# Methodology: Half of B2B website tools are gone within 4 years

By Cristian Frunze, Founder of Kenso · Nov 15, 2026 · [Kenso Research](/research) No. 24 · Websites and pricing

The full methodology behind [Half of B2B website tools are gone within 4 years](/research/website-tool-lifespan.md): the data, every method, every number and every limitation. The same text as the [methodology PDF](/research/website-tool-lifespan.pdf).

Data through the web archive's 2026 crawls.

## Question

How fast do B2B website tools (chat, analytics, marketing automation, forms, scheduling, consent) spread from early adopters to the mainstream, how fast do they decline, and which tools replaced which?

## Data

- **Companies.** The shared homepage panel built for Kenso Research: a seeded simple random sample of 10,000 records from Kenso internal data company records (Kenso's dataset of public company records) in seven software and internet industries, headquartered in the US, UK, Canada, Australia, Ireland or New Zealand, founded 1950-2018, with 11 to 5,000 employees today. Platform domains and domains shared by several records were removed before sampling.
- **Homepages.** One web archive capture of each company's homepage per calendar year, 2014-2026: the crawl nearest mid-year, with a second crawl in the same year as a fallback when the first had no capture. The captures were fetched once for every study that uses the panel; this study read the stored archived page records and fetched nothing.
- **Usable captures.** HTTP 200, not a parked or for-sale page, and at least 50 words of visible text. 55,285 usable company-years in total:

| Year | Usable homepages |
|---|---|
| 2014 | 2,570 |
| 2015 | 2,689 |
| 2016 | 3,114 |
| 2017 | 4,103 |
| 2018 | 4,550 |
| 2019 | 4,692 |
| 2020 | 4,886 |
| 2021 | 5,083 |
| 2022 | 5,103 |
| 2023 | 4,899 |
| 2024 | 4,896 |
| 2025 | 4,437 |
| 2026 | 4,263 |

Early years have fewer usable captures because Kenso's web archive captured fewer of these sites then.

## Tool detection

1. **Kenso's offline technology detector** (detector version 1.2.0, a fingerprint corpus of 7,539 technologies and 7,723 patterns) ran on every capture with HTML comments removed. The stock command line runs each pattern inside a sandbox with a timer, which took about 28 seconds per page; we called the same detector functions with that per-pattern timer switched off, which leaves matching unchanged. On 48 random captures both ways returned identical results (48 of 48 documents, 369 technology hits each).
2. **Live-signature check.** Offline, the detector sees only the static page. Chat widgets and many marketing tools are loaded by a short inline script, which the detector's patterns mostly do not match (Drift, for example, has only JavaScript-runtime fingerprints). For 67 common tools we matched each vendor's loader address in script `src` attributes and in the code of executable inline scripts.
3. **Counting rule (stale tags).** A tool counts on a capture when the live-signature check fires, or when the detector reports evidence that only a live loader produces: a script source, a response header or a cookie. Mentions in HTML comments, page text, ordinary links (such as a link to a Calendly page) and tools only implied by another tool do not count.
4. **Categories.** From the detector's categories: live chat, cookie compliance (called cookie consent here), appointment scheduling, form builders and surveys (one category, forms and surveys), marketing automation, tag managers and analytics. A tool in several of these gets one, in that order. Following the detector, any Google tag (`gtag.js`) counts as Google Analytics and also as a tag manager. Universal Analytics and Google Analytics 4 are told apart by their tracking ids.
5. **Focal tools.** Tools on at least 1% of usable homepages in at least one year: 40 tools.

