THE DATA STUDY · 93 PAGES SCANNED
We scanned 93 famous SaaS landing pages. 81% hide their hottest zone outside the hero.
Same model, same day, same viewport: the landing pages of well-known SaaS products, run through a peer-reviewed attention model (UNISAL) and scored with the exact formula behind our public scanner. Here is everything the data says, including the names.
44.1
AVERAGE SCORE /100
67%
SCORE BELOW 50
0
PAGES ABOVE 80, ZERO
63%
WEAK VISUAL HIERARCHY (<30)
The distribution: mediocrity is the norm
Median: 41/100. Not a single page reached 80. These are companies with design teams.
81% of pages: the eye lands somewhere you didn't design for
The single most consistent finding: on 81% of pages, the predicted primary gaze zone is NOT in the top band of the screen, the hero everyone spends their design budget on. The eye goes where contrast, faces and isolation pull it. Rarely where the roadmap said.
TOP-LEFT
8%
TOP-CENTER
12%
TOP-RIGHT
0%
CENTER-LEFT
19%
CENTER
55%
CENTER-RIGHT
6%
BOTTOM-LEFT
0%
BOTTOM-CENTER
0%
BOTTOM-RIGHT
0%
19% start in the hero band81% start outside it
The names
BEST OF THE BATCH
- pandadoc.com78/100
- lemonsqueezy.com78/100
- softr.io74/100
- clickfunnels.com73/100
- semrush.com68/100
ROUGHEST OF THE BATCH
- slidebean.com21/100
- tome.app22/100
- workday.com23/100
- airtable.com23/100
- convertkit.com25/100
A low score isn't a verdict on the product. It measures how a first-time eye triages the page. Big brands survive weak heroes; your startup probably can't.
Methodology (fully reproducible)
- 01Sample: well-known SaaS landing pages, scanned the same day at 1440×900 desktop viewport. Pages behind aggressive bot protection failed and are excluded. N is the count of successful scans, nothing else.
- 02Model: UNISAL, a peer-reviewed saliency model trained on public eye-tracking datasets (SALICON, MIT300). Deterministic: the same screenshot always produces the same map.
- 03Score: 0.35 × attention intensity (×4, capped) + 0.35 × visual hierarchy + 0.30 × focus clarity, the exact formula of our public scanner. Zones use a 3×3 grid; 'hero' = the top band.
- 04Bias note: we sell the scanner. Which is why every number is recomputable: scan any page on the list yourself and compare.
CITE THIS STUDY, AND COME BACK NEXT YEAR
This is the 2026 edition of the Attention Barometer: we re-scan the same 93 pages every July with the same model and publish the deltas: who improved, who regressed. Cite it as: “Heatpoints Attention Barometer 2026, N=93, predictive saliency (UNISAL), desktop 1440×900.” Link this page; the data and method stay up permanently.
THE OBVIOUS QUESTION
Where would YOUR page land in this distribution?
Free, instant, no account needed.