Get an email — with the full report attached — when a QA run finishes. Each rule is yours only. You can scope it to one profile or all, set a threshold, and send to one or more people.
Loading…
Add a rule
Notify me when a run has
Limit to specific error types (optional — default: all types)
What's new
Recent improvements to What's Wrong? — newest first.
🚀 Set up profiles from Dropbox in one click new
In Settings › Integrations, Set up everything from Dropbox runs the whole flow for you: creates the account × language profiles, binds each one's newest checklist, and learns from every LQA scorecard and every past Xbench / LTB QA report — all in order, with live progress. Profiles set up this way also auto-route Dropbox jobs by account & language. Prefer control? Each step still has its own Preview / Apply button. Everything learned stays a suggestion until a PM/Linguist approves it.
📚 Learns from your past QA reports too new
Discover & learn now reads two sources across Dropbox, not one: LQA / MQM scorecards (human corrections → terminology, notes, DNT) and your years of Xbench / LTB QA reports — the findings reviewers marked Ignore / FP become the brands & codes each account keeps untranslated, queued as DNT terms in the right profile's Learn tab. So the new profiles start with the do-not-translate knowledge your reviewers already built up.
🔎 Discover & learn across Dropbox new
In Settings › Integrations, one button walks your whole Dropbox, finds every LQA / MQM scorecard (Amazon, Apple, Sony… any layout), routes each to the right profile by account & language, and queues the corrections into that profile's Learn tab as suggestions. Preview counts first; Discover & queue ingests. New/unknown scorecard layouts are read by an AI fallback, so nothing is missed.
✨ AI brief from your LQA history new
In a profile's Learn tab, Generate AI brief from LQA reads that account's whole correction history and drafts a house-style brief — terminology, register, and whether brands are kept in English or localized (it differs per account). Edit it, then apply it to the profile's AI instructions; the deep-review then follows it automatically.
🗂️ Create account × language profiles new
In Settings › Integrations, Create profiles sets up the agreed account × language profiles (Apple, Amazon CCM / WWCS / AdLoc / General / ATI × TR/AR, Google, iHerb, Netflix…) with structured routing, so the right profile is picked by account and language. Preview first, then create the missing ones.
🎯 Fewer “identical to source” false flags new
Pure template segments common in Amazon content — ${INSERT ASIN(s)}, #if(...), #end — are no longer wrongly flagged as untranslated. Learned from years of reviewer “Ignore” decisions; a real word next to a placeholder still gets caught.
🏷️ One-click brand / not-translatable new
See a brand, product name or code flagged as “identical to source”? Right in the report, click + Protect as brand or + Not translatable. Recurring terms are pre-tagged likely brand and one click clears every occurrence. Anyone can suggest; the term lands in that profile's Learn tab and a PM or Linguist approves it — then it's remembered for every future run (brands are also enforced as must-keep-verbatim).
🔤 Spelling now catches more files
When a file arrives without a declared target language (some Lionbridge / Google exports), the tool now detects the language from the text so the Spelling report runs instead of being silently skipped.
🎯 Fewer false positives
Smarter handling of Arabic numbers & length, segment-split sentences, “#1”-style ranks, dates like “Q1'24 → 2024”, and URLs / placeholders kept identical — so reports stay focused on real issues.
📥 LQA inbox: drop & learn new
In Settings › Integrations you can pair Dropbox folders with profiles. Drop LQA scorecards, style guides or 101 briefs into a paired folder and they're pulled in automatically and turned into rule & note candidates in that profile's Learn tab — ready for you to approve.
🔔 Notifications, now richer new
Under your avatar ▸ Notifications: choose which severities (errors / warnings / info) and even which error types (numbers, tags, terminology, spelling…) should trigger an email — or just AI-confirmed findings. Set a threshold, send to several people, name / edit / pause rules, see when each last fired, and send yourself a test email.
✨ Learn from feedback new
Open a profile and use the Learn tab to teach the tool an account's standards. Upload LQA scorecards (.xlsx) and now also style guides and "101" briefs (.docx, .pdf): it finds recurring corrections and DO / DON'T guidance and proposes ready-made rules and AI notes — you approve the good ones. Prefer to do it in bulk? Settings › Integrations lets you upload many files to a chosen profile at once. Approved rules start checking automatically and can be downloaded as an Xbench checklist.