## Validation

On the 2026 crawl we compared detection with the technology field in Kenso's company website index, built from the same September 2026 crawl, for 3,448 companies with a usable 2026 homepage in both. That index can include a domain's other pages (median two pages) and evidence a static page does not show, so it is a cross-check rather than ground truth. Positive agreement is 2 x both / (ours + theirs).

| Detection | Tool | This study | Website index | Both | Agreement |
|---|---|---|---|---|---|
| detector, all evidence | All tools in this table (pooled) | 6,430 | 5,277 | 4,823 | 0.82 |
| detector, all evidence | Google Tag Manager | 2,206 | 1,605 | 1,485 | 0.78 |
| detector, all evidence | Google Analytics | 1,422 | 1,394 | 1,387 | 0.99 |
| detector, all evidence | Site Kit | 99 | 321 | 99 | 0.47 |
| detector, all evidence | Contact Form 7 | 325 | 320 | 306 | 0.95 |
| detector, all evidence | Linkedin Insight Tag | 312 | 307 | 305 | 0.99 |
| detector, all evidence | Cloudflare Browser Insights | 291 | 290 | 288 | 0.99 |
| detector, all evidence | Google Ads Conversion Tracking | 256 | 257 | 254 | 0.99 |
| detector, all evidence | HubSpot | 636 | 225 | 205 | 0.48 |
| detector, all evidence | CookieYes | 205 | 192 | 190 | 0.96 |
| detector, all evidence | Gravity Forms | 102 | 122 | 101 | 0.90 |
| detector, all evidence | Cookiebot | 67 | 69 | 67 | 0.99 |
| detector, all evidence | Complianz | under 50 | 62 | under 50 | 0.59 |
| detector, all evidence | OneTrust | 61 | 59 | 58 | 0.97 |
| detector, all evidence | Ahrefs | 52 | 54 | 52 | 0.98 |
| detector, all evidence | Facebook Pixel | 250 | under 50 | under 50 | 0.00 |
| detector, all evidence | Sirvoy | 120 | under 50 | under 50 | 0.00 |
| study counting rule | All tools in this table (pooled) | 6,223 | 5,290 | 3,931 | 0.68 |
| study counting rule | Google Tag Manager | 2,229 | 1,605 | 1,436 | 0.75 |
| study counting rule | Google Analytics | 1,562 | 1,394 | 1,388 | 0.94 |
| study counting rule | Site Kit | under 50 | 321 | under 50 | 0.00 |
| study counting rule | Contact Form 7 | 325 | 320 | 306 | 0.95 |
| study counting rule | Linkedin Insight Tag | under 50 | 307 | under 50 | 0.01 |
| study counting rule | Cloudflare Browser Insights | 291 | 290 | 288 | 0.99 |
| study counting rule | Google Ads Conversion Tracking | under 50 | 257 | under 50 | 0.01 |
| study counting rule | HubSpot | 589 | 225 | under 50 | 0.12 |
| study counting rule | CookieYes | 212 | 192 | 190 | 0.94 |
| study counting rule | Gravity Forms | 85 | 122 | 84 | 0.81 |
| study counting rule | Cookiebot | 69 | 69 | 68 | 0.99 |
| study counting rule | Complianz | 76 | 62 | 62 | 0.90 |
| study counting rule | OneTrust | 70 | 59 | 59 | 0.91 |
| study counting rule | Ahrefs | under 50 | 54 | under 50 | 0.00 |
| study counting rule | Leadfeeder | 58 | under 50 | under 50 | 0.00 |
| study counting rule | Hotjar | 135 | under 50 | under 50 | 0.00 |
| study counting rule | Intercom | 74 | under 50 | under 50 | 0.00 |
| study counting rule | Microsoft Clarity | 172 | under 50 | under 50 | 0.00 |
| study counting rule | HubSpot Forms | 274 | under 50 | under 50 | 0.00 |

Like for like (detector, all evidence types), agreement is 0.82 over the tools in the table, and 0.94 to 0.99 for Google Analytics, Contact Form 7, CookieYes, Cookiebot and OneTrust. Two gaps are explained: the website index lists HubSpot mostly on sites hosted on HubSpot's CMS (199 of its 225 HubSpot sites also carry the CMS key), while this study counts the HubSpot tracking script; and tools such as the LinkedIn Insight Tag are found there from evidence a static homepage does not hold. The study's own counting rule agrees less (0.68) by design, because it drops HTML-only matches and adds inline loaders.

