Body text size
type:Smallest body text is 12px (base ~17px). below the ~16px readable floor.
FixKeep body copy ≥15-16px; reserve sub-13px for fine print only.
01
Homepage / Lander
▲ strongest
4 tactics
saas · Aug 5, 2026
devin.ai
73/100 Partial
Scores higher than 88% of 1,391 multi-step funnels we have measured. Among the 1,074 saas funnels in that group, higher than 85%.
How this was measured
Measured by Roentgen 1.19 on 2026-08-05. A capture is read once and its measurements freeze there, so the stamp says which ruleset read this one.
Coverage is detection coverage, not a verdict — it counts the conversion tactics we could prove are present, out of the 11 that apply to a saas funnel. Per device: desktop 75 · mobile 70.
Walk 34 links on the entry page, 27 walkable. 2 matched a funnel step; we captured 3 pages in all, this one included.
What we recorded
Progress
2 scans
this funnel’s coverage score across measured re-scans
measuredextracted directly from the live page. a fact, not an opinion
Unchanged across 2 scans
Same 6 conversion tactics detected since 2026-08-03.
- 6 tactics 2026-08-03
- 6 tactics 2026-08-05
The funnel, frame by frame
3 steps walkedIn their own words
29 quotedthe entry page’s own title, headline, buttons and prices — quoted, never paraphrased
measuredextracted directly from the live page. a fact, not an opinion- Page title
Devin | The AI Software Engineer
- Meta description
Devin is an AI coding agent and software engineer that helps developers build better software faster. Parallel cloud agents for serious engineering teams.
- Headline
Devin, the AI software engineer
- Section headings
How Nubank refactors millions of lines of code to improve engineering efficiency with Devin
Overview
The Problem
The Decision: an army of Devins to tackle subtasks in parallel
The Solution: Custom ETL Migration Devin
Build with Devin
Use cases
Learn & work together
+8 more, not shown- Buttons and links
Try Devin
Contact sales
Hear from our customers
Approved new knowledge: When working in the backend repo
Rejected new knowledge: When working in the backend repo
Product
Solutions
Resources
Get started
Download
What we found
6 tacticsconversion tactics we could prove are present, with the on-page text that evidences each
measuredextracted directly from the live page. a fact, not an opinionOffer clarity1 of 1
Makes the pricing ladder explicit
free, $80/mo, $40/mo, $0/mo
Action path2 of 4
Offers a free trial or free tier
Free plan
Offers a demo or docs entry point
Documentation
Trust & proof2 of 2
Shows customer logos or a trusted-by claim
logo wall (8 distinct marks)
Puts a number on adoption
1,000 engineers
Friction reduction2 of 2
Answers buyer objections
Frequently Asked Questions
Offers a free trial or free tier
also shown under Action path
Value reinforcement1 of 2
Puts a number on adoption
also shown under Trust & proof
What we didn’t find
5 of 11 applicableapplicable tactics we checked for and did not see on any step we reached
measuredextracted directly from the live page. a fact, not an opinionWhat we didn’t see5 of 11 applicable
Press / "as seen in"
508 of the 1,960 funnels we checked
Lead-capture form
342 of the 1,393 funnels we checked
Us-vs-them comparison
332 of the 1,968 funnels we checked
Proof above the fold
314 of the 1,968 funnels we checked
Scarcity / urgency copy
101 of the 1,960 funnels we checked
Absence is a measurement, not a recommendation: we do not know why a tactic is missing, whether it belongs here, or whether adding it would change anything. Where a walk stopped early, a tactic living deeper would also read as not detected.
How well it’s built
18 run · 6 flaggedcontrast, mobile, typography, accessibility, weight — each with what would fix it
measuredextracted directly from the live page. a fact, not an opinion6 of 17 measured checks need work1 could not be measured — not counted either way
Text contrast (WCAG AA)
contrast:75% of measurable text passes WCAG AA. 4+ runs fail. 32% of text was unmeasurable (over images/gradients). not counted either way.
worst: “GitHub” at 1.05:1 (34px, #fcfcfc on #f7f6f5). needs 3:1
FixRaise text/background contrast to ≥4.5:1 (≥3:1 for large text), especially on CTAs and body copy.
