This month's releases focus on clearer performance signals, review effort spent where it matters, and faster navigation across the platform. Here's what's new in the platform.
The metrics section on My Tasks has a new layout. You get a Speed chart showing daily annotation points per hour against your project's target line, plus three KPI cards: Correction ratio, Time allocation ratio, and Quality score with a hover breakdown of precision, recall, geometry accuracy, and property accuracy. A project selector scopes everything to one project, defaulting to the one you worked on most recently.
The KPI cards cover the last 7 days and carry trend arrows, so you can tell at a glance whether your numbers are moving the right way.
Available in: My Tasks · Docs
A new Study tab in Project Management follows a group of project members over a set period and compares their work against the rest of the project. Choose P2P coaching, one coach and their students, or Group tracking for any group. Each study tracks Annotate Speed against target, Quality score, Time allocation ratio, and Correction ratio.
The speed change since start only shows a colored arrow when it exceeds the group's normal day-to-day movement, so a noisy week does not read as a win or a loss. Two tables give you the per-member view: a summary of speed and quality from start to end, and a daily KPI table colored against baseline. Use studies to see whether a coaching session, a new feature, or a course actually moved the numbers.
Available in: Project Management · Study tab · Docs
Quality Review phases can now select inputs based on estimated quality instead of a flat share of every annotator's work. Set a target quality, say 95 percent: inputs at or above it skip manual review, while inputs below it, and any input without an estimate yet, go to review. Estimates build on reviewed work from roughly the last 60 days, so annotators whose work keeps getting skipped cycle back into review as history ages out.
Flat-percentage review spends reviewer time on work that is already good. Quality-based sampling points it at the inputs most likely to have problems, reducing review volume without lowering the quality bar.
Available in: Request Settings · Phases · Quality Review
Review phase settings now offer a middle ground between fully editable and fully locked. The "Block review edits" option on each review phase card lets reviewers adjust existing object geometry while property changes and creating or deleting objects stay blocked. Reviewers can correct what matters without risking unintended changes to classifications or the object set.
Available in: Request Settings · Phases
Plus: the Quality Review and Expert Verification overview has been tidied up. Gate status now sits on the Overview, and two new cards show the actual and planned input review samples.
The omnibar now searches pre-labels by UUID, finds scenes without inputs, pending scenes, or scenes with errors, and jumps to a workspace's project page by workspace name or UUID. Leading and trailing spaces no longer break a search. Scene and pre-label tables also load noticeably faster, down from waits of up to 30 to 40 seconds, by fetching only the last three months of history.
Available in: Omnibar (cmd+K)
A set of annotation-tool updates is already live on the next environment and lands in the main app after summer: an expanded camera drawer that shows every 2D camera in one grid, a toolbar and direct property editing in the 2D grid view, a group-by-frame-or-camera control in the image grid, and live checkers that validate every change instantly, no Validate button needed.
Questions? Reach out to your account team or contact us at support@kognic.com.