A/B Test Bucket Assignment
Units travel down an intake rail to a splitter and drop alternately into a control bucket and a variant bucket, both fill levels rising to an even split before the cohort clears.
Units travel down an intake rail to a splitter and drop alternately into a control bucket and a variant bucket, both fill levels rising to an even split before the cohort clears.
A finished run waits at a three-way gate while the picker weighs the outcomes, then the chosen route swings open along its lane and takes the run to its terminal.
A visitor key enters a hashing dial whose pointer sweeps the notched slot ring and stops on the same slot every cycle, that slot latching lit while the rest stay dim.
An audience rule sweeps across a population grid column by column, filling the cells that qualify for the test and striking out the ones that fail it.
A variant panel paints in and only once it has rendered does an exposure tick drop down the guide onto the log rail beneath it, the rail advancing one notch per render.
A staircase of rollout stages is drawn one tread at a time as exposure climbs from a sliver of traffic to half of it, the dashed level rule stepping up the axis with each stage.
An override toggle is thrown and the points blade swings off the bucketed main line, diverting the waiting unit up the forced variant spur to its terminal.
A live metric trace runs inside a guardrail corridor until it drops through the floor rule, where the breach point lights and the halt square latches the run to stopped.
A ring of population cells turns over to the treatment arc by arc while one reserved wedge stays outlined and untouched, the holdback still there when the ring completes.
A unit drops into a free slot on the first experiment layer and the matching slot on the layer below crosses out, so no unit is ever claimed by two running tests at once.
A marker ring steps along a track of lifecycle stations from drafted to concluded, drawing the run line behind it and latching each station as it leaves.
Three return visits arrive one per row of a two-column assignment grid and each lands in the same column, a bracket closing down the side once the assignment has held.
An evaluation head steps down a stack of targeting rules, marking each one it falls through, until the first rule that matches outlines and claims the unit.
A handle is dragged along the seam of a two-segment allocation bar, widening the variant share against the control while the pointer below reads the new split off a fixed percentage scale.
The carryover left by a finished run drains out of the washout window on the timeline, and only once the gap is clear does the next run fill its planned slot.