Categories
Named and ordered by you. A preferential category can carry a weight of zero, so a reviewer records a preference without it costing the delivery a point.
Structured human review against MQM and DQF-style templates. Categories, severities and weights are yours to rename and re-weight, and the automatic check pre-fills the card so the reviewer starts from evidence instead of a blank grid.
LQA is in early access. Ask for it, we test it with you on your own template, and switch it on for your workspace.
A template is the whole scoring contract, not a set of tick boxes. If your client counts a critical as a hard fail and ours does not, you change the template rather than argue with the number. The default card weights minor 1, major 5 and critical 25, per 1,000 words.
Named and ordered by you. A preferential category can carry a weight of zero, so a reviewer records a preference without it costing the delivery a point.
Any number of levels with your own penalties. Mark a level as a hard fail when a single instance should sink the delivery outright.
Penalty per 1,000 words on a 100-point or fractional scale, with your pass mark and the review band underneath it.
The whole file, a random sample or a stratified slice of a set size, with a seed so the same sample can be drawn again.
You decide which automatic rule lands in which category and how an engine severity translates into yours. The reviewer opens a card that is already populated.
Each job keeps a copy of the template it was created with, so editing the template tomorrow never silently re-scores work you already finalised.
Each of these is a working template you can clone and edit. They exist because real programmes score differently, not to pad a feature list.
The general-purpose card: accuracy, fluency, terminology, style, locale convention, design, verity, other and a zero-weight preferential lane.
Modelled on a large advertising programme: nine categories, a critical level that hard-fails the delivery and a fractional scale rather than points.
For copy where correct is not the point: brand voice, transcreation, register and fluency as first-class categories, next to accuracy and terminology.
Build a card from scratch or paste your client's. Categories, severities, weights, bands, sampling and the mapping from automatic findings are all editable.
The automatic run maps its findings into your categories and severities, so the card opens with the mechanical work already done.
The linguist keeps, edits or drops each seeded finding and adds what only a person can see. Every entry carries a span and a comment. Each finding keeps the reviewer who last changed it.
Penalties roll up per 1,000 words against the frozen template. Pass, review band or fail, with the arithmetic shown rather than asserted.
The card is finalised and exported for the client. Only a user with manage rights can reopen it. Finalised scorecards aggregate by account, so a programme can be read as a trend.
An 87 on your card and an 87 on a client's card are different statements. Comparing vendors or months means running them on the same template, which is exactly why templates are versioned and frozen.
Seeded findings are a starting point. A linguist who drops one is not fighting the system, and the record keeps both the original finding and the decision.
Send us your scorecard and one delivery. We set the template up, run the automatic pass into it and show you the card a reviewer would open.