Conversion audit & proposal
The website through a product, growth and UX lens. Baseline: the live prototype, which is good and stays the foundation. This is not a redesign. It is a re-aim.
Until checkout exists, this site cannot take money. It has exactly one conversion currency: an email address with intent data attached. Every section either earns that email, builds desire toward it, or is dead weight.
Today the page is dressed like a finished store without a cash register: tier cards say "Get started" and scroll you to a free email form. Instead of hiding the gap, we use it. The site runs in explicit pre-launch mode: pricing becomes a research instrument (tier buttons reserve the launch price and tag the tier in MailerLite), the free pack is the north-star conversion, and September 1 becomes real, honest urgency. On launch day the same buttons flip to buy buttons. Nothing else moves.
The hero concept
Built photo next to the self-drawing plan. No competitor shows built proof beside the drawing. Sharpen, never replace.
The design restraint
Architects judge us like a portfolio. The quiet is an asset.
Free pack as trial
The right mechanism for non-refundable digital goods.
One-time pricing wedge
"You already pay Autodesk monthly" is the best line on the page.
Mobile buries the message
High impactEffort STraffic comes from Instagram and LinkedIn, meaning phones, meaning the in-app browser. On a 380px screen the visitor gets ~1.5 screens of media before a single word. Beautiful, but mute.
Mobile-only reorder: kicker, headline, descriptor and CTA first, then the plan video, then the photo. Desktop unchanged. Pure CSS.
Nobody says what this is
High impactEffort SThe hero has a promise but no category. A first-time visitor cannot tell if this is software, a plugin, a course or a stock library. The five-second test fails.
One descriptor line under the headline: "Parametric AutoCAD blocks and the method behind them. Drop, stretch, done." The mock dropped this line as a composition study, not a copy decision.
The photo-plan pairing is implicit
High impactEffort SThe hero's entire persuasion is "this drawing became that room," and we never say it. The connection is left to luck.
Paired captions: photo "Built: Tra.ma, Madrid" / video "Drawn: the same project, real speed."
Riqui must confirm the video plan is the same project as the vanity photo. If not, swap the photo or soften the caption.
A 20MB first impression
High impactEffort MMeasured on the live deploy: the hero video is 14.5MB, the Method poster is 4.7MB rendered at 35% opacity, and 24MB of unused images ship alongside. On 4G inside Instagram's browser, the wow arrives after seconds of beige.
Two-tier video: the hero gets a ~2MB sped-up 15-second loop (the drawing as time-lapse); the full 2:25 real-speed video moves to the Method section, click to play. Compress every image, delete the dead ones. Target: under 3MB initial payload. Speed is a conversion feature.
The Method section is a broken promise
High impactEffort SA play button that does nothing and "coming soon." Dead controls teach visitors to stop clicking, and they generalize the lesson to the buy buttons.
The full real-speed video lives here today, with controls and an honest label: "The full draw. 2:25, real speed, no cuts." Swap in the master walkthrough when it exists.
Placeholder quotes are anti-proof
High impactEffort S"[Name], [Studio, City]" reads as template debris to friends and as fabrication to strangers. Meanwhile our strongest borrowed authority, Never Too Small, is a footnote at the bottom of the page.
Cut the quotes section. Add a slim proof strip under the hero: "Designed in daily practice at Tra.ma, Madrid · Featured on Never Too Small." Real quotes return when they exist; the friends round is the collection mechanism.
Pricing CTAs write checks the page can't cash
High impactEffort M"Get started" on a €199 card that scrolls to a free email form feels like bait once noticed. And we throw away the intent signal.
Tier buttons become "Reserve launch price" and open the capture with the tier tagged in MailerLite. We get a segmented launch list and willingness-to-pay data before writing checkout code. On Sep 1 the same buttons become buy buttons.
Tier paralysis with no safety net
High impactEffort SThree tiers, no refunds, no on-page way to undo a wrong choice. Choice risk suppresses entry. The 15-day upgrade policy already exists but is hidden in the email flow.
One quiet line under the grid: "Start anywhere. Upgrade within 15 days and pay only the difference." A hidden support policy becomes the answer to the number one pricing objection.
Reverses the earlier email-only decision. Founders should nod.
