Nobody contacts you through your website? Eight checks to run tonight

Your website loads, it looks fine, and nobody enquires. Nine times out of ten that is not a traffic problem. It is that a stranger on a phone cannot tell what to do next, so they leave.

Below are eight checks you can run tonight on your own homepage, with no tool and nothing to buy. Read the table, then read the section for any check you failed. Our own free audit grades five of the eight; the other three you have to run yourself, and this guide says which is which.

The eight checks

Checks 1 and 2: is the next step obvious, and is it on screen?

Check 1 — one button that names the next step. Our crawler reads every link and button on your homepage and looks for action words in both languages: 免費, 立即, 預約, 報名, 登記, 查詢, 諮詢, 聯絡, WhatsApp, "get started", "book", "contact us", "learn more". If nothing on the page matches, the check fails.

Test it free: read only your buttons aloud and ignore everything else. If the list is "Home, About, Services, More", you fail.

Fix tonight: rename one button to the actual next step. 「WhatsApp 查詢價錢」 beats 「了解更多」. Worth noticing: 「了解更多」 is on our word list, so it passes the check while telling a customer nothing. Passing an automated check and being clear are two different things.

What we miss: we read only a and button elements. A button built out of a div is invisible to us, and to a screen reader. Fix the markup.

Check 2 — the step is on the first screen. The check is called "above-fold value", but the code measures one thing: whether an action button sits in the top 700 pixels. Not your headline, not your offer. Geometry.

Test it free: load your homepage and do not scroll. Is there a button?

Fix tonight: move one button up into the hero. That is the whole fix.

What we miss: we measure those 700 pixels in a desktop browser window, not on a phone, so a button that clears the fold on a laptop and sits under it on a handset still passes. Nielsen Norman Group's eyetracking work found people spend 57% of their page-viewing time above the fold — on the screen they are actually holding.

Checks 3 and 4: does a stranger have a reason to believe you?

Check 3 — proof from a named customer. We search the page source for testimonial, review, rating or customer story, and count blockquote elements plus any element whose CSS class contains testimonial, review or quote.

Test it free: does a named human say you did the job well, anywhere on the homepage?

Fix tonight: three real customers — first name, district, one sentence about the actual job. Never invent one. A fabricated review is a trade-description problem, not a marketing tactic.

What we get wrong, both ways: the word list is English, so a 「客戶評價」 section passes only if it sits in a blockquote or carries an English class name. Meanwhile the word "review" anywhere in the source — a script, a cookie banner — passes the check on a page with no testimonial at all.

Check 4 — who you legally are. Our trust check matches four buckets in both languages: security (ssl, 安全, 加密), guarantee (保證, 退款), certification (認證, 獲獎, 資格) and media (合作夥伴, 媒體報導). Almost everyone passes, because "secure" turns up in most sites' scripts, so a green tick here means very little.

Test it free: ask the Hong Kong version of the question instead. A company incorporated here must display its registered name and its liability status on its website — the Companies (Disclosure of Company Name and Liability Status) Regulation, Cap. 622B, sections 4 and 5.

Fix tonight: put your registered company name, a real street address and a real phone number in the footer. Someone deciding whether to send you a deposit reads that, not a padlock graphic.

Check 5: can they finish the form on a phone?

We count every input, textarea and select on the page and fail you above five. Zero fields is not an automatic pass: a page with no form passes only if we can find another route to you — a contact link, an email capture, or a phone number we recognise.

Baymard Institute measured an average of 11.3 form fields in e-commerce checkouts in 2024 and concluded most sites need only eight. That is a checkout, not an enquiry, so read it as direction rather than a target. An enquiry form needs three fields: a name, one way to reply, and one line saying what they need.

What we miss: the count is page-wide, not per form. A search box, a language selector and two cookie toggles all count towards your five.

Check 6: is there a WhatsApp link that opens a chat?

In Hong Kong the contact route is WhatsApp, not a form. 74.7% of internet users aged 16 to 64 used it in the second quarter of 2025, in a market where 7.16 million people — 96.8% of the population — were online that October. A site with no WhatsApp link is losing enquiries no form will recover.