## Methods

- **Adoption share:** the share of usable homepages in a year with the tool, or with any tool in the category, with 95% Wilson intervals.
- **Bass diffusion:** Bass's (1969) discrete least-squares form, the yearly change in share regressed on last year's share and its square, giving market potential m, innovation p and imitation q. Reported only where all three come out positive.
- **Rise and decline:** years from half of the peak share to the peak, and from the peak back to half, by linear interpolation between yearly captures.
- **Switching:** in consecutive years where a company has usable captures in both, a tool gone and another tool of the same category new.
- **Churn:** among company-years with the tool and a usable capture the next year, the share without the tool the next year.
- **Survival:** Kaplan-Meier curves for new installs (a tool appearing after the company's first usable capture). An install ends at the first later usable capture without the tool; a year with no usable capture is bridged when the tool is present on both sides; installs still present at a company's last capture are censored. Durations count from the first capture with the tool to the first capture without it, so they are upper bounds by up to a year.
- **Suppression:** no published count under 50 companies; every published share is over at least 50 companies.

## Results

### Adoption by category

| Year | Usable homepages | Analytics | Tag managers | Marketing automation | Forms and surveys | Live chat | Cookie consent | Scheduling |
|---|---|---|---|---|---|---|---|---|
| 2014 | 2,570 | 71.9% | 3.4% | 10.1% | 10.7% | 6.0% | 0.3% | 0.0% |
| 2015 | 2,689 | 73.1% | 6.5% | 13.0% | 14.1% | 7.1% | 0.4% | 0.0% |
| 2016 | 3,114 | 72.9% | 11.1% | 14.6% | 18.4% | 8.7% | 0.8% | 0.0% |
| 2017 | 4,103 | 69.7% | 13.0% | 15.3% | 22.1% | 10.5% | 1.0% | 0.1% |
| 2018 | 4,550 | 67.7% | 27.0% | 18.5% | 24.3% | 12.9% | 4.5% | 0.2% |
| 2019 | 4,692 | 66.7% | 40.4% | 20.4% | 26.3% | 14.5% | 6.8% | 0.7% |
| 2020 | 4,886 | 62.2% | 46.6% | 20.2% | 26.8% | 14.1% | 8.0% | 0.9% |
| 2021 | 5,083 | 59.8% | 55.1% | 21.3% | 29.5% | 12.7% | 8.8% | 1.1% |
| 2022 | 5,103 | 59.9% | 60.4% | 22.5% | 27.5% | 11.9% | 10.5% | 1.3% |
| 2023 | 4,899 | 59.0% | 63.8% | 22.6% | 25.6% | 10.6% | 12.1% | 1.8% |
| 2024 | 4,896 | 57.9% | 65.4% | 22.4% | 25.3% | 9.5% | 14.1% | 2.0% |
| 2025 | 4,437 | 56.6% | 65.0% | 22.6% | 23.7% | 8.7% | 15.0% | 1.7% |
| 2026 | 4,263 | 55.9% | 63.5% | 21.7% | 21.0% | 7.8% | 16.8% | 1.5% |

### Survival of new installs

Pooled over the focal tools: 18,951 new installs, 9,301 removals. 79.3% survive to the next yearly capture, 53.0% to year 3 and 44.3% to year 4; the median is 4 years. Median by category: analytics 4, cookie consent 3, forms and surveys 3, live chat 3, marketing automation 3, scheduling 3, tag managers 8 years.

