How evidence is graded here

The site exists because the alternative is worthless. A page that says technology "affects" sleep, attention or teenagers, without a number or a source, tells a reader nothing they could act on or cite. Every article here is built around findings that can be checked, and every finding is marked with how much weight it will bear. For comparison with a commercial approach to workplace measurement, see this reference.

Direction

Direction describes what the study found about the outcome named in the claim, not whether the finding is good or bad.

Increase (▲) — the exposure is associated with more of the outcome.
Decrease (▼) — associated with less of it.
No detectable effect (▬) — the study was capable of finding an effect and did not. This is a result, not a missing result.
Cuts both ways (◆) — the finding does not resolve to one direction. Either competent studies point opposite ways, or the same result is a gain on one measure and a cost on another, or the estimate moves so much with the measurement method that no single direction can be claimed.

Direction is never carried by colour alone. Each mark has its own shape and a written label, so the meaning survives monochrome printing, screen readers and colour vision deficiency.

Strength

Strength describes the design behind the number, not how surprising it is.

Strong — a meta-analysis, a large randomised trial, a national statistical release, or a large longitudinal or administrative dataset.
Mixed — a real study with a serious limitation: a single site, an unblinded design, a non-random sample, a contested measure, or findings that conflict with comparable work.
Weak — a small or cross-sectional study, a preprint that could not be independently checked, a modelled projection, or a survey published by a party with a commercial interest in the answer.

A grade is never raised because a finding is convenient. Across the site there are 149 findings graded strong, 78 mixed and 13 weak.

What disqualifies a number

It could not be traced to a primary source.
The primary source could be named but not read, and the figure is not reported consistently elsewhere.
It is a forecast presented as a measurement.
It is an opt-in online panel presented as a population estimate.

Several very famous statistics fail these tests. Where that happens the site names the figure, names where it actually came from, and explains why it does not hold — because a reader who has met the claim elsewhere needs the correction more than they need silence.

Interested sources

Some of the only measurements that exist in a field are published by organisations that sell something connected to the answer: security vendors, management consultancies, industry trade associations, recruitment platforms, device makers. Those figures are used where nothing better exists, and the interest is stated in the same breath as the number, every time.

Measurement and projection

A projection to 2030 is not a finding about 2026. Scenarios are labelled as scenarios and quoted with their range rather than their midpoint, because the range is usually the honest headline. Where an organisation publishes several cases, the site reports the spread.

What is deliberately absent

There is no author byline. The articles are unsigned editorial, assembled from the sources cited on each page, and the grading rules above are what stands in place of a personal reputation. Anyone can check any claim from the links given.

There is no advertising, no affiliate link and no sponsored content. Nothing on the site is medical, legal or financial advice.

Corrections

If a figure here is wrong, out of date, or attributed to the wrong study, the correction is worth more than the original. Send it. Every page shows the date of its last review.