A number that changes mid-glance interrupts exactly the moment a reader was trying to lock onto it

What it is

Auto-refreshing real-time dashboards change data mid-read: numbers swap, bars resize, the curve's tail extends. These changes strike exactly when attention is on the content—the user has just locked onto a number to interpret it, and the refresh replaces it with a different one. The result is not "seeing updated data" but "the previous glance and this glance show different values—which one do I trust?" The more frequent the jumping, the stronger the disruption and the lower the reading comprehension.

Why it happens

The disruption comes from the refresh cost of visual working memory: reading a number involves perceptual encoding → working memory maintenance → semantic understanding, spanning roughly 1-2 seconds. If an auto-refresh changes the visual input mid-stage, the value held in working memory conflicts with the new value on screen—pre-attentive processes detect the change (the inverse of change blindness: changes during fixation are extremely hard to ignore), and attention is yanked from understanding back to re-confirmation, breaking the comprehension chain. Jumping also compounds: each disruption forces the user to re-locate where they were reading, and the re-location cost stacks into significant reading-efficiency loss across multiple refreshes. Layout shift is a second layer: if a refresh causes reflow (a longer number pushes neighbors aside, bar ordering changes), the user's attention anchor ("I'm reading the second column") fails with the layout, and the re-anchoring cost rises further.

Studying it

The standard paradigm for studying jump disruption is a timed reading task: present a set of numbers or charts on screen, trigger data updates at varying frequencies during the task, and measure how comprehension accuracy and reading speed decay with update frequency. Dependent variables include single-read accuracy, cross-refresh memory retention, and subjective interference ratings. Eye tracking adds fixation-break counts and regressive-saccade rates, revealing where in the reading chain the disruption strikes. On real dashboards, A/B testing refresh intervals against task completion time and error rate provides a more ecologically valid measurement. A methodological caveat: lab updates are controlled discrete events, while real data streams update irregularly—decay curves do not extrapolate directly to irregular update schedules.

Where it stops holding

The disruption's severity depends strongly on the user's task: operators monitoring a wall display expect data to change (the change is the signal), while an executive glancing at a dashboard is interrupted by it (the change is noise). Likewise, a jumping number disrupts more than a trend chart's extending tail—the former replaces the value being read, while the latter only changes the curve's shape (history stays fixed), confining the disruption to the tail region where attention already concentrates. For scrolling trend charts, the trailing auto-extension is usually anticipated and tolerated; a full-chart redraw is not.

Applying it

• Turn off auto-refresh by default; offer a manual refresh button or a low-frequency auto-refresh (≥30 seconds), letting the user choose when to accept updates.
• Fade transitions (300ms) on refresh instead of hard cuts, helping the visual system link old and new states as one object updating.
• When numbers change, update only the changing digits (e.g., 3,412 → 3,415 animates the last two digits), keeping the leading digits stable as a visual anchor.
• Verification: have a colleague complete one reading task on the dashboard; count how many auto-refreshes interrupted them and ask what value was on screen each time—misses are evidence of disruption.

Related

• Same group: U7.09.2 Update frequency should match decision frequency · U7.09.3 Data must carry a visible timestamp
• Nearby: U1.02.1 Position along a shared scale is the most accurately read channel · U7.07.3 One-screen capacity is bounded by interpretation ability
• Search terms: real-time update distraction · layout shift · auto refresh