### Tools

| Tool | Category | Peak share (year) | 2026 share | Rise / decline years | Annual churn | Median years (new installs) | Bass fit |
|---|---|---|---|---|---|---|---|
| Google Analytics | Analytics | 71.1% (2015) | 45.4% | before window / beyond window | 12% | 5 | |
| Cloudflare Browser Insights | Analytics | 8.7% (2026) | 8.7% | 0.8 / beyond window | 14% | | |
| Hotjar | Analytics | 6.2% (2019) | 4.0% | 2.9 / beyond window | 29% | 2 | p 0.256, q 0.45, m 5.5% |
| Microsoft Clarity | Analytics | 4.6% (2026) | 4.6% | 2.4 / beyond window | 30% | 3 | p 0.015, q 0.68, m 6.1% |
| MonsterInsights | Analytics | 4.2% (2019) | 1.4% | 1.6 / 6.4 | 31% | 3 | p 0.203, q 0.05, m 3.2% |
| Crazy Egg | Analytics | 2.5% (2016) | 0.5% | 1.3 / 4.7 | 39% | 2 | |
| Zoominfo | Analytics | 2.3% (2023) | 0.7% | 2.0 / 1.7 | 31% | 2 | p 0.068, q 0.62, m 1.7% |
| Leadfeeder | Analytics | 2.1% (2024) | 1.6% | 3.9 / beyond window | 28% | 3 | p 0.047, q 1.13, m 1.9% |
| ExactMetrics | Analytics | 1.6% (2020) | 0.1% | 0.8 / 2.3 | 43% | 2 | p 0.080, q 0.30, m 0.9% |
| Lead Forensics | Analytics | 1.4% (2018) | 0.4% | 1.3 / 5.0 | 22% | 4 | p 0.243, q 0.05, m 0.9% |
| Mixpanel | Analytics | 1.3% (2015) | 0.5% | before window / 5.2 | 29% | 2 | |
| Statcounter | Analytics | 1.2% (2014) | 0.1% | before window / 1.7 | 20% | | |
| CookieYes | Cookie consent | 5.7% (2026) | 5.7% | 4.4 / beyond window | 27% | 3 | p 0.009, q 0.37, m 7.6% |
| Cookie Consent (open source, Osano) | Cookie consent | 2.5% (2019) | 0.6% | 1.2 / 5.0 | 24% | 3 | |
| Complianz | Cookie consent | 2.1% (2026) | 2.1% | 3.0 / beyond window | 22% | 4 | p 0.011, q 1.01, m 1.9% |
| Cookiebot | Cookie consent | 2.0% (2026) | 2.0% | 5.2 / beyond window | 21% | 4 | p 0.040, q 0.32, m 2.3% |
| Cookie Notice | Cookie consent | 2.0% (2020) | 0.7% | 2.0 / 0.9 | 44% | 2 | p 0.095, q 0.54, m 1.3% |
| OneTrust | Cookie consent | 2.0% (2026) | 2.0% | 5.0 / beyond window | 17% | 4 | p 0.003, q 0.93, m 1.9% |
| Contact Form 7 | Forms and surveys | 19.4% (2019) | 9.4% | before window / 6.8 | 21% | 4 | |
| HubSpot Forms | Forms and surveys | 9.5% (2021) | 7.9% | 0.9 / beyond window | 29% | 2 | |
| Gravity Forms | Forms and surveys | 3.1% (2025) | 2.6% | 8.0 / beyond window | 26% | 3 | p 0.115, q 0.34, m 2.9% |
| WPForms | Forms and surveys | 2.7% (2024) | 0.5% | 2.9 / 0.7 | 46% | 2 | |
| Marketo Forms | Forms and surveys | 1.1% (2020) | 0.8% | 5.3 / beyond window | 28% | 3 | |
| Intercom | Live chat | 4.7% (2019) | 2.2% | 2.3 / 5.5 | 25% | 3 | p 0.245, q 0.15, m 3.2% |
| Olark | Live chat | 2.6% (2015) | 0.2% | before window / 2.0 | 34% | 2 | |
| Drift | Live chat | 2.6% (2020) | 0.2% | 2.5 / 2.7 | 38% | 2 | p 0.143, q 0.04, m 1.4% |
| Tawk.to | Live chat | 2.2% (2020) | 1.1% | 2.7 / beyond window | 23% | 3 | p 0.145, q 0.47, m 1.7% |
| Zendesk Chat | Live chat | 2.2% (2016) | 0.8% | before window / 7.8 | 25% | 3 | |
| LiveChat | Live chat | 1.2% (2018) | 0.3% | 3.9 / 2.8 | 30% | 2 | |
| Zoho SalesIQ | Live chat | 1.2% (2021) | 1.0% | 3.4 / beyond window | 23% | 4 | p 0.105, q 0.83, m 1.1% |
| HubSpot | Marketing automation | 16.7% (2026) | 16.7% | 8.8 / beyond window | 21% | 4 | p 0.051, q 0.31, m 16.8% |
| Marketo | Marketing automation | 3.0% (2015) | 0.6% | before window / 2.5 | 28% | 3 | |
| Salesforce Marketing Cloud Account Engagement | Marketing automation | 2.5% (2016) | 1.1% | before window / 9.3 | 28% | 3 | |
| BDOW | Marketing automation | 1.7% (2017) | 0.0% | 1.6 / 0.8 | 48% | 2 | |
| Sumo | Marketing automation | 1.7% (2017) | 0.0% | 1.6 / 0.8 | 48% | 2 | |
| ActiveCampaign | Marketing automation | 1.4% (2022) | 0.9% | 2.9 / beyond window | 29% | 2 | p 0.120, q 0.42, m 1.1% |
| MailChimp | Marketing automation | 1.3% (2021) | 0.7% | 3.5 / beyond window | 29% | 3 | p 0.165, q 0.22, m 1.1% |
| Act-On | Marketing automation | 1.2% (2015) | 0.2% | before window / 3.7 | 32% | 3 | |
| Calendly | Scheduling | 1.5% (2024) | 1.2% | 2.9 / beyond window | 28% | 3 | p 0.034, q 0.70, m 1.4% |
| Google Tag Manager | Tag managers | 65.4% (2024) | 63.5% | 5.6 / beyond window | 9% | 8 | p 0.024, q 0.63, m 64.5% |

