Localization Services

Forrester's difficulty with localization services was never evidence. It was the unit of measurement. The industry sells human judgment wrapped around software, and a scorecard has to pick something countable.

On 23 September 2026 it picked. The Forrester Wave: Localization Services, Q3 2026 is Forrester's first evaluation of the category, and most of the names it graded had already been graded once, on the technology side.

Inside The Forrester Wave: Localization Services, Q3 2026

The evaluation published on 23 September 2026 under report code RES200827, subtitled "The 11 Providers That Matter Most And How They Stack Up" and authored by Kathleen Pierce with three contributing analysts. It scored eleven providers against twenty-seven criteria after roughly eighteen weeks of research.

RWS, Smartling, Centific and Welo Global were named Leaders. TransPerfect, Lionbridge and Argos Multilingual were placed as Strong Performers. Acolad, Acclaro, Vistatec and LanguageWire were placed as Contenders.

Two details in the published positioning figure carry more weight than the tier labels. The figure draws three zones and two dividing curves, so it shows no Challengers band, which is a statement about the shape Forrester gave the market rather than a claim that no vendor was assigned to a fourth zone. And exactly two vendors carry the customer-feedback ring, RWS and Smartling, which corroborates Smartling's own statement that only two of the eleven received above-average customer feedback. Neither ring is doubled, so this edition has no Customer Favorite.

The first services scorecard crowned the firms that also sell the software

RWS and Smartling won Leader placements here. Both were also scored in The Forrester Wave: Translation Management Systems, Q3 2025, the inaugural technology evaluation of the adjacent market.

Centific makes it three. It took a Leader placement in the services evaluation and appeared in the technology field as well. So three of the four Leaders in this Wave are vendors that sell the software as well as the service, and the fourth, Welo Global, is the only Leader that did not appear in that technology field.

That outcome answers a question the technology Wave raised without settling. When machine translation compressed per-word rates and AI agents started compressing the workflow costs around the translation itself, the open question was whether the services layer would be hollowed out by the software layer or would end up owning it. Forrester's first services scorecard places the firms that do both at the top.

The services evaluation's own framing points the same way. The shift it describes is away from faster translation and toward technology that understands context, culture and specialist demands, with integration, governance and AI risk management as the differentiating criteria. Those are software disciplines. A services provider being scored on governance and AI risk is a services provider being scored on its platform.

What the page that stood here got wrong, and what it got right

This article replaces one that argued the evaluation did not exist and could not exist yet. Its central claim was accurate when it was written and expired on 23 September 2026: no Forrester Wave had ever been published under the name Localization Services.

Its open questions, though, survived contact with the answer. It asked what the unit of being scored would be, offering three candidates: the linguist network, the AI orchestration layer, or the account management that survives a merger. Forrester's twenty-seven criteria span all three, and the Leader rung resolves the ambiguity in favor of vendors that operate the orchestration layer themselves.

It also asked which economics the evaluation would choose, cost per word or cost per outcome across an automated pipeline. That one is now answerable from the criteria rather than from speculation, and the answer favors the second: a twenty-seven-criteria scorecard that weights integration, governance and AI risk is not a per-word scorecard.

TransPerfect finished on the boundary it has been straddling

The most acquisitive vendor in the field landed on the edge. TransPerfect was resolved as a Strong Performer, but its position in the published figure sits essentially on the line between the Leaders and Strong Performers zones, and the placement should not be read as comfortable.

The history makes the position legible. TransPerfect bought Unbabel, announced just before the technology Wave published, which is the classic services-industry move: consolidation to buy the technology margin back rather than build it. It was scored separately from Unbabel in that 2025 evaluation because the deal closed too late to be treated as one vendor.

A firm that buys its way into the technology layer and lands on the boundary between the top two bands is a firm whose integration work is not finished. That is not a criticism of the acquisition. It is what a boundary placement means in a market where the top band is occupied by vendors that already had both halves.

Where the technology scorecard still stands

Nothing in the Translation Management Systems evaluation changed. It remains the Q3 2025 edition, Forrester's inaugural assessment of that market, announced on 10 September 2025 by Kathleen Pierce, scoring twelve vendors across nineteen current-offering criteria and seven strategy criteria.

Its Leaders were Phrase, XTM and LILT. Phrase took the highest possible scores in twenty-one of the twenty-six criteria, on configurable workflows, continuous localization and AI agents that automated ninety-nine percent of workflows for some reference customers. XTM was recognized for orchestration, AI agent roadmaps, workflow controls and fit for regulated industries. LILT was recognized for custom AI models paired with human-verified precision.

The two evaluations must be read as distinct. RWS, Smartling, Centific, TransPerfect and Lionbridge appear in both, which is precisely the overlap that makes them easy to conflate. The technology Wave scores platforms. The services Wave scores the firms that run programs on top of those platforms. A buyer who needs the platform and a buyer who needs the program are shopping in two different reports that happen to share a vendor list.

Where the market is actually moving

The most transferable signal in the services evaluation is the customer feedback result. Only two of eleven providers received above-average feedback from the references Forrester interviewed, and both of those carry the ring in the published figure.

Read that against the size of the field. Eleven providers cleared the inclusion threshold and nine of them could not muster above-average reference feedback. In a market built on long-term client relationships and high switching costs, that is a striking result, and it maps onto the integration lock-in that the technology Wave already documented, where reference customers wanted to change providers but were held back by the integration investment they had already made.

There is one more piece of housekeeping that matters to anyone working from older vendor lists. Welo Global is Welocalize. The company rebranded in April 2026 and now describes itself as delivering multilingual content, data and AI solutions through five brands. Any list of this market compiled before that date carries the old name.

Three questions for a market that finally has a scorecard

Does your provider own the orchestration layer, or rent it? The Leader rung is dominated by vendors that sell both halves. Ask which parts of your workflow run on the provider's own technology and which run on a competitor's platform, and get the answer in writing. A services firm reselling someone else's orchestration layer has a margin structure that will eventually show up in your renewal.

What is your automation number, and what does it cost per outcome? Ninety-nine percent workflow automation is a platform claim. Ask your services partner what percentage of your content still requires human review, in your language pairs, and what that review costs per unit delivered rather than per word. The criteria Forrester chose suggest per-outcome economics are now the ones that separate the field.

What happens to your workflow when the contract ends? This question outlived the old article because the lock-in finding outlived it. Ask what happens to your automations, glossaries and translation memory at termination, and get the export paths in writing. A provider that cannot answer is already planning to keep you through inconvenience.

For European buyers, one standing clause is worth repeating regardless of tier placement. Translation memory and feedback data are records of your customers' communications, and a provider that has changed ownership needs its data processing agreement reviewed rather than assumed.

Forrester does not endorse any vendor named here, and tier placement should not be read as a recommendation to buy.

A close neighbor in Forrester's own coverage is Marketplace Development Platforms, where this Wave scores one category built from four distinct businesses, selling physical goods to consumers versus digital services to businesses, and its two Leaders each dominate one end of that ladder and barely compete at the other.