quotes one axis first, lets you anchor to that number, and quotes the second axis after you have mentally signed.

The fix is to model both axes in one spreadsheet before the kickoff call. Pull projected seat count from headcount planning and projected unit volume from product usage data. Multiply each axis through to a 12-month total and add the totals. Compare against a per-seat-only alternative and a usage-only alternative in the same category. If the hybrid number sits within 10 percent of either pure model, take the pure model and remove an entire dimension of variance from the budget.