Total Cost of Ownership (TCO)

Executive Summary: What is it?

Total Cost of Ownership (TCO) is the full cost of software over its life, in three stages. Adopting it means moving your data and your people in and standing the tool up. Running it has five parts: the license or subscription, the infrastructure it runs on, the engineering time to operate it, support, and any third-party service it pays per use. Leaving it means getting your data out in a form the next tool can read. Moving to open source shifts most of the running cost from the license to the other parts, and it changes who decides how you leave.

CFO / Business Impact: What does it cost/risk?

The "Free" Illusion. "Free" describes a license model, not a cost model. Self-hosting trades a license fee for other running costs: the infrastructure, and the engineering hours that go into updates, backups, monitoring, patching and incident response. That time is real spend even though no invoice arrives, which is why it is easy to leave out of a budget.

• Example: A tool with no license fee that needs five engineering hours a month costs five hours at your loaded engineering rate, every month, for as long as it runs.
• The Duties Behind the Hours: Patching, backups and incident handling can also be legal duties, depending on your organization and the data the system holds. Once you self-host, their work is yours as the operator; the duties of making and supplying the software stay with its maker, where the law names one. See the operator's compliance checklist.
• The Cost of Leaving: The terms of leaving are set when you sign up, not when you leave. A hosted vendor decides how your data comes out: in which format, whether a bulk export exists, whether all of it comes out, whether export sits in a higher tier, and whether moving the data out is billed. A self-hosted tool leaves that decision with you, but only as far as the project documents a way out. See Data Portability.

Technical Reality: How does it work?

• SaaS Model: The vendor runs the software, so infrastructure and maintenance are bundled into the subscription. The bill is predictable, and it grows with the vendor's pricing unit: seats, active users, usage or data volume. Some vendors bill a second meter beyond the plan, such as API calls or overages.
• Self-Hosted Model: You carry the infrastructure and the engineering labor, on your servers or, for a desktop tool, on your users' computers. That cost grows with the load and data the tool handles and the work it needs, not with a price per seat, and it is front-loaded: setup, hardening and migration come first.
• The License: Many self-hosted tools carry no license fee, but not every one is free for every use. An open core project sells a paid edition with some features. A dual-licensed project sells a license for terms other than its open source license. A source-available license, one that is not open source, can require a paid license above a revenue or size threshold written into it, or restrict certain uses. Some features of a tool may also need paid editions of products it bundles.
• Support: Help costs money wherever it comes from: the vendor's paid support or its own managed service of the software, a paid service provider, or community support paid for in your team's time.
• Third-Party Meters: A self-hosted tool can still pay per use. An email relay or an SMS gateway bills for what the tool sends through it, and an AI feature's model provider bills per unit, such as tokens (the pieces of text a model reads and writes). These grow with usage, not with your infrastructure. Running the models locally moves that cost into GPU infrastructure instead.
• Getting In: Adoption has a one-off cost whichever model you choose: an importer for the old tool, a drop-in replacement that speaks the same protocol, or a manual migration. The vendor or a service provider can also be paid to do that work.
• Getting Out: Leaving costs whatever the export path demands: a database dump or files on disk you already hold, an export tool, an API someone has to script against, or no documented path at all. A tool that stores no data of its own leaves only its configuration behind.
• TCO Formula: Adoption + License + Infrastructure + (Engineering Hours × Loaded Rate) + Support + Third-Party Meters + Exit = Real Cost. Every term is local. Your region, your contracts and your team's rates decide the numbers, so no published price list can fill them in for you.

On a Solution Profile: Where Each Cost Appears

A Solution Profile's Cost of Ownership section states what the self-hosted software costs to adopt, run and leave: models, paths and footprints, never amounts. Each term of the formula has its place on the Solution Profile.

• License: the Pricing Model row, which states whether the self-hosted build carries a license fee and what it withholds or restricts, if anything. A paid edition's features are named in the Executive Summary, or under User Management, in the Profile list, for access controls.
• Adoption: the Adoption row, when getting in has a one-off cost or a named path: an importer, a drop-in replacement or a manual migration. The effort to stand the tool up is graded under Complexity, in the Profile list.
• Infrastructure: the Operations row: the footprint the tool needs and what makes it grow.
• Engineering Hours: the work is named under Maintenance, in the Profile list. The security duties it includes are named under The Compliance Shift, in the Risk Assessment, which links to the operator's compliance checklist.
• Support: the Support row: what the vendor offers so your team carries less, and where else help comes from.
• Third-Party Meters: the Usage Meters row, and for a tool that depends on a model the AI Models row, which states that your model provider bills, and per which unit, or, for local models, points to the GPU footprint in Operations.
• Exit: the Exit row: the export path, with its format or tool.

Where the data sits and which law applies are stated in the Profile list and the Risk Assessment of the same Solution Profile.

💡 Executive FAQ

• Question: When is self-hosting cheaper?
• Answer: When the subscription grows faster than your operating cost. Per-seat pricing rises with every user you add, while self-hosted cost stays roughly flat until the deployment needs more infrastructure or more engineering time, apart from any third-party meter, which grows with use. Where the two lines cross depends on your rates, your scale and the vendor's pricing model, so work it out with your own numbers.
• Question: Can we ignore TCO for small tools?
• Answer: No. Forgotten instances ("zombie servers") keep accumulating infrastructure cost and unpatched security risk long after anyone uses them.
• Question: Why does a Solution Profile on this hub name no prices?
• Answer: Prices differ by region, tier, contract and date, so a single figure is never true for every reader and goes stale quickly. A Solution Profile states how the software charges and what makes the cost grow; your own rates and your providers' quotes supply the numbers for your situation.