We sell blocks nobody can see
High impactEffort LThe core ask is €129-399 for a catalog, sight-unseen. The page shows three 24px generic icons. The catalog is the product.
Phase B: a drafting-catalog strip on the homepage with real block line art and per-tier counts, linking to a full catalog page by Aug 1. Revives the already-validated catalog concept in the new design language.
The Try overlay demos the wrong thing
Med impactEffort MDragging three static shapes proves nothing; any PDF viewer can drag. The product's magic is parametric stretch. A weak demo lowers perceived product quality.
Port the door simulator from the previous build (drag to stretch, type dropdown, already well-reviewed), or hide "Try KitCAD" until it's ready. A missing demo beats a shrug.
The page ends without an ask
Med impactEffort SFAQ, then footer. The visitor who read everything is the highest-intent person on the site, and we say nothing to them.
A closing dark CTA section, plus the dismissible sticky bar after the hero scrolls away (existed before, lost in the port).
We are flying blind
High impactEffort SNo analytics of any kind. Nothing above can be validated or tuned, and design arguments have no referee.
Cloudflare Web Analytics: cookieless, free, no consent banner required, one script tag. KPI set defined below.
The current page is tasteful but could belong to any premium SaaS. The goal is recognition: an architect should feel "these people draw like me" within one screen. Four moves, all cheap, all restrained. This document you are reading uses three of them.
1. Real block line art as the icon language
Replace the generic icons in Free pack and What's included with actual KitCAD block drawings. The assets already exist. The most "us" move available, and it doubles as product preview.
2. The footer as a drawing title block
Project, scale, date, revision letter, drawn-by. Architects will smile; nobody else will notice. Signature move.
3. Kickers as sheet numbers
"A-01 · The method", "A-02 · Free pack". Quiet, systematic, ours.
4. Dimension-line section dividers
Thin rules with end ticks and a centered label, used in 2-3 places. Discipline rule: never more than two identity moves visible per screen. No blueprint kitsch.
- Exit-intent popups, countdowns, fake scarcityBrand-corrosive for a design-literate audience. The only urgency we use is real: launch pricing ends Sep 1.
- Gating the video or catalog behind emailThe wow is what earns the email. Locking it inverts the trade.
- Splitting into multiple pages nowA single page keeps narrative momentum with cold traffic. Split when the catalog is real.
- Moving pricing above method and proofNew brand, considered purchase: belief before price.
- Fabricated social numbers"Join 500 architects" when it isn't true. We show counts when they are real.
1-2 working sessions
- Leaks 01-08, 11, 12
- Identity moves 1-2
- FAQ additions as placeholders
- Leak 08 ships with founder sign-off
Show the goods
- Catalog strip + catalog page (leak 09)
- Parametric door demo (leak 10)
- Real testimonials from friends round
- Founder photos, statement time-math
Turn the key
- Checkout decision executes
- Reserve buttons flip to buy
- Launch-price deadline messaging live
- Riqui: is the video plan the same project as the vanity photo? Determines the hero captions. Optional: a clean capture without the right-click menus.
- Founders: OK to put the 15-day upgrade-difference promise on the page? Reverses the email-only decision.
- Alex: MailerLite form action URL (still pending), and confirm groups/fields are available for tier intent.
- Riqui / Adri: honest average redraw minutes per element for the statement math. No number, no math section.
- Anyone: does hello@kitcad.com exist? The footer and the Studios CTA point at it.
| Metric | What it tells us | Feeds decision on |
|---|---|---|
| Email capture rate | North star. Visitors to free-pack signups. | Everything; the referee metric. |
| Tier-reserve clicks | Willingness to pay, by tier, pre-checkout. | Pricing, tier structure, launch stock of upgrade codes. |
| Full-video play + completion | Does the method story hold attention? | Video length, hero loop cut. |
| Sticky bar clicks | Recovered intent from deep scrollers. | Keep or kill the bar. |
| FAQ opens | Which objection blocks purchase. | Copy priorities, new sections. |
Tooling: Cloudflare Web Analytics, cookieless, no consent banner needed. The friends round is structured, not vibes; the feedback pill asks three fixed questions: what did you think this was in the first five seconds, would you pay and which tier, and what almost stopped you.