How to sell a translation or localization company, and what to do with its records

How do you sell a translation company, and where do its records fit?

You sell a translation or localization company by preparing three things buyers price: recurring client revenue, a repeatable delivery process and clean ownership of the assets. Alongside the sale, the firm's own operational records, such as project management history and vendor performance data, can be licensed to AI developers, but client content cannot unless the client consents.

That last distinction decides everything in this sector. A language services provider (LSP) holds two very different piles of material. One pile is the firm's own work records. The other is its clients' words, stored in translation memories, glossaries and files. Only the first pile, plus any client-consented assets, can be licensed.

This guide is for owners and the advisors who sit beside them. It covers what buyers look at, which LSP records may qualify for a data license, and how to sequence a licensing conversation around a sale.

What do buyers look at when they value an LSP?

Buyers of language services firms look for revenue that repeats, margins that survive vendor rate pressure and a delivery process that does not depend on the founder. Machine translation and AI post-editing affect how buyers read a per-word pricing model, so expect questions about it.

Plan on a quality-of-earnings review that separates machine-assisted volume from fully human volume. Owners who can show the trend with data sell with fewer surprises.

Which LSP records could be licensed for AI training?

The records most likely to qualify are the ones the firm created about its own work, not the text it was paid to translate. A company with 50+ full-time employees at peak (contractors excluded) and several years of history often holds a lot of this.

• Project management history: quotes, job tickets, scheduling, deadlines, handoffs and delivery confirmations in the TMS or PM tool.
• Vendor management records: linguist onboarding, test translations, scoring, assignments and payment disputes, subject to privacy limits for individuals.
• Quality assurance logs: error categories, reviewer comments, corrections and sign-offs, where the text excerpts can be removed.
• Client communication: account-manager email and ticket threads about scope, changes and escalations, after redaction.
• Sales and finance operations: CRM opportunity history, proposals, pricing decisions and invoicing exceptions.
• Internal process documents: style-guide management workflows, onboarding SOPs and escalation playbooks.

AI buyers value these because they show multi-step work with decisions and outcomes: a quote becomes a job, the job meets a problem, someone resolves it, the client accepts or complains. Strong companies typically keep records across 10-15+ systems, and archived systems from earlier years help.

The translation memory problem: whose words are they?

Translation memories (TMs) and termbases are the asset owners most often assume they can license, and the one most likely to fail the rights test. A TM stores source and target segments that came from client documents. Even when the LSP built the database, the segments carry client content, and client contracts frequently restrict reuse, require confidentiality or assign ownership of deliverables to the client.

Apply this rule: if a segment came from a client, the client's permission governs it.

Where an asset class fails, leave it out. Where a client is willing, written consent can bring it back in, and the scope is agreed with the company before anything is prepared. SourceX flags rights questions during the inventory stage so owners learn early what is in and what is out.

How should an owner sequence a sale and a data license?

Run the licensing review before you launch the sale, because the answers change what you can tell buyers. A licensed dataset with an exclusive AI-training term should be disclosed to the acquirer, and the buyer will want to know what survives closing.

1. Inventory systems and years of history with the person who administers the TMS, PM tool, CRM and finance system.
2. Sort every asset into "ours", "client-consented" and "client-owned, excluded".
3. Remove owner personal data and personal accounts from company systems; the guide on removing owner personal data from company systems before a sale walks through it.
4. Decide with your M&A advisor whether a license should close before, during or after the sale.
5. Add data licensing to the buyer conversation alongside the usual strategic buyers; the guide to building an M&A buyer list shows where AI data buyers fit.

Nothing is binding until the company agrees price and terms and signs. If the business has already closed or wound down, records can still qualify if they still exist, as covered in licensing data from a wound-down company.

How does a SourceX introduction work for an LSP?

SourceX is the enterprise data transaction layer for AI. It manages the licensing process from sourcing and rights review to delivery and payment, and it does not train AI models. Companies keep ownership; data is licensed, not sold.

For an LSP the sequence runs like this: an owner or a partner such as an advisor or broker introduces the firm; SourceX qualifies size, history, data breadth and rights; the firm completes a data inventory, which is where translation memory and client-file questions surface; and price and terms are agreed as one all-in price with SourceX's fee included. Buyers then review, typically responding within about two weeks once the firm is deal-ready. On close the data is delivered under the agreed redaction rules and the firm is paid once, typically within about 60 days of invoicing after the buyer selects the data.

Deals are typically exclusive for AI training for an agreed term, so an owner must be willing to consider that. The company fit checker gives a preliminary, non-binding screen without contact details.

When is a license the wrong move for a translation company?

Pause if any of these describe the firm:

• Almost all valuable records are client text with no client consent.
• Client contracts prohibit any reuse and the clients will not budge.
• Archives were deleted or the TMS was cancelled without an export.
• The firm never reached 50+ full-time employees at peak (contractors excluded).
• The data has already been licensed for AI training.
• Nobody can run the exports.

Rights not cleared today may be cleared later by seeking client consent, and the who qualifies page lists the full baseline.

Next step

Advisors, brokers and consultants who serve LSP owners can register as a partner and introduce firms that pass the screen. Partners earn 25% of the eligible platform fees SourceX actually collects from the referred company's licensing deals, capped at $100,000 per referred company, and only after the buyer pays and SourceX receives its fee. Owners can also apply directly at sourcex.si/apply.