⚡ Snappier Settings & Reports
The provider connection check now runs in the background and is cached, so pages no longer wait on a slow AI model — your settings and reports appear right away.
🤝 AI: two models must agree
Turn on “require two models to agree” (per profile, or as the org default in Settings) so an AI finding is only kept when a second model also confirms it. Fewer false positives.
⚡ AI: faster, clearer model choice
Pick the AI model from a list labelled by speed (“fastest / fast / slow”), with an Auto option that always picks a fast one. Heavy models no longer time out.
🎯 More accurate checklist results
Fixed how some Xbench-style checklist rules were read (negations, quotes, placeholders and tags), so you get far fewer wrong flags on tag-heavy files.
🧹 Cleaner reports
Spelling now has its own separate report. Checklist findings show a clear “what & how to fix” line. Confusing labels were removed.
📥 Dropbox auto-reports show up reliably
Reports generated from your Dropbox folders now reliably appear in the panel’s history (not only next to the file in Dropbox).
📊 Excel export & glossary import
Download any report as an Excel file. Import a glossary straight from a spreadsheet, mapping your own columns.
🔐 Accounts & roles
Sign in with your own account. Roles (Admin / PM / Linguist / Translator) each land on the right home and see what’s relevant. Forgot-password by email and self-service name/password changes.
📁 Folder picker & auto-profile
Map Dropbox folders to the right profile so the correct checks run automatically; smoother folder browsing.
💬 Send feedback
Spotted something off? The Send feedback button (top bar) reaches us directly.
Help & docs
A quick orientation. The tabs across the top are where everything lives.
Run QA
Drop a bilingual file (XLIFF, SDLXLIFF, mxliff, txlf, xlz, wsxz). Deterministic rules run first; AI review only re-checks rule-flagged segments, so it stays cheap.
Profiles
A profile is a per-client / per-language-pair rule set. The “How profiles work” panel on the Profiles tab walks through creating one.
Reports
Every run is saved here. Open one to see findings side by side with the source, filter by category, and export to Excel.
Settings
AI engine, integrations (Dropbox watcher, Dispatch adapter), auto-profile rules and branding. Admin / PM only.
Found something off? Tell us what’s wrong.
In-editor QA extension —
Run What's Wrong QA straight from the CAT editor — no Dropbox export, no upload. Supported editors: …. More are being added.
1 · Install
Download & unzip. In Chrome open chrome://extensions, turn on Developer mode, click Load unpacked, pick the folder.
2 · Connect with your login recommended
Link the extension to your qa.eltur.co account — it then runs as you, no token to paste.
My connections
—
Advanced · shared team token (admin break-glass)
3 · Use it
Open a job in a supported editor → Run What's Wrong button (bottom-right) → report opens, with the auto-picked profile and a change link. Read-only.
Run QA
two layers · strict rules first, the model only where it earns its keep
Get started
1 · Upload a package
Drop any XLIFF file · format is read from content, not the extension. Works for every account and language pair. Reports and this panel are in English; source and target text stay in their original languages.
Drop file(s) here or click to browse
One or many files, or a zip/wsxz package · all go into one report.
2 · Matched profile & options
The profile is auto-detected from the file (account → division → language pair, most specific wins). Override it below if the detection is off. Profile settings are edited on the Profiles tab.
Scope override (PM/admin) · skip segments entirely. By default every QA already excludes Locked + ICE/Context from the score, and translator-folder jobs also exclude 100%.
Result
Open:Download:
How it works
Layer 1Rule engine · deterministic
Tags, numbers, placeholders, missing targets, brand/DNT protection, punctuation, spacing, length, consistency. Cheap, exact, instant. Findings here are facts.
Layer 2AI deep-review · only the exceptions
Runs only on segments the rules flagged as suspicious · meaning shifts, omissions, additions, register. Produces suggestions you can copy; it never overwrites the master file. Runs for panel, Dropbox and Dispatch alike.
In your editorWorking in ATMS, WorldServer, fbCAT, Smartling or XTM? Run QA without exporting.
