Abstract
We analyze 20,042 behavioral events across 3,557 sessions pulled from a production export of yaaf.site’s live server — real visitor traffic on the site’s business-template demos, logged between 17 and 26 July 2026, not a designed or simulated sample. We set the resulting behavioral profile against published, real-world benchmarks for Western web traffic: device mix, time on page, and contact-conversion rate. The comparison points in a consistent direction: this is an overwhelmingly mobile, narrow-screen, WhatsApp-leaning audience, one that spends less than half the Western benchmark time per page yet converts to contact at nearly ten times the Western form rate. At n=3,557 every headline figure rests on thousands of observations, and every business category carries at least 70 sessions — but the window is short, ten days, and traffic inside it grew more than twentyfold from first day to last. We report sample sizes throughout, because the principal risk in this dataset is composition, not size.
Why this snapshot matters
yaaf.site sells ready-made websites to small Cameroonian businesses — the corner salon, the pharmacy, the private school. Every template in the catalogue is instrumented the same way, logging page views, scroll checkpoints, clicks, and a final exit event with total time on page. This report reads ten days of that real instrumentation — 3,557 actual visits across ten trades and three pricing tiers — against the same Western web-analytics benchmarks that quietly justify most default design advice: contact forms over messaging apps, desktop-first breakpoints, hover states built for a mouse.
Cameroon is still in the early-middle of its internet transition: 12.6 million people online, 41.9% penetration, against 29.0 million active mobile connections covering 96.4% of the population6. That gap — near-universal mobile connections, much thinner internet usage — is the backdrop every number below should be read against. Where an earlier, much smaller sample could only hint at the shape of that behavior, 3,557 sessions are enough to state it.
Data & method
The source file, catalog_events_new.sql, is a production dump of the catalog_events table — a live log, not a designed sample. Each row records one of six event types — page_view, scroll_depth, click, contact_click, whatsapp_click, page_exit — against session ID, template (“catalog site”), business category, device type, screen width, and timestamp. We grouped the 20,042 rows into 3,557 sessions by session ID, then derived per session: device type, total engaged seconds (summed from every page_exit.seconds in that session), maximum scroll depth reached, and whether the session contained any contact-oriented action. Every definition used below is stated explicitly, so each figure is directly reproducible from the raw file.
Two wrinkles worth flagging up front. First, 3,242 of 3,557 sessions (91.1%) recorded at least one page_exit event; the remaining 8.9% ended without one — a tab closed abruptly, a connection dropped — and are excluded from duration statistics but counted everywhere else. Second, the window contains no extreme idle-tab outliers of the kind that can wreck a duration average: the longest single session ran 16 minutes 46 seconds, and removing the top two sessions moves the mean by less than a second. We still report medians as the headline statistic throughout — the duration distribution is heavily right-skewed (median 34s, mean 76s) — and note the mean separately where it matters.
For the Western side, there is no single equivalent dataset — no company publishes raw session logs for “the Western web.” Instead we use the closest independently published aggregate benchmark for each metric: StatCounter for device share34, Contentsquare’s Digital Experience Benchmark for engagement and bounce12, and cross-industry form-conversion studies for contact-intent1516. Every Western figure in this report is footnoted to its source.
Findings
3.1 The mobile-only web
86.3% of sessions opened on a phone, 11.9% on a desktop or laptop, and 1.8% on a tablet. In the published Western benchmarks, the pattern runs the other way: desktop still edges out mobile in both the United States (59.5%) and Europe (53.9%) as of June 202634. This is not a finding sample size can threaten in either direction: the gap between 86% and 41–46% is roughly a factor of two, measured across thousands of sessions.
Figure 1 — Device share by region
Share of sessions/pageviews by device · Cameroon = real yaaf.site production export, n=3,557 · US/EU = StatCounter, June 2026
3.2 Real sessions are brief
Median total session length was 34 seconds; median dwell on a single page was 23 seconds — identical on mobile, the dominant device — against a Western cross-industry average of roughly 52–54 seconds13. Visitors here spend well under half the Western benchmark time per page. 30.3% of sessions ended after a single page with almost no scrolling and under 30 seconds spent — a real quick-exit rate, still meaningfully below the Western single-page exit rate of about 53%2, but no longer a rounding error.
Scroll behavior is where this audience departs most sharply from the assumptions built into a long marketing page. Across the 2,812 sessions with scroll tracking, the average deepest checkpoint reached was 56.4%, and only 4.2% of them logged the 100% checkpoint; 45.9% got as far as 75%. The exit-reported maximum, which is finer-grained than the four checkpoints, tells the same story: a median deepest scroll of 58%, with 20.1% of sessions reaching 90% or more. Most visitors here see the top half to two-thirds of a page and no further — which puts a hard practical premium on what sits above the fold.