BDOW and Sumo are the same product under two names in the detector corpus and appear as two rows.

### Switching

| Category | Removals | Replaced in the same year | Replacement installs | To point tools | To HubSpot | To Google | To Salesforce |
|---|---|---|---|---|---|---|---|
| Scheduling | 120 | 2% | under 50 | | | | |
| Marketing automation | 2,410 | 11% | 249 | 47% | 38% | 0% | 14% |
| Tag managers | 1,646 | 0% | under 50 | | | | |
| Forms and surveys | 2,685 | 11% | 306 | 61% | 39% | 0% | 0% |
| Cookie consent | 863 | 14% | 119 | 100% | 0% | 0% | 0% |
| Analytics | 5,533 | 11% | 558 | 80% | 0% | 20% | 0% |
| Live chat | 1,439 | 12% | 176 | 99% | 0% | 0% | 1% |

Tool-to-tool flows of at least 50 companies:

| Category | From | To | Companies |
|---|---|---|---|
| Forms and surveys | Contact Form 7 | HubSpot Forms | 81 |
| Analytics | Google Analytics | Hotjar | 52 |

Most common replacement installs (at least 50 companies):

| Category | Tool | Replacement installs |
|---|---|---|
| Forms and surveys | HubSpot Forms | 120 |
| Analytics | Google Analytics | 111 |
| Analytics | Hotjar | 99 |
| Marketing automation | HubSpot | 95 |
| Analytics | Cloudflare Browser Insights | 63 |
| Forms and surveys | Contact Form 7 | 57 |

### Universal Analytics and GA4

Universal Analytics peaked at 73.6% of usable homepages and was on 12.6% in 2026; GA4 was on 29.7% in 2026.

