Field study ·
What 246 SaaS homepages showed our Growth Scan
We pointed the Growth Scan at the homepages of 246 SaaS companies listed on Wikidata. It checks seven public-page facts, from the page title to horizontal overflow. Most homepages passed all seven. The problems it did find were mostly structural, and some disappeared once a full browser loaded every resource.
- got no Growth Scan finding
- 76.8%
- 189 of 246 homepages passed all seven rules.
- show more than one H1
- 6.9%
- 17 homepages, and the same 17 in a full-browser recheck.
- ask for a demo in the first screen
- 43.6%
- 78 of 179 English homepages show a demo request above 1,200 px.
- median homepage length
- 9,298 px
- About 7.7 window heights at 1,440 × 1,200 px.
One finding per page
What the Growth Scan flags first
The Growth Scan reports one finding per page: the first of its seven rules that fails, in a fixed order. Search basics come first because they are quick to fix and easy to verify. This table shows what the scan would have emailed for each homepage.
Two structural rules lead: horizontal overflow (17 homepages) and more than one H1 (15). Missing titles and descriptions are rare in this sample.
| First finding | Homepages | Share | 95% interval |
|---|---|---|---|
| Missing page titleNo non-empty title element in the rendered page. | 3 of 246 | 1.2% | 0.4–3.5% |
| Missing meta descriptionNo non-empty meta description. | 7 of 246 | 2.8% | 1.4–5.8% |
| No visible H1No visible top-level heading. | 8 of 246 | 3.3% | 1.7–6.3% |
| More than one H1Two or more distinct visible H1 texts. | 15 of 246 | 6.1% | 3.7–9.8% |
| No call to actionNo visible button, and no visible link with an action word such as start, try, book, or contact. | 1 of 246 | 0.4% | 0.1–2.3% |
| Long first-step formA visible form with 8 or more fields, or 6 or more required fields. | 6 of 246 | 2.4% | 1.1–5.2% |
| Horizontal overflowContent more than 8 px wider than the 1,440 px window. | 17 of 246 | 6.9% | 4.4–10.8% |
| No findingAll seven rules passed. | 189 of 246 | 76.8% | 71.2–81.7% |
Recheck
Each rule on its own, in two renders
A first finding hides the rules behind it, so we also counted every rule separately. Then we loaded the same pages again in a full browser that allows every host and request type.
The Growth Scan render is deliberately bounded: it loads only same-host HTTPS resources, makes no API calls, and stops at 150 requests or 8 MB. The comparison below covers the 244 homepages that rendered both ways.
| Rule | Growth Scan render | Full render | In both |
|---|---|---|---|
| Missing page title | 3 (1.2%) | 0 (0.0%) | 0 |
| Missing meta description | 10 (4.1%) | 8 (3.3%) | 8 |
| No visible H1 | 11 (4.5%) | 6 (2.5%) | 6 |
| More than one H1 | 17 (7.0%) | 17 (7.0%) | 17 |
| No call to action | 4 (1.6%) | 2 (0.8%) | 1 |
| Long first-step form | 8 (3.3%) | 12 (4.9%) | 8 |
| Horizontal overflow | 19 (7.8%) | 11 (4.5%) | 8 |
| No finding | 187 (76.6%) | 195 (79.9%) | – |
- More than one H1 is the most stable result. The same 17 homepages failed it in both renders.
- Missing titles came from the render, not the pages. With every host allowed, all 3 pages without a title in the Growth Scan render showed one. The same effect explains 5 of the 11 pages without a visible H1.
- Overflow depends on the render. 8 homepages overflowed in both, 11 only in the Growth Scan render, and 3 only in the full render.
- Long forms go the other way. 4 long first-step forms appeared only in the full render, because they load from another host.
First screen
What the first screen asks visitors to do
We read the action words in visible buttons and links above 1,200 px on the 179 homepages that declare English. A page can use several phrases, so the shares add up to more than 100%.
86.6% of these homepages (155 of 179) show at least one action phrase in the first screen.
| Phrase family | Homepages | Share | 95% interval |
|---|---|---|---|
| Demo requestBook, request, get, or schedule a demo | 78 of 179 | 43.6% | 36.5–50.9% |
| Free or trialFree trial, start free, try for free | 57 of 179 | 31.8% | 25.5–39.0% |
| Contact or salesContact us, talk to sales | 52 of 179 | 29.1% | 22.9–36.1% |
| Get startedThe exact phrase “get started” | 27 of 179 | 15.1% | 10.6–21.1% |
| Sign upSign up or signup | 17 of 179 | 9.5% | 6.0–14.7% |
| Self-serve demoWatch, try, or start a demo | 15 of 179 | 8.4% | 5.1–13.4% |
| Learn moreThe exact phrase “learn more” | 14 of 179 | 7.8% | 4.7–12.7% |
| Download or installApps and extensions | 5 of 179 | 2.8% | 1.2–6.4% |
Sales-led, self-serve, or both
The first-screen ask splits almost evenly: 44 English homepages lead only with a demo request and 44 only with a self-serve start. About one in five offers both. Of the 57 pages with neither, 22 show a contact or talk-to-sales link, and 24 show no action phrase above 1,200 px.