The mean session length, 76.3 seconds, runs more than twice the median. That gap is not the work of a couple of abandoned browser tabs — the longest session in the whole file is 16 minutes 46 seconds, and dropping the two longest moves the mean by less than a second. It is simply a long right tail: the top decile of sessions runs past 212 seconds while the bottom three deciles finish inside 16.
Figure 2 — Time spent, Cameroon (real) vs. Western benchmark
Cameroon values are medians (n=3,242 sessions with exit data, robust to right skew) · Western values are published cross-industry means — not a perfectly matched statistic, but the best available on each side
Twenty-three seconds is not indifference. On a metered connection, it might be exactly how long it takes to decide whether the rest of a page is worth loading.Reading the real session data
3.3 Contact intent: high, and WhatsApp-led
16.6% of sessions (589 of 3,557) took some contact-oriented action, roughly four to ten times the 1.7–4% Western form-conversion range1516. The WhatsApp figure stands on 259 clicks across 205 sessions: 5.8% of all sessions opened a WhatsApp chat — more than three times the 1.7% Western contact-form average, and on its own a higher conversion rate than most Western forms achieve. A further 7.8% of sessions used an on-page contact control and 7.2% clicked something else instrumented as a generic click; the three overlap, which is why they sum to more than 16.6%.
Conversion is also remarkably flat across pricing tiers: 17.3% for Starter templates (n=1,792), 16.9% for Pro (n=472), 15.4% for Premium (n=1,293). Whatever drives a visitor to make contact, on this evidence it is not the amount of design polish on the page.
Figure 3 — Contact-intent rate
Share of sessions that took a contact-oriented action · n=3,557 for Cameroon figures
3.4 A narrow, two-cluster screen
Median viewport width was 412px. The distribution is essentially two clusters with a near-void between them: 78.5% of sessions sit at 430px or below, a hard zero across the entire 481–767px band, a thin 1.8% tablet presence at 768–1024px, and 11.9% spread across the 1025–1919px laptop range. Not one session in 3,557 arrived on a viewport of 1920px or wider. The design implication is direct: the layout that matters is the one that renders between roughly 360 and 430 CSS pixels, where 71.1% of all sessions live.
Figure 4 — Screen-width distribution
Share of sessions by first-touch viewport width · Cameroon (yaaf.site, real, n=3,557)
screen_width per session, n=3,557.3.5 Ten days, ten categories
Restaurants (1,222 sessions) and salons (996) account for 62.4% of the sample, but every one of the ten business categories carries enough sessions to support a percentage worth reading: the smallest, churches, has 70. The category spread of contact intent is narrow — every category lands between 12.8% and 22.9% — but it is not flat. Churches (22.9%, on the smallest cell in the table at n=70), clinics (22.1%) and real estate (21.6%) convert at roughly half again the rate of restaurants (14.9%) and boutiques (12.8%). Real estate and auto also show the heaviest WhatsApp lean, at 11.3% and 10.5% of sessions respectively, against 5.1% for restaurants — consistent with categories where the next step is a negotiation rather than a booking.
| Category | Sessions | Avg. time engaged | Contact action |
|---|---|---|---|
| Restaurants | 1,222 | 1:09 | 14.9% |
| Salons | 996 | 1:21 | 18.5% |
| Hotels | 403 | 1:21 | 15.6% |
| Schools | 330 | 1:13 | 16.1% |
| Boutiques | 180 | 1:22 | 12.8% |
| Clinics | 104 | 1:25 | 22.1% |
| Real estate | 97 | 1:26 | 21.6% |
| Professional services | 79 | 1:30 | 15.2% |
| Auto | 76 | 1:18 | 15.8% |
| Churches | 70 | 1:03 | 22.9% |
Discussion
Three independently documented facts about Cameroon plausibly explain the shape of this behavior:
- The infrastructure is mobile by default, not by preference. 96.4% of the population holds a mobile connection against 41.9% who count as internet users at all6 — for most people online, the phone isn’t the second screen, it’s the only one. Smartphone adoption across Sub-Saharan Africa sits at 54%, projected to reach 81% by 20308, and a 4G-capable handset still costs roughly 26% of monthly GDP per capita in the region, against 16% elsewhere in the low-and-middle-income world7. A 412px median screen width, with nothing at all above 1919px, is what that affordability ceiling looks like in a viewport.
- Short sessions do not mean disengaged visitors. A 23-second median page view, read next to a 16.6% contact rate, is consistent with someone deciding quickly whether a page is relevant and, having decided, acting rather than lingering — a plausible pattern for a visitor budgeting mobile data rather than one with unlimited broadband to spare. The scroll data sharpens the design brief considerably: with a median deepest scroll of 58% and only one session in five passing 90%, anything that has to be seen has to be in the top half of the page.