## Hypotheses (pre-registered 2026-10-19)

1. **Spread slow, decline fast. Not supported.** Of 19 focal tools with a peak of at least 2% inside the window, 7 have both a measurable rise and decline; 2 of them declined faster than they rose. Live chat as a category rose from half its peak to the peak in 3.9 years and, 7 years after the peak, is still above half of it.
2. **Half of installs gone within 4 years, median under 4. Not supported as stated.** The median is exactly 4 years (survival 53.0% at year 3, 44.3% at year 4), and the pre-registered test was a median under 4. Chat and marketing automation (median 3) churn faster than analytics (4) and tag managers (8), as expected.
3. **Consent tools spread fastest in 2019-2026, more than doubling. Supported** on relative growth (2.47 times, ahead of scheduling at 2.07 and tag managers at 1.57). In percentage points tag managers gained more (23.1 against 10.0).
4. **Live chat peaked and fell at least 20%. Supported.** Peak 14.5% in 2019, 7.8% in 2026 (-46%). Drift rose and fell (2.6% in 2020 to 0.2%); Intercom held up better (4.7% to 2.2%). HubSpot's chat was too rare on homepages to measure.
5. **Replacements run inside the suites. Mostly not supported.** Suites took 53% of replacement installs in marketing automation and 39% in forms, but 20% in analytics and almost none in chat or consent. The largest single flow is Contact Form 7 to HubSpot Forms.

## Robustness

| Variant | Median years (new installs) | New installs | Live chat 2026 vs peak | Cookie consent 2026/2014 | Tag managers 2026/2014 |
|---|---|---|---|---|---|
| main | 4 | 18,951 | -46% | 54.1 | 18.5 |
| primary crawl only | 4 | 17,379 | -48% | 55.7 | 18.9 |
| detector only | 4 | 13,281 | -22% | 48.2 | 950.7 |
| all evidence | 4 | 22,788 | -44% | 54.2 | 19.8 |
| balanced panel | 3 | 2,811 | -56% | 135.0 | 16.8 |
| no tag manager | 4 | 4,873 | -63% | 31.9 | |

- **Primary crawl only** drops the fallback captures. **Detector only** drops the live-signature check; it misses most inline chat loaders, which is why its chat decline is smaller. **All evidence** counts every detector match. **Balanced panel** keeps only companies with a usable capture in every year. **No tag manager** keeps only company-years with no tag manager, where no tool can hide inside one.

## Verification

An independent re-analysis, written from the study plan alone without seeing our code, recomputed the headline numbers from the study extract. All 37 checked numbers matched to the reported precision, as did the usable homepages in each year and the median survival in each category.

## Limitations

- **Homepage only, static HTML only.** Tools inside the product, on other pages or loaded by a tag manager are not seen. Tag managers spread from 3.4% to 63.5% of homepages, so some later declines could be tools moving inside one; the no-tag-manager variant shows the chat decline there too.
- **Live loaders, not use.** A loader on the page shows the tool was installed, not that anyone used it or paid for it.
- **Survivor sample.** The companies exist in today's records, so companies that closed before 2026 are missing, more so in early years. Size, industry and founding year are as recorded today.
- **Crawl coverage.** Kenso's web archive does not capture every site every year; small sites are captured less often, and sites that block crawlers are missing. A year with no usable capture is bridged in survival only when the tool is on both sides.
- **Yearly snapshots.** Install and removal dates are known to the year, and survival durations are upper bounds.
- **Detector coverage.** The live-signature list covers 67 common tools; rarer tools rely on the detector alone. Validation against the website index is for 2026 only.
- **Industry labels** do not separate B2B from B2C software.

## Sources

- Kenso's web archive, monthly crawls June 2014 to September 2026.
- Kenso internal data: Kenso's dataset of public company records (company sample only; no person data used).
- Bass, F. M. (1969). A new product growth for model consumer durables. Management Science 15(5), 215-227.
- Kaplan, E. L. and Meier, P. (1958). Nonparametric estimation from incomplete observations. Journal of the American Statistical Association 53(282), 457-481.
- Kenso Research shared homepage panel (built for the Kenso Research study *Half of B2B software homepages now say AI. Did it pay off?*).