Skip to main content
AGENCYSCRIPT
CoursesEnterpriseBlog
đź‘‘FoundersSign inJoin Waitlist
AGENCYSCRIPT

Governed Certification Framework

The operating system for AI-enabled agency building. Certify judgment under constraint. Standards over scale. Governance over shortcuts.

Stay informed

Governance updates, certification insights, and industry standards.

Products

  • Platform
  • AI Scripts
  • Certification
  • Launch Program
  • Vault
  • The Book

Certification

  • Foundation (AS-F)
  • Operator (AS-O)
  • Architect (AS-A)
  • Principal (AS-P)

Resources

  • Blog
  • Agency Archetype Quiz
  • Free Live Training
  • Build AI Agents Masterclass
  • Build with AI Challenge
  • OS Plugin Install
  • Verify Credential
  • Enterprise
  • Partners
  • Pricing

Company

  • About
  • Contact
  • Careers
  • Press
© 2026 Agency Script, Inc.·
Privacy PolicyTerms of ServiceCertification AgreementSecurityCookies

Standards over scale. Judgment over volume. Governance over shortcuts.

On This Page

Stage One: TriageSorting Content by the Cost of ErrorWhy Triage Comes FirstThe Output of TriageStage Two: InstrumentBuilding the Foundation Before VolumeWhat Instrument OwnsThe Temptation to Skip InstrumentStage Three: ExecuteTranslating at Tiered DepthParallelizing Where PossibleThe Discipline of Post-EditingStage Four: RefineClosing the Feedback LoopWhy Refine Makes the System CompoundApplying TIER in PracticeSequential First, Then CyclicalWhen to Compress the ModelDiagnosing a Struggling Project With TIERMapping TIER to RolesWho Drives Each StageAvoiding the Two Classic FailuresFrequently Asked QuestionsWhy use a named model instead of a checklist?Which stage do teams skip most often?Can I reorder the stages?Does TIER assume a particular tool?How does TIER handle ongoing content after launch?Key Takeaways
Home/Blog/How the TIER Model Structures Localization Work
General

How the TIER Model Structures Localization Work

A

Agency Script Editorial

Editorial Team

·June 11, 2017·7 min read
ai translation and localization toolsai translation and localization tools frameworkai translation and localization tools guideai tools

Most localization advice arrives as a pile of tactics: build a glossary, tier your content, test your layouts. The tactics are sound, but without a structure to hang them on, teams apply them inconsistently and forget half of them under deadline. A named model gives the work a shape that survives pressure.

This article introduces the TIER model: Triage, Instrument, Execute, Refine. It is a deliberately small framework, four stages, each owning a clear set of decisions. The goal is not novelty for its own sake but a vocabulary a team can share, so that "we skipped Instrument" becomes a sentence anyone can act on.

The model is sequential the first time through and cyclical thereafter. You move through all four stages to launch a market, then loop back to Refine continuously as content and markets grow.

Stage One: Triage

Sorting Content by the Cost of Error

Triage is the decision about how much human attention each piece of content deserves. The core question is the cost of being wrong. Legal disclosures, medical instructions, and brand-defining marketing sit at the top; internal changelogs and generic product specs sit at the bottom.

Why Triage Comes First

Every later decision depends on this sort. The tooling you choose, the reviewers you staff, and the timeline you promise all flow from how your content distributes across the risk tiers. Teams that skip triage end up applying uniform effort to non-uniform stakes, which is both wasteful and risky. The trade-off reasoning behind these decisions is detailed in Weighing the Real Costs Behind Localized Copy.

The Output of Triage

Triage is done when you can point at any piece of content and name its tier and the reason. The deliverable is a simple map: content types down one axis, risk tier across the other, with a one-line justification each. This map is not bureaucracy; it is the contract that prevents arguments later about why the legal page got full review and the changelog did not. When stakeholders disagree about effort, you return to this map rather than relitigating each item.

Stage Two: Instrument

Building the Foundation Before Volume

Instrument is the setup stage: glossary, translation memory, locale configuration, and quality metrics. It is the least visible and most leveraged work in the model, because everything done here multiplies across every language and every future piece of content.

What Instrument Owns

A populated glossary locks terminology. Seeded translation memory captures prior approved work. Locale settings handle dates, currency, and number formats. And the quality metrics you will track later are defined now, so you measure from the first market rather than retrofitting later. The metric definitions belong to the approach in Reading the Signal in Localization Quality Numbers.

The Temptation to Skip Instrument

Instrument is tedious and produces nothing a stakeholder can see, which is exactly why teams under deadline skip it. The cost of skipping is invisible at first and brutal later: every language repeats the same terminology errors, reviewers correct the same mistakes over and over, and quality never improves because nothing is captured. The rule is that Instrument time is never the place to cut. If the timeline is tight, cut the number of markets in the first wave, not the foundation that serves all of them.

Stage Three: Execute

Translating at Tiered Depth

Execute is where content meets the engine. Each tier follows its triage assignment: machine-only content ships after automated checks, mid-tier content gets spot-checks, and top-tier content gets full human review of machine drafts. The key discipline is that reviewers post-edit drafts rather than translate from scratch, which is where the speed gain lives.

Parallelizing Where Possible

Once the foundation from Instrument is in place, languages can run in parallel. Machine drafts populate overnight; human effort concentrates on correction during the day. This is the stage where the project visibly moves, but its pace is entirely set by how well the earlier stages were done.