Named controls
a11y:2 buttons/links have no accessible name (icon-only, no aria-label).
FixAdd aria-label to icon-only buttons and links.
Oversized images
weight:9 images are shipped at more than 2x the width shown.
“hero_usecase_2.webp” is 870px wide, displayed at 329px
FixServe responsive sizes (srcset/sizes) matching the rendered width.
Landmark regions
structure:No <main> or <footer> or <header> landmark. screen-reader users cannot jump between regions.
FixWrap the page in <header>, <main> and <footer> (or matching ARIA roles).
Tap-target size (mobile)
mobile:56 of 61 interactive targets are under 44×44px on mobile.
smallest: “Devin home” at 28×28px
FixGive tappable controls a ≥44×44px hit area. the touch minimum.
Worth knowing8
Line length
6 paragraphs run past ~95 characters per line.
FixCap measure at ~60-75ch for comfortable reading.
type:Form labels
2 form fields rely on placeholder text as the only label.
FixBack placeholders with a real <label>. placeholder text vanishes on input.
a11y:Heading levels
3 headings skip a level (e.g. <h2> followed by <h4>). Screen-reader users navigate by this outline.
FixStep heading levels one at a time; use CSS for size, not a deeper level.
a11y:Skip link
No same-page “skip to content” link was found. Keyboard users tab through the whole nav on every page.
FixAdd a first-in-DOM <a href="#main"> that is visible on focus.
a11y:Label in name
6 controls have an aria-label that does not contain its visible text, so voice-control users cannot activate it by saying what they see.
FixMake the accessible name start with the visible label text.
a11y:Render-blocking resources
6 stylesheets/scripts in <head> block first paint.
FixDefer non-critical JS (defer/async) and inline only critical CSS.
weight:Keyboard reachability
937 elements look clickable but cannot be reached by keyboard.
FixUse <button>/<a>, or add role + tabindex="0" and a key handler.
structure:Autocomplete hints
No autocomplete attributes across 2 fields. browsers cannot autofill.
FixAdd autocomplete="email"/"tel"/"name" to identity fields.
forms:Not measurable1
Page weight
At least 2.5 MB across 63 measurable resources, but 93 cross-origin resources did not expose a size. The true weight cannot be determined from the page.
weight:Clear3
Brand kit
5 colour roles
palette by on-page usage · type rendered live
measuredextracted directly from the live page. a fact, not an opinionfull palette ordered by on-page usage %
primary#1d2e56
accent#1d2e56
bg#f7f6f5
surface#efefef
text#121111
See the roles applied
accent
Body copy in their text role, on a surface card.
the roles, doing their job
The quick brown fox64px
Jumps over the lazy dog34px
A subheading in their voice27px
Body copy set at their measured paragraph size. This is how their product reads to a first-time visitor. same family, same scale, same weight.
- List item one, at body scale
- List item two
- List item three
Type scale, weights & families
SIZE SCALE
11pxAa
13pxAa
14pxAa
15pxAa
16pxAa
17pxAa
21pxAa
24pxAa
26pxAa
27pxAa
34pxAa
64pxAa
WEIGHTS
Aa400
Aa500
Aa700
FAMILIES
NB International Pro
rendered in their own type system · webfont not loaded. system fallback if uninstalled
Measured, found none
2 of 9 layerslayers we ran on this funnel that came back empty
measuredextracted directly from the live page. a fact, not an opinionElements — the atomic pieces — price anchors, ratings, guarantees — with their verbatim on-page text
Playbook match — which funnel play this page runs — measured, evidence-quoted
We ran these passes on the pages we captured and found nothing to show. That is a measurement of this funnel, not a recommendation — and it is a different statement from a layer we never ran, which does not appear here at all.
X-ray a funnel
