If everything is high priority, nothing is
Aliases: urgency collapse · everything-is-urgent · top-tier saturation
What it is
When most items in the send queue are labeled top-tier, priority inflation has already happened: the scale no longer discriminates, and the next item is treated as "those so-called urgent ones were harmless." Everything-high is, in the experience, no priority at all—not a missing tag, but a ranking function collapsed to a constant.
This entry is about exhausting the grading scale. It is not about how consequence and latency should be scored, and not about false positives poisoning true alarms. Inflation can occur when every item is "technically high" even if none is a false positive.
Why it happens
Urgency is judged relatively. Working memory keeps a recent sample of what counted as urgent, not an absolute lexicon. Once everyday pushes raise the top-tier baseline, a genuinely costly event cannot rise above it; banners, sounds, and marks are read as wallpaper. Engineering has an isomorphic pressure: teams default to "high" because "low" feels like a demotion. Without a quota, the local optimum is to promote one's own events; the global result is a full top tier.
Inflation is not the same as "make it louder." Loudness is about detection. Inflation destroys comparison. Turning up stimulation on an already inflated channel only speeds adaptation.
Studying it
Count tier share over a window, then match opens, dismissals, and in-session breaks. The signal that the scale has failed is top-tier share rising while top-tier open rate falls—the label remains, the contrast is gone.
Independent variables: share of items in the top tier, number of tiers, whether the default tier is high or low. Dependent variables: open-rate gap across tiers, marginal open rate of the top tier, time to resume the interrupted task.
A lab that balances counts per tier will not see inflation. Use production mix, or deliberately mark 80% as top-tier. Mean satisfaction is a poor substitute for between-tier contrast: people can feel "fine overall" and still treat every banner alike.
Where it stops holding
Sparse, single-class, mandated mail (a monthly bill, an annual inspection) does not exhaust the scale even at top-tier, because the comparison set has almost no competitors. Incident mode (outage, accident) may saturate the channel temporarily, but it needs a declared end and a roll-back or the emergency grade becomes the new everyday. Two-tier systems inflate faster; many tiers with identical presentation are one tier.
Applying it
- Quota the top tier per user per day as a small fraction of sends; overflow drops a tier or goes silent rather than keeping the "urgent" stamp.
- Default to mid or low; promotion to the top requires a written consequence and deadline, with periodic audit.
- When the open-rate gap between top and mid approaches zero, cut top-tier volume before adding visual emphasis.
- Verify with four weeks of top-tier share plotted against that tier's open rate. Rising share with falling opens is scale failure. Halve the top tier and check whether the gap between tiers reopens.