FINternship, Mentor-led apprenticeship in Singapore

The order a first-time visitor should meet your page in, built from what you sell. Each badge says what to do with that block.

Your page rewritten section by section, with a couple of options for each line and a positioning line to aim at. These are starting points, so pick one and edit.

The hero has a real point of view. Condensed all-caps display type set in navy and gold, a wide cream field, and a photo of actual students in a Singapore office give you a page that does not look like a template, which satisfies the visual snap-judgment and aesthetic-usability principles better than most sites in this category. The clarity test is where it slips. "LEARN THE SKILLS SCHOOL NEVER TAUGHT YOU." could sit on a study-skills course, a personal finance newsletter, or a coding bootcamp. Your visitor has to read the small gold badge and then the subline to learn this is a six-week mentor-led apprenticeship. The single most persuasive fact on the whole page, that it is free, appears only in a checkmark row that sits below the desktop fold. Worse, at 1440x900 the headline is so tall that both hero buttons are sliced in half by the fold line, so the primary action fails the above-the-fold principle on desktop even though mobile passes it cleanly.

Composition is the strongest part of this page. Dark navy bands alternate with cream and warm grey, which chapters a 9,800px scroll into sections a skimmer can navigate, and that satisfies the common-region principle properly. Card grids sit on a consistent grid, section eyebrows sit at a fixed offset above each heading, and the vertical rhythm between sections holds steady down the whole page. Whitespace is genuinely generous rather than merely empty. The problem is the top of the funnel. Your headline is so dominant that it beats the action it is supposed to lead to, and the buttons fall off the screen entirely on a 900px-tall desktop viewport. Von Restorff works against you when the loudest element is not the one you want clicked. Lower down, the signal-to-noise ratio drops: the six-card "SUPPORT THAT DOES NOT STOP AT WEEK SIX." grid repeats the four cards from "PRACTICAL EXPERIENCE FROM DAY ONE." almost point for point, adding another full screen of scroll before the closing CTA.

Type is the reason this page does not look generated. The display face is a tightly tracked condensed grotesque with leading close to 1.0, exactly what the display-tracking principle asks for, and it gives the brand a voice before a word is read. Body copy is set around 17px with comfortable line height, left aligned, and the measure holds near 60 characters in the hero and the section intros, all well inside the readable range. Two families at most appear across the page and each keeps a clear role. What costs you is uniformity in the section headings. Every H2 down the page is set in the same all-caps condensed slab at close to the same size, so "WHY WE BUILT FINTERNSHIP." shouts exactly as loud as "WHAT PARTICIPANTS SAY." and "THREE STAGES, ONE TRAJECTORY.". A skimmer scrolling nine thousand pixels gets twelve equally loud shouts and no sense of which section matters. Nothing ranks itself.

The palette itself is disciplined. Navy, cream, and a single mustard gold carry the entire page, with warm grey as a fourth surface, and nothing clashes. Legibility is where it breaks. That gold sits at roughly 2.2:1 against the cream background, which fails the 4.5:1 floor for small text and even the relaxed 3:1 floor for large display type. Every section eyebrow on the page uses it: "SINGAPORE · MENTOR-LED APPRENTICESHIP", "WHAT YOU GET", "WHY WE BUILT IT", "FROM PAST COHORTS". These are your wayfinding labels and they are the hardest text on the page to read. Half of your headline has the same problem, so "SCHOOL NEVER TAUGHT YOU." reads noticeably fainter than "LEARN THE SKILLS" above it. The same gold works well against navy in the dark bands, which proves the colour is fine and the pairing is not. Action colour is also inconsistent, with a navy button at the top and a gold one at the close.

The path works when your visitor finds it. Clicking "Apply" from the header took our test visitor to a page headed "APPLY FOR THE NEXT COHORT." with the line "10 minutes · Every submission reviewed manually" sitting right under the button, which is genuine risk reversal and near-perfect message match. The apply action is also repeated at the bottom of the page, so a convinced reader does not scroll back. Three things cost you conversions. On desktop the hero buttons are cut off by the fold, so the first action a visitor sees is the small header pill. Beside "Apply Now" sits "Read the blog" at the same size and shape, and our test visitor clicking it landed on the blog index, out of the funnel. Then there is the unanswered question: the page says "next cohort" everywhere and never gives a start date, an application deadline, a weekly time commitment, or how long a reply takes. A student cannot tell whether this fits around their semester.

Craft buys you a lot here. The page looks like someone designed it on purpose, the copy is free of typos, images all render, and nothing overlaps or breaks. Three named testimonials with schools attached beat the unattributed quotes most programme sites run. The "HONEST FIT CHECK" section, which openly lists "Only want something to add to your CV" under "NOT FOR YOU IF", reads as confidence rather than sales pressure and earns real trust. What is missing is the organization behind it. The page promises "Founder-led mentorship" and "direct feedback from experienced operators" and never names one of them, even though the nav has a Mentors link. A stranger being asked to hand over six weeks cannot see who they would be learning from. The footer gives you a company blurb and "© 2026 FINternship · Singapore" but offers no email address, no phone, no privacy policy, and no terms. "1,000+ students mentored" carries no cohort count or year to anchor it.