Add the browser extension and a Run What's Wrong button appears right inside the editor — it QAs the open job from your own session and gives you a report. No Dropbox export, no upload.
Profiles
one profile per account · division · language pair · edited here, honoured by panel, Dropbox and Dispatch alike
New here? How profiles work
A profile is a saved set of checks for one kind of work, usually one account, division and language pair, for example Amazon · CCM · EN→DE. When a file comes in, the matching profile is picked automatically from the folder it lands in, so nobody has to choose checks by hand each time.
Most people start with one profile per language pair, and that is perfectly fine. As you grow you can add more specific ones (a whole account, or a single division inside it) and the most specific match wins.
To create one: click + New profile, set the account, division and language pair it covers, turn on the checks you want (glossary, numbers, locale format, spelling, AI review), and save. You can also copy an existing profile as a starting point.
The same profile is honoured everywhere: the panel, Dropbox auto-runs, and Dispatch. Set it once, it applies across all three.
Pick a profile to edit
Profiles are grouped by account.
Reports
every run kept · reopen a report whenever you need it
Insights
Findings by category
Errors by account & type
No data for this period yet.
Run history
When (UTC)
File
Account · Division
Pair
Profile
Trigger
Findings
Crit
Warn
Info
AI
AI ✓
No runs yet.
Report
Download:
Settings
org-wide defaults and connectivity · most values are set in the server environment and shown here read-only
AI deep-review
The AI layer only reviews segments the rules flagged. It produces suggestions and never edits the master file. API keys are set in the server environment (GEMINI_API_KEY / OPENAI_API_KEY / DEEPSEEK_API_KEY) · values are never shown here.
Org-wide default for every run. A profile's own AI provider (Style & AI tab) overrides this.
Require two models to agree (default)
When on, an AI finding is kept only if a second, independent model also confirms it (cuts false positives). New profiles default to this; each profile can still turn it off or pick its own models.
Daily AI-run limit per role
Caps how many AI deep-reviews each person can trigger per day (resets at midnight TR time). Over the limit, QA still runs the rule checks — only the paid AI step is skipped. Admin & PM are always unlimited.
Set to -1 for unlimited.
Approx. cost per 200-segment file
Gemini ~$0.015 · GPT mini ~$0.02 · DeepSeek ~$0.008. AI only runs on rule-flagged segments, so real cost is usually well below this.
General
Fallback profile
Used when no account · division · language-pair matches a job.
Integrations
This covers three things: knowledge upload (feeds Learn), set up from Dropbox (profiles + checklists + learning), and connections & automation (Dropbox watcher · LQA inbox · Dispatch · auto-profile routing). All secrets live in the server environment; the value is never shown.
Knowledge (Learn) — bulk ingest
Upload many feedback files at once and route them to a chosen profile: LQA scorecards (.xlsx), style guides or “101” briefs (.docx, .pdf). Recurring corrections and DO / DON'T guidance become candidate rules & notes (AI-vetted). This only queues them — you review and approve in that profile's Learn tab, where nothing goes live until you approve it.
LQA inbox (Dropbox) — auto-pull
Pair Dropbox folders with profiles. Any LQA file (.xlsx, .docx, .pdf) dropped into a folder is distilled into that profile's Learn queue automatically — you just review & approve. Processed files move to a _ww_done sub-folder. The watcher checks on its normal schedule.
Dispatch adapter token
Shared secret Dispatch presents on every /qa-adapter/v1 call (WW_ADAPTER_TOKEN).
checking…
Dropbox
Refresh-token auth for the watched folder (DROPBOX_REFRESH_TOKEN).
checking…
DB
Dropbox watcher
Watches the team Projects tree: when a file lands in a from* folder, a QA report is generated and written back into the same folder. A webhook makes this near-instant; a cron sweep is the fallback. Use the auto-profile rules above to control which profile each folder uses.
Set up profiles from Dropbox
one flow: create · enrich from every source · auto-route
One button does it all: creates the account × language profiles, binds each one's newest checklist, and learns from every LQA scorecard and every past Xbench / LTB QA report across Dropbox — queuing terminology, notes and DNT / untranslatable terms into each profile's Learn tab. Everything queued is a suggestion a PM/Linguist approves before it changes a check. Profiles created this way also auto-route Dropbox jobs by account & language. You can also run any single step below. Admin.