| First-screen ask | Homepages | Share | 95% interval |
|---|---|---|---|
| Demo request onlyA demo request, and no free, trial, get-started, or sign-up phrase | 44 of 179 | 24.6% | 18.9–31.4% |
| Self-serve start onlyA free, trial, get-started, or sign-up phrase, and no demo request | 44 of 179 | 24.6% | 18.9–31.4% |
| BothA demo request and a self-serve start | 34 of 179 | 19.0% | 13.9–25.4% |
| NeitherNo demo request and no self-serve start | 57 of 179 | 31.8% | 25.5–39.0% |
Structure
Forms, length, and search basics
These facts come from the same Growth Scan render of all 246 homepages. They describe the pages; they say nothing about how the pages perform.
- Visible forms
- 35.4% of homepages (87 of 246) show a visible form. The median visible form has 1 field; the middle half has 1 to 4 fields (124 forms). Only 8 homepages show a form with 8 or more fields or 6 or more required fields.
- Page length
- At a 1,440 px window, the median homepage is 9,298 px long. The middle half runs from 7,188 to 12,887 px.
- Main headings
- 88.6% show exactly one visible H1, 4.5% show none, 3.3% show two, and 3.7% show three or more.
- Titles and descriptions
- The median title has 56 characters (middle half 48 to 61, 243 titles). The median meta description has 151 characters (middle half 139 to 159, 236 descriptions).
- Canonical and indexing
- 93.1% (229 of 246) declare a canonical URL. One homepage told search engines not to index it.
- Rules per page
- 189 homepages failed no rule, 48 failed one, 6 failed two, and 3 failed four.
Method
How we built the sample and ran the scans
On 8 November 2026 at 18:19 UTC we asked Wikidata for every item whose industry (P452) is software as a service (Q1254596) and that lists an official website (P856). We visited each website, followed redirects to the final URL, and kept one homepage per site: the root path, or a language root such as /en (15 homepages).
The scans used detector public-scan-detector-v1 in headless Chromium at 1,440 × 1,200 px with JavaScript on. The Growth Scan render waits for the page structure, up to 2 seconds of network quiet, and 150 ms more, as the product does. It ran from 18:22 to 18:25 UTC; the full-browser recheck, with the same viewport, waits, extraction code, and rules, ran from 18:28 to 18:31 UTC.
| Companies with the SaaS industry tag and a website | 322 |
|---|---|
| Dissolved, removed | −1 |
| Websites visited | 321 |
| No answer (DNS, TLS, or timeout) | −20 |
| No HTTPS version | −5 |
| Homepage disallowed by robots.txt | 0 |
| Refused by the scanner's network-safety check | −3 |
| Error page served to the headless browser | −21 |
| Over the scanner's limit of 150 requests or 8 MB | −15 |
| Scanned | 257 |
| Listed website was a product, about, or for-sale page | −10 |
| Same site listed twice | −1 |
| Homepages analysed | 246 |
Conduct. Each website got one robots.txt request, two HTML requests, and two browser visits. We followed each site’s robots.txt rules for all crawlers (the * group); none disallowed a homepage. We did not log in, submit forms, or keep screenshots.
Language. We read the lang attribute of each homepage: English 179, French 22, Spanish 10, German 7, Portuguese 5, 16 pages in 11 other languages, and 7 without a declared language.
Sample. Wikidata records a country for 199 of the companies: United States 68, France 28, United Kingdom 12, Canada 8, India 8, Australia 5, Brazil 5, Singapore 5, Denmark 5, and 55 in 30 other countries. Where a founding year is recorded (213 companies), the median is 2017; the middle half runs from 2011 to 2024.
Intervals. Shares use Wilson 95% intervals. Percentages are rounded to one decimal place.
Limits
- The sample is not the SaaS market. It holds companies notable enough for a Wikidata entry with this industry tag, which favors established companies and some countries: France supplies 28 of 246 homepages.
- We scanned one desktop render of each homepage at one moment. Mobile layouts, campaign landing pages, and signed-in pages are not covered.
- 64 of the 321 websites could not be scanned, often because of bot protection or the scanner's limits. The results describe the pages that rendered.
- The rules record facts, not quality. One H1 does not make a good headline, and a call-to-action candidate is any visible button or a link with an English action word.
- We have no visitor, conversion, or revenue data for these sites. Nothing here shows that a rule affects conversion or that a fix would help.
- The first-screen phrase families use English keywords and can overlap. Homepages in other languages are left out of that section.
Corrections: write to [email protected] with the number and the reason. We log material corrections on this page with the date.
Run the same seven rules on your page
Growth Scan checks one public page with the rules above and emails one finding, or tells you that none applies. It is free and needs no account.