WhatsApp documents the link format: https://wa.me/85298765432, using the full international number with no plus sign, no leading zero, no brackets and no dashes. Add ?text= and a URL-encoded message to pre-fill the chat with the page name, so you can tell which page produced the enquiry.

Test it free: tap your own link on your own phone. It should open a chat with your number already in it. A wa.me link carrying the local eight digits without 852 opens nothing.

What we miss: we spot WhatsApp only by finding wa.me in the page. A LINE or WeChat QR image, a Messenger button and a plain tel: link all score nothing in our contact check. Nor does a Hong Kong phone number written the local way: our detector needs at least ten digits, so 2345 6789 does not register, while +852 2345 6789 does. Write the international form everywhere — ours is not the only machine reading your page.

Checks 7 and 8: speed, and the thumb test

We do not measure either one. Until September 2026 our report printed "Mobile UX" and "Loading speed" as passes: both were hardcoded to true, with no measurement of any kind behind them. We removed the rows rather than keep printing a tick we had not earned. Measure them yourself tonight, free:

• PageSpeed Insights, mobile tab. Google's "good" thresholds are Largest Contentful Paint within 2.5 seconds, Interaction to Next Paint of 200 milliseconds or less, and Cumulative Layout Shift of 0.1 or less, taken at the 75th percentile of page loads and measured separately for mobile and desktop.
• Your own phone, on mobile data. Among Hong Kong households with home internet access in 2025, the smartphone was the most popular device used to connect, at 99.9%. Tap every button with a thumb. WCAG 2.2 sets a minimum pointer target of 24 by 24 CSS pixels at level AA.
• Google's page-experience guidance. It asks whether the page is served over HTTPS, displays well on mobile, has good Core Web Vitals and avoids intrusive interstitials — and it says plainly that there is no single signal.

What our audit sees, and what it does not

A green tick you cannot trust is worse than no tick, so here is the whole scoring sheet.

Our conversion score is 100 points across nine signals. The report on screen shows four of them. The PDF shows five. Four signals worth 35 points between them are scored and never shown to you at all.

The four hidden ones are also the crudest. Live chat is matched by brand name, so ten points go to whoever installed Intercom. Pricing matches the word "plan" anywhere on the page. The FAQ needle is English only, so a Hong Kong site with a 「常見問題」 section loses five points it has earned.

Three checks left the report in September 2026. "Mobile UX" and "Loading speed" were the unconditional passes above, and "Navigation clarity" returned the same result as the call-to-action check without ever looking at a menu. If you are holding an older PDF with eight ticks on it, three of them meant nothing.

Two more detectors are worth knowing about. Our blog check matches any link containing /blog, /news, /articles or /posts — so /newsletter reads as a blog — and it never opens the page or reads a date, so a blog last touched in 2019 counts as present while a live one at /insights counts as absent. Our email-signup check wants type="email" or the words newsletter, subscribe, signup, 訂閱, 電子報 or 電郵通訊, so a sign-up box built as type="text" under the label 「留下電郵」 is invisible to it.

That is our list to fix. It is also the lesson: a machine reads your page the way a hurried stranger reads it, by labels and structure, in about five seconds. If the label does not say what the thing is, neither of them finds it.

What we would skip

Skip the live chat widget. Our own score gives it ten points and, for a three-person Hong Kong business, we think that is wrong. An unstaffed chat bubble is worse than none.

Skip exit-intent popups and stock "trust badge" images. Google's page-experience guidance names intrusive interstitials as a problem, and a padlock graphic proves nothing.

Skip the famous conversion statistics. Most of the "88% of users…" numbers circulating online trace back to nothing checkable, which is why every figure here is named and dated in the sources below.

Run the checks on your own site

Paste your URL into the free WITTI Grow audit. It grades the conversion checks above plus ten SEO checks in about three minutes, with no sign-up and no card, and tells you which ones your homepage fails today. Work down this list with the report open, and treat checks 7 and 8 as yours to measure by hand.

Next guide: Your page title is costing you customers: the ten SEO checks we run