The real cost of localization at scale: pay-per-use execution vs. seat-based TMS licensing

Every VP of Localization has sat through the same budget conversation. Finance wants a number for next year. You give them one based on last year's content volume, current headcount, and a guess about how many markets you'll add. Then Q3 hits, a product launch triples your string count in three markets, a video division decides dubbing is now a priority, and the number you gave Finance in January is wrong by a wide margin, in either direction. Nobody budgets localization well, not because the teams are bad at forecasting, but because the pricing models underneath the budget were never built to flex.

The actual argument is this: enterprise localization spend is locked into cost structures, seat-based TMS licenses and fixed-scope vendor quotes, that price the infrastructure, not the work. A pay-as-you-go execution model, priced against words and minutes actually processed, changes the shape of the cost curve so it tracks what you're asking the system to do. That distinction matters more as content volume becomes uneven across markets, quarters, and modalities, which for any enterprise operating past a handful of languages, it always eventually does.

Two cost architectures, one budget line

A traditional TMS is licensed like most enterprise software, per seat per year, often with tiered access to features such as connectors and workflow automation. You pay for the seat whether the linguist assigned to it is translating 50,000 words that month or 5,000. The license does not track volume; it tracks how many people are logged in and what tier you signed for. Project-based vendor quotes add a per-project fee scoped to an estimated word count, often with minimums and rush surcharges. The result is a cost structure built around two units that have little to do with actual throughput: seats and projects.

This is not a minor inefficiency. Your fixed costs are set by your peak anticipated need, not your average one. If you provisioned ten TMS seats to handle a product launch surge, you pay for ten seats in the quiet months that follow. If you locked a vendor into a fixed-scope contract sized for last year's roadmap, you either pay overage fees when a market accelerates or absorb an idle contract when it doesn't.

How Ollang prices execution: words and minutes, not seats

Ollang is priced as an execution layer, and its pricing mechanics reflect that. You only pay for the words or minutes processed, with no upfront fees. There is no seat count to negotiate, no annual license to true up, no project minimum to hit before a vendor will quote you. Ollang provides a pay-as-you-go model with no upfront costs, you simply pay per minute or word for the services you use.

Ollang is not a portal your team logs into; it is built to be called. Send translated content through APIs, MCP, automation, and CI/CD pipelines, and localization runs inside the deployment workflow you already use. On the developer side, MCP lets AI agents run workflows, SKILLS provides reusable agent actions, the SDK scans and applies translations, and the API builds end-to-end localization pipelines. Practically, a release pipeline can push a string bundle to Ollang, receive reviewed, localized output, and merge it, without a PM opening a ticket with a vendor or a linguist logging into a TMS.

Billing follows the same logic as invocation: every call that processes words or minutes generates a metered unit, and that is what you are charged for. There is no separate cost for the seat that made the API call, because there is no seat.

For enterprises that need more than metered access, such as SSO, custom connectors, or dedicated support, Ollang provides custom pricing upon request, including advanced integrations and team features. Usage-based pricing and enterprise governance are not mutually exclusive. The metered rate covers throughput, and the custom layer covers the identity, integration, and team-management requirements that come with running localization as infrastructure at enterprise scale.

Modeling cost behavior under uneven volume

Run the two models against a realistic content calendar and the difference stops being theoretical.

Seasonal launch surge. A retail or consumer brand doubles its localized content volume for eight weeks around a major launch, new product pages, campaign assets, updated app strings, across a dozen markets. Under seat-based TMS licensing, you either provision for that peak all year, paying for idle capacity the other 44 weeks, or you scramble to add seats and pay setup and onboarding costs for temporary capacity. Under fixed-scope vendor contracts, the surge blows past the estimated word count in the original quote, triggering change orders, rush fees, or a renegotiation mid-launch, exactly when your team has the least time to manage a vendor conversation. Under a pay-per-word model, the surge appears as a higher metered bill for those eight weeks, calculated against the same rate you use in a quiet month. No renegotiation, no idle capacity to justify to Finance afterward.

Video-heavy quarter. If your organization shifts investment toward training video, product demos, or marketing video for two quarters and then pulls back, dubbing and subtitling work, typically quoted per minute, can spike spend dramatically relative to text-only quarters. If your TMS seat licenses were sized around text translation throughput, they do not help with video work, you end up paying a separate vendor bill for the video work on top of the fixed license you are already carrying. In Ollang's model, video and audio work is priced on the same metered logic as text: pay per minute or word for the services you use, so a quarter that is 80 percent video and a quarter that is 80 percent documents both settle against the same rate card, without a separate contract for each modality.

The quiet quarter. This is where seat-based models perform worst. A market goes quiet, a regional office pauses expansion, a product line sunsets a language, and the seat licenses and fixed retainer keep billing regardless. Usage-based pricing has no equivalent drag: if nothing is processed, nothing is billed.

A framework for true cost-per-market

Most localization budgets compare vendors on rate card alone, cost per word or cost per minute, without accounting for the fixed costs wrapped around that rate. That is the wrong comparison. The framework that reflects total cost of ownership looks like this:

Under seat-based/fixed-contract models:

True cost per market = (annual seat license cost ÷ number of markets actively supported) + (per-project vendor fees, including minimums and rush surcharges) + (idle capacity cost, seats or contract capacity paid for but not used, prorated to the market)

Under usage-based execution:

True cost per market = (words processed for that market × per-word rate) + (minutes processed for that market × per-minute rate) + (any custom enterprise fees for integration and governance, allocated across markets)

The variable that seat-based models systematically hide is the idle capacity term, the seats and contract minimums you pay for regardless of whether that market generated volume this quarter. Run both formulas against your actual per-market volume data for the last four quarters, not your projected volume, and the gap becomes visible. Markets with steady, high volume often look comparable under either model. Markets with bursty, unpredictable, or seasonal volume, which describes most secondary and tertiary markets in a global rollout, are where seat-based and fixed-contract pricing quietly overcharges, because you carry capacity sized for the peak while paying for it every week of the year.

Another variable worth modeling is time-to-scale cost: what does it cost, in dollars and in procurement cycle time, to add a new market or modality under each model? Under a TMS, it is a new seat negotiation. Under a fixed vendor contract, it is a new scope and quote cycle. Under metered execution, it is the same rate card applied to a new content stream, with no new negotiation required to start.

The argument this framework supports

Usage-based pricing is not automatically cheaper in every scenario. A market with very high, stable, predictable volume might be cheaper under a negotiated fixed rate, and that is a legitimate outcome of running the framework above rather than assuming otherwise. The argument is narrower: the cost structure should match the shape of the work. Localization volume is not stable across an enterprise's markets and modalities, and AI-generated content pipelines are making it more uneven as more source content of more types gets created faster than translation teams can plan for.

Seat-based licensing and fixed-scope vendor quotes were built for a world where localization was procured project by project on a planning cycle. Pricing tied to words and minutes processed treats localization the way infrastructure is priced elsewhere in the enterprise stack, compute and storage and API calls, where the bill reflects consumption, not provisioned capacity. The real shift a VP of Localization should model for is a cost structure that stops penalizing you for the fact that content volume was never going to be flat; a better rate alone does not solve that.