- WhatsApp reads as the more natural contact channel, and the data says so directly. With 400+ million active users across Africa and adults messaging businesses at 80%+ weekly rates910, and open rates above 90% against 20–25% for email10, a 5.8% WhatsApp-open rate on 3,557 sessions — more than triple the Western contact-form average15 — is what that ecosystem looks like at the level of an individual small-business page.
For yaaf.site’s own template roadmap, the practical read stays simple and is now better supported: design for a narrow phone viewport first, keep the WhatsApp button visible without scrolling, and don’t expect a first-time visitor to linger — the page has to make its case in the first handful of seconds and the first half of its own height, not the first few minutes.
Limitations
- Ten days is a short window, and it was not a steady one. Daily sessions rose from 39 to 823 across the ten days. Whatever drove that growth — a campaign, a share, seasonal timing — means later days dominate the totals and the visitor mix is shifting inside the sample. This report describes 17 June–26 July 2026, not “Cameroonian behavior” in general.
- The acquisition mix changed mid-study. 17.0% of sessions arrived from external referrers — 12.7% from Facebook link shims, 2.9% from Google, 1.4% from in-app Android browsers — alongside 37.8% from yaaf.site’s own catalogue and 45.1% with no referrer. Referred and direct visitors behave similarly here (14.5% vs 17.2% contact rate, identical median duration), but the external share climbs over the ten days, so the visitor mix that produced the last few days of data is not quite the one that produced the first few.
- 8.9% of sessions have no duration data. 315 sessions ended without a
page_exitevent and are excluded from every timing statistic. If abrupt exits skew short — which is plausible — the true median session length is lower than 34 seconds, not higher. - Scroll depth is measured two ways and they disagree in emphasis. Checkpoint events say 4.2% of sessions reached 100%; exit-reported maxima say 20.1% reached 90% or more. Both are in the file; we report both rather than pick the flattering one. Templates also differ in page length, so “depth” is not a constant quantity of content across categories.
- The Western side is still aggregate, not template-specific. StatCounter, Contentsquare, and the conversion-rate studies describe the general Western web, not visitors specifically evaluating small-business website templates. The device and time-on-page comparisons are like-for-like enough; the conversion comparison sets a WhatsApp tap against a submitted form, which are different acts of commitment.
Conclusion
Three and a half thousand real sessions describe a consistent visitor: on a narrow, budget-Android phone, deciding in twenty-odd seconds whether a page is worth their data, rarely scrolling past its midpoint, and reaching for WhatsApp at more than triple the rate Western visitors submit a contact form. Every one of those tendencies is more pronounced than the Western benchmarks describe, and each rests on hundreds or thousands of observations. The honest brief for yaaf.site’s template roadmap is to build for that visitor specifically — phone-first, WhatsApp above the fold, the argument made in the top half of the page — and to keep collecting real production data like this over longer windows, so that the next report can separate what Cameroonian visitors do from what a ten-day growth spurt made them look like.
References
- Contentsquare. “The 2026 Digital Experience Benchmark Report.” contentsquare.com/guides/digital-experience-benchmark
- Contentsquare. “2025 Digital Experience Benchmarks: Rising Acquisition Costs and More.” contentsquare.com/press/2025-digital-experience-benchmarks
- StatCounter Global Stats. “Desktop vs Mobile Market Share — United States of America,” June 2026. gs.statcounter.com
- StatCounter Global Stats. “Desktop vs Mobile Market Share — Europe,” June 2026. gs.statcounter.com
- StatCounter Global Stats. “Desktop vs Mobile Market Share — Worldwide.” gs.statcounter.com
- DataReportal. “Digital 2026: Cameroon.” datareportal.com/reports/digital-2026-cameroon
- GSMA. “Pioneering Affordable Access in Africa.” gsma.com
- GSMA Intelligence. “The Mobile Economy Africa 2025.” gsmaintelligence.com
- MyNewsGH. “How WhatsApp Business API Is Reshaping Customer Communication in Africa.” mynewsgh.com
- ElectroIQ. “WhatsApp Business Statistics By Users And Downloads (2025).” electroiq.com
- Leadferno. “Website Contact Forms and Lead Management: Uncovering Costly Mistakes.” leadferno.com
- Digital Silk. “Top 50 Small Business Website Statistics.” digitalsilk.com
- HubSpot. “What Is the Average Time Spent On a Website?” blog.hubspot.com
- Databox. “Average Session Duration: Benchmarks, Definitions, & Pro Tips,” 2025. databox.com
- Zuko. “25 Conversion Rate Statistics You Need in 2025” (citing Ruler Analytics). zuko.io
- Lucky Orange. “What Is a Good Conversion Rate? Benchmarks by Industry.” luckyorange.com