The Discipline of Post-Editing

Execute has one easy-to-miss discipline: reviewers must post-edit, not retranslate. The temptation, especially for skilled translators, is to discard the machine draft and write their own version, which throws away the speed advantage entirely. Train reviewers to fix what is wrong and leave what is right. When a draft is so poor that post-editing is slower than starting fresh, that is a signal to fix Instrument, the glossary or the engine choice, rather than a reason to abandon the approach. Protecting the post-editing discipline is what keeps Execute fast.

Stage Four: Refine

Closing the Feedback Loop

Refine is the stage teams most often drop, and it is what separates a one-time launch from a durable system. Every human correction feeds back into translation memory and the glossary, so the same error never recurs. Post-launch correction tickets are monitored per language to catch systemic issues.

Why Refine Makes the System Compound

Without Refine, quality plateaus and every new market repeats old mistakes. With it, the system gets cheaper and better over time, because the reusable assets grow with every project. This is the stage that turns localization from a cost center into infrastructure.

Applying TIER in Practice

Sequential First, Then Cyclical

The first time a team localizes a product, they walk through Triage, Instrument, Execute, and Refine in order. After launch, Refine runs continuously, and any significant new content re-enters at Triage. A worked example of the full cycle appears in One Team, One Quarter, and Forty Markets to Reach.

When to Compress the Model

For tiny, low-stakes projects, Triage and Instrument compress into an afternoon and Refine may be informal. The model still applies; it simply scales down. The fastest entry path is sketched in Shipping One Localized Page Before You Buy a Platform.

Diagnosing a Struggling Project With TIER

The model's real payoff is diagnostic. When a localization effort is struggling, the four stages give you a checklist of causes. Inconsistent terminology points to weak Instrument. Effort wasted on low-stakes content points to absent Triage. Slow reviewer throughput points to Execute translating from scratch rather than post-editing. Quality that never improves points to missing Refine. Instead of a vague sense that something is wrong, the team can name the failing stage and fix it directly. That shared diagnostic vocabulary is the main reason to adopt a named model over an ad hoc list of tactics.

Mapping TIER to Roles

Who Drives Each Stage

The model also clarifies ownership. Triage is led by whoever understands the business stakes of the content, often a product or marketing lead, because they can judge the cost of error. Instrument is shared between localization and engineering, since glossary and memory are linguistic while locale configuration and metrics instrumentation are technical. Execute is owned by reviewers working inside the tooling. Refine is owned by whoever holds the long-term process, since it is the stage that runs forever. Naming an owner per stage prevents the common failure where Instrument and Refine, the unglamorous stages, fall to nobody.

Avoiding the Two Classic Failures

Two failure patterns recur often enough to name. The first is front-loaded enthusiasm: a team does Triage and Instrument well, ships through Execute, and then never returns to Refine, so quality plateaus and the next project repeats old mistakes. The second is the rush: a team under deadline skips Instrument entirely and jumps to Execute, producing fast output riddled with inconsistent terminology that costs more to clean up than the foundation would have cost to build. The model's value is that it makes both failures visible and nameable before they cause damage, which is the entire point of having a shared structure rather than a loose collection of tactics.

Frequently Asked Questions

Why use a named model instead of a checklist?

A checklist tells you what to do; a model tells you why and in what order. The shared vocabulary lets a team diagnose where a project went wrong, such as "we under-invested in Instrument," which a flat list does not support.

Which stage do teams skip most often?

Refine. The launch feels like the finish line, so the feedback loop never gets built, and quality stops improving. Skipping Instrument is the second most common and most damaging omission.

Can I reorder the stages?

Not for a first launch. Triage informs Instrument, which enables Execute, which produces the corrections that feed Refine. The dependency chain is the point of the model.

Does TIER assume a particular tool?

No. It is tool-agnostic. Any translation management platform with glossary, translation memory, and a neural engine can support the model. The framework governs how you use the tool, not which one you buy.

How does TIER handle ongoing content after launch?

New content re-enters at Triage and flows through the stages again, while Refine runs continuously in the background. The model is cyclical by design, not a one-time waterfall.

Key Takeaways

  • TIER stands for Triage, Instrument, Execute, Refine, applied sequentially then cyclically.
  • Triage sorts content by the cost of error and drives every later decision.
  • Instrument builds glossary, translation memory, and metrics that compound across all languages.
  • Execute post-edits machine drafts at depths set by triage, parallelizing where possible.
  • Refine feeds corrections back so the system improves rather than plateaus.

Search Articles

Categories

OperationsSalesDeliveryGovernance

Popular Tags

prompt engineeringai fundamentalsai toolsthe difference between AIMLagency operationsagency growthenterprise sales

Share Article

A

Agency Script Editorial

Editorial Team

The Agency Script editorial team delivers operational insights on AI delivery, certification, and governance for modern agency operators.

Related Articles

General

Rolling Out AI Hallucinations Across a Team

Most teams discover AI hallucinations the hard way — a confident-sounding wrong answer makes it into a client deliverable, a legal brief, or a published report. The damage isn't just to the output; it

A
Agency Script Editorial
June 1, 2026·11 min read
General

A Model Behind an API Is Only Potential

Large language models don't do much on their own. A model sitting behind an API is potential, not capability. What converts that potential into something useful—something that drafts, classifies, summ

A
Agency Script Editorial
June 1, 2026·11 min read
General

Case Study: Large Language Models in Practice

Most teams that fail with large language models don't fail because the technology doesn't work. They fail because they treat deployment as a one-time event rather than a discipline — pick a model, wri

A
Agency Script Editorial
June 1, 2026·11 min read

Ready to certify your AI capability?

Join the professionals building governed, repeatable AI delivery systems.

Explore Certification