You have the right instincts and the wrong evidence. "WHAT PARTICIPANTS SAY." gives three quotes with full names and institutions, which is more honest than most, but every quote is a feeling rather than a result. "I became more confident speaking and presenting ideas." tells a sceptical student nothing they can weigh against six weeks of their time. Reciprocity is genuinely strong, since the programme is free and the blog gives away real career writing before asking for anything. Shared identity lands well too, because "STUDENTS", "NSFS", and "career switchers" name your reader precisely, and the fit-check section speaks their language. Authority is absent. There are no mentor names, no company logos, and no named businesses behind the "real briefs from Singapore businesses" claim that appears twice. Scarcity is left on the table entirely: a free cohort with limited places and a manual review is naturally scarce, yet the page never says how many spots there are or when applications close.

The writing has a voice and mostly avoids hype. Sentences run short, verbs are active, and lines like "You stop doing the work and start owning it" earn their place. "THIS PROGRAMME REWARDS OWNERSHIP." with its two-column fit check is the strongest copy on the page because it disqualifies people, which makes everything else more believable. Subheads carry meaning on their own, so a skimmer can navigate by heading alone. The failure is repetition. Five separate sections make overlapping arguments: the four benefit cards near the top, the finance crossover section, the finance core grid, the three-stage roadmap, and the six-card ecosystem grid all circle the same promises of projects, mentors, money skills, and network. By the fourth pass your reader is not learning, they are scrolling. Specificity is also thin where it matters most. "Live business briefs" and "real Singapore businesses" appear repeatedly without one example of a brief, a company, or a deliverable a student produced.

The structural basics are in good shape. There is exactly one H1, heading levels nest properly with no skips across thirty-three headings, the page declares its language, and every interactive control we checked carries an accessible name. Link and button text reads clearly out of context, since "Apply Now" and "Read the blog" say where they go. Text stays legible at 375px without pinch-zoom. Two things need fixing. One of the eight images on the page, the group photo of a cohort walking in the CBD, has no alternative text, so a screen reader user gets nothing where a sighted visitor gets a picture of the community you are selling. Tap targets are undersized on mobile: the header "Apply" button measures 70 by 36 pixels and the logo link 81 by 32, both under the 44-pixel comfortable target and only just clear of the 24-pixel floor. Since that header button is the only apply action visible on most of the mobile scroll, its size matters more than the number suggests.

Speed is not costing you anything. The document was ready in 0.73 seconds and the load event fired at 1.02 seconds, comfortably inside the 2.5-second threshold for main content. The hero rendered complete in our first capture with the headline, subline, and photo all in place, so there is no blank-then-pop moment for your visitor. Every button click registered immediately and navigated to a real page, with no stalls or dead clicks in the whole run, which means interaction responsiveness and feedback are both fine. Layout held stable between the first capture and the settled page. The only soft spot is total asset weight. Network activity kept going for about six seconds after navigation started, which fits a page carrying a large hero photograph plus roughly a dozen more photos down a nine-thousand-pixel scroll. That does not delay the first screen, but on a mid-range phone over mobile data the lower sections will fill in slowly while your reader scrolls.

Mobile is handled better than desktop, which is fortunate given who your audience is. At 375px there is no sideways scrolling, the layout genuinely reflows into a single column rather than shrinking the desktop view, and the first screen carries the badge, the full headline, the subline, both buttons, and all four trust checkmarks including "Free 6-week apprenticeship". That is more persuasive content in one phone screen than the desktop fold delivers. Body text stays readable without zoom. Two issues remain. Tap targets are tight, with the header apply button at 70 by 36 pixels, below the 44-pixel size that a thumb hits reliably. Then there is the scroll itself. On a phone this page runs many screens deep, and after the hero buttons scroll away the only apply action is that small pill at the top until the very bottom. A student reading the testimonials or the fit check has no action within reach.

• The visual identity is genuinely distinctive. Tightly tracked condensed display type in navy and gold on cream, with alternating dark bands chaptering the scroll, looks designed rather than assembled from a template.
• "THIS PROGRAMME REWARDS OWNERSHIP." with its "NOT FOR YOU IF" column is the best copy on the page. Openly disqualifying people who "Only want something to add to your CV" makes every other claim more believable.
• The apply page keeps the promise the button made. Our test visitor landed on "APPLY FOR THE NEXT COHORT." with "10 minutes · Every submission reviewed manually" sitting right under the button, which answers the two questions a hesitant applicant has.
• Testimonials carry real names and institutions, with Jun Xi at SMU, Gabriel Kow post-ORD, and Kenny Soh at NTU. Attribution this specific is rare on programme sites.
• Speed is excellent. The page was ready in under a second and the hero rendered complete, so nobody bounces waiting.
• Mobile reflows properly and puts the headline, both buttons, and the free offer in one phone screen.

Load performance

Accessibility quick-checks (desktop)

Mobile tap targets (375x812)

Deterministic signals measured in-browser at capture time, not model output.

Make the changes above, then run coldeye on the new version. It reads the old and new side by side, blind, twice with the positions swapped, and tells you which one actually converts better. A score on its own drifts about half a point between runs, so the blind verdict is the part to trust.