Run a step on its own (advanced) — preview, or see per-profile / unrouted detail
1 · Discover accounts
scan Dropbox for active client / division work
Walks the whole Projects tree and lists each client · division with its latest delivery date, languages, and matching checklist / style-guide — so you can see who's active (last 12 months) and set up profiles. Runs in the background (a few minutes) and writes a report to Dropbox (_WW_account_discovery.md). Admin.
2 · Create profiles
the account × language profiles from the Dropbox inventory
Creates one profile for every account that has a checklist in Dropbox (~55) — each Amazon workstream kept separate (CCM, WWCS, AdLoc, ATI, AWS, HPB, Music, D2AS, Alexa, Policy, Vendor, General × TR/AR), plus Apple, Google, iHerb, Netflix, Riot, TikTok, WhatsApp, VMware, Sony, Shopify and the rest. Each is auto-marked with structured routing. Preview lists them and shows which already exist (it never duplicates `apple_tr` / `amazon_en_tr`); Create makes only the missing ones. Admin.
3 · Bind checklists
attach each profile's newest Dropbox checklist
For every profile (existing & new), finds its account's newest Xbench checklist in Dropbox (by modified date — drafts / experimental / obsolete / archived copies skipped) and binds it. Also sets the org base layer to the newest Elturco_General_v2, which every profile inherits. Preview shows what would bind; Bind now applies. Admin.
4 · Discover & learn from Dropbox
QA reports first, then LQA scorecards → each profile's Learn
Learns from two separate sources, each its own button: past Xbench / LTB QA reports (reviewer Ignore / FP marks → DNT terms) and LQA / MQM scorecards (human corrections → terminology + notes). Each file is routed to the right profile and queued to its Learn tab as a suggestion — nothing changes until you approve. Preview just counts. Re-runs are incremental — only files new since the last scan are processed. If something fails, the exact error is shown. Admin.
Ongoing
Auto-profile rules
all folders → profiles, in one place
This is the global view of every folder → profile rule. You can also manage a single profile's folders inside it, under Profiles ▸ (open a profile) ▸ Routing — same rules, scoped to that profile. Folder (or sub-path) → profile: when a job's path contains the folder(s), that profile is used, after Dispatch's account/division match and before the generic fallback. Most specific wins (so Amazon/adloc beats Amazon). Admin / PM.
Folder token
Profile
Source
Test a path
Profile patterns (read-only · these take precedence over folder rules)
DI
Dispatch
When a translator sends a job to QA, Dispatch passes the project’s account, division and language pair and the right profile is picked automatically · most specific wins. If a profile has AI enabled, the deep-review runs too.
Run What's Wrong QA straight from the CAT editor — no Dropbox export. The translator clicks one button; the extension reads the open job from their own logged-in session and gets a report back. Supported editors: …. More editors are being added.
1 · Install the extension
Download the package, unzip it, then in Chrome open chrome://extensions, turn on Developer mode (top-right), click Load unpacked and select the unzipped folder.
2 · Connect with your login recommended
Click below to link the extension to your qa.eltur.co account. The extension then runs as you — reports are attributed to you and your role's daily limits apply. Nothing to copy or paste.
My connections
—
Advanced · shared team token (admin fallback)
2b · Point it at this server
Click the extension icon and paste these two values, then Save.
3 · Use it
Open a job in a supported editor → a Run What's Wrong button appears (bottom-right) → click it → the report opens. The report shows which profile was auto-picked, with a change link if it's wrong. Read-only: the extension never changes the job.
Force a minimum version admin
Block older extension builds from running QA until the user updates. Turn the block ON only after the team is on a build that reports its version (current shipped: —), or everyone is locked out at once.
Branding
White-label the panel and reports for your organization. Blank fields keep the defaults.
Logo
Shown in the footer of the panel and reports, and on the sign-in screen. PNG / SVG / JPEG, up to 2 MB.
Users & roles
Admins manage who can sign in. Roles: admin (everything), pm (settings except General + manage auto-profile), linguist (create/edit profiles), translator (run QA + see only their own reports).