The default sort determines what most users will ever see
Aliases: sort default bias · ranking default · default ordering
What it is
In query logs, people who actively change sort are a minority. The default sort determines what most users will ever see, because most sessions treat the order of the first screen as the whole of what they will look at. A default is not a neutral placeholder; it is the product choosing, for the silent majority, which kind of “better” to use. Relevance, time, popularity, personalization—whichever is default puts that kind of answer into almost everyone’s field of view, and parks the rest in a menu that is rarely opened.
This leaf does not argue whether relevance and recency should be switchable—switching is a path for the few who act. The default decides what the people who do not act see.
Why it happens
Changing sort requires noticing that the current order is wrong, finding the control, and picking another option. All three steps cost, and if the first screen looks “good enough,” noticing never happens. The default then enjoys enormous position bias: objects on that first screen become representatives of the collection; objects not on it do not exist in most users’ model. Commerce that defaults to “comprehensive / popular” keeps new and cold-but-on-topic items out of the silent majority’s sight; news that defaults to relevance parks a breaking latest row under an older, more on-topic long read until someone switches to time.
The default also reverse-shapes what “found it” means. People judge search success by whether the first screen has something clickable; if the default criterion is systematically misaligned with the task type, failure is charged to search, though the right dimension sits in the menu. Because so few change it, A/B-ing the default is almost changing the product’s visible world, not a secondary preference.
Studying it
Measure the baseline rate of changing sort before deciding which dimension the default should land on.
- Paradigms: share of sessions whose sort param differs from default, stratified by task type (known-item, time-critical, price comparison); a swap of default between relevance and time, watching where the silent majority’s successful clicks land. Hearst notes that most people do not touch advanced controls.
- Independent variables: default criterion, control visibility (which modulates change rate but does not cancel the default’s dominance), whether the task has time pressure.
- Dependent variables: sort-change rate, hit rate on the default first screen, rate of missing a time-critical target because sort was not changed.
- Methodological note: a lab task that says “find the latest by time” drives change rates toward one hundred percent and does not represent the field. A low field change rate cannot be read as “everyone likes the default”—the control may be unseen. Separate “did not change sort and succeeded” from “did not change sort and missed the target”; the latter is the cost of a wrong default.
Where it stops holding
Users of expert tools and trading terminals change sort often; the default’s dominance drops, but the default should still be the most common professional task, not an uncommitted “comprehensive.” With no reliable time field, a time default lies to the majority. Personalized defaults that differ per person mean the silent majority no longer shares a visible world, and collaboration or support-by-dictation fails; if personalized defaults remain, the control must show “this is ranked for you.” Under legal or fairness constraints, the default must not systematically deliver paid rows or a class of sellers onto most people’s first screen.
Applying it
- Choose the default from the real mix of task types: time-critical work defaults to time, known-item and topical exploration default to relevance; do not use “comprehensive” to avoid the choice.
- Review the default as a product decision that will determine the silent majority’s field of view, not as a minor preference to tune later.
- When the default conflicts with the second-most-common task, put that second dimension one tap from the collapsed state—but do not count on this to save the majority.
- Verify in logs: if sort-change rate is below ten percent, the default first screen almost is the product’s search result. Sample sessions that never changed sort and failed; if failure is default-dimension versus task-dimension mismatch, changing the default outperforms teaching people to open the menu.