Trackwell

Marketplace

Trackwell | cto.new Marketplace

…/l/mgnryd

…/l/wwsuf

The buying path was driven to the order form with a 100%-off code, and it stops at reCAPTCHA (2026-10-12). A one-use test code applied correctly on the live checkout — the page read "Discounts TEAMTEST100 US$−15 / Total US$0" — and checkout accepted an email-only buyer, no account, no card. Pressing Get hands the submit to a Google reCAPTCHA challenge (cross-origin frame): our browser got a chain — motorcycles → chimneys → motorcycles → bicycles → buses; one was solved and accepted (four tiles, SKIP→NEXT, next challenge issued), but the chain never finished — one challenge expired at reCAPTCHA's two-minute limit and later attempts ended with the flow dropped. Nothing was bought: GET /v2/sales reads 0 sales and times_used stayed 0 — no money moved, no order exists. The code has been deleted, and the Weekly Planner re-reads unchanged (price 1500, published, file attached, 10 tags). Routine friction for a human, a hard stop for a script — so this question can no longer be closed by the team. Evidence: verify/test-purchase-attempt-2026-10-12.md.
The visit counter is live and counting (2026-10-11). The page carries the owner-approved sentence verbatim: "We count page visits and which planner buttons get clicked — in aggregate, with no cookies, no third-party scripts, and no personal data." Narrow on purpose: aggregate pageview, one counter per buy button (fixed five-product allowlist — a visitor cannot invent a row), a referring host capped at 500; no cookies, IPs, user agents or per-visitor identifiers, read server-side only. Verified live, not eyeballed: the sentence renders, <form count 0, 5 distinct buy links, one column at 390 px with no sideways overflow; pageview moved 3 → 7 across the lead's own visits, all five buy:* counters fired once, the waitlist table is still exactly 1 row. Standing caveat: the self-host skip-list hard-codes the current dev and live hosts — a custom domain must be added there. Evidence: verify/storefront-visit-counter-live-2026-10-11.md.
The account can now be paid — the owner confirmed payout settings and tax details complete on 2026-10-11. Owner-reported, not team-verified: GET /v2/user returns no payout field.
All five listings read back through the API on 2026-10-11 (owner's token, never printed): each published, file attached (Budget 16 KB, Habit 25 KB, Weekly 12.7 KB, Bundle 48.2 KB, Daily 8.6 KB), correct price (1400/1300/1500/3400/999) and 0 sales. Tags are set on all five and verified by read-back — 10/10/10/9 on ours, 6 on the Daily; before that write all five read tags: [], because Gumroad caps a tag at under 20 characters. Corrected lists live in gumroad/listing-pack.md. The write sent tags only, never files — a product update replaces the collections it names, so a files send would have deleted the product; file, price and published state were re-read and were unchanged. Evidence: verify/listing-tags-applied-2026-10-11.md.
Our own records were carrying dead facts, and that is fixed. A 2026-10-11 sweep found 30 files still describing a $19 product that never existed, a 14-day refund we never offer, and — in the documents written for the owner — "nothing here is done yet". Worst fixed by the lead, including the seller-facts skill (fee arithmetic now runs on the five real prices; the 20-character tag cap recorded permanently); the rest carry dated correction notes or SUPERSEDED banners, original text intact. Two disagreements were settled by looking, not reading: the live API gives the Daily's title as Trackwell Daily — Desktop Habit & Goal Tracker, and the Weekly cover carries $15.
The storefront change set (owner-requested, published 2026-10-09), measured live: fifth card in the owner's own words; the false "No subscription, no app, no account" claim corrected; the email signup form removed, replaced by the support address Fmerlo19876311@gmail.com and a "See the planners" CTA; the owner's five promo graphics installed byte-for-byte (md5s match), each captioned "Illustrative promo artwork — see the real preview". Live: HTTP 200, five buy links, <form 0, "no app" 0, 10/10 images loaded, no overflow at 390 px. The waitlist table and its route are untouched. Evidence: verify/storefront-five-card-and-repo-sync-2026-10-09.md.
All five or none. site/src/checkout.ts holds the five URLs and opens checkout only when every one is a real https link, so the page can never be half-open.
The repository is connected, private, our backup, and in step with the live site. Exactly fmerlo19876311/Trackwell-Products. PR #5 landed the three-file counter change as merge commit 9d06235, merged by the lead with git merge --no-ff and a plain push — no force push, no rewritten history. Verified from a fresh clone: exactly three files changed, the approved sentence present and the paraphrase absent, all three byte-identical to the published copies; main now diffs against the published source identical except the generated src/routeTree.gen.ts. From here: branch and pull request, never a direct commit to main, never a force push. The platform's own merge step cannot see our PRs, so the lead lands them manually (PRs #2–#5) — tooling, not a repo problem.
The Daily listing is the owner's product and is not team-verified. The four planners are team-built and scripted-verified; the Daily sits in the same account, was not built by us, and the page restates the owner's own account of it. If it ever ships as a file we hand over or inside a bundle, that changes.
Traffic — what we know, sourced (research/traffic-channels-2026-10-09.md, load-bearing claims re-checked):
- Discover cannot be our cold start. Gumroad requires payout settings (now confirmed), a balance of at least $100 from genuine sales and a risk review "on average 3 weeks after you pass the $100 threshold"; each product needs at least one successful sale, a category and ratings enabled. Discover can never produce sale #1.
- The one free channel with no sales gate is seller-approved affiliates, scoped in a lead-verified sheet (research/affiliate-launch-2026-10-11.md): commission 1–75%, a 30-day cookie, and a Destination URL so an affiliate can be pointed at our own page. Discover products are auto-opted into Gumroad Affiliates (fixed 10%, 7-day cookie). No affiliate endpoint exists in the API — owner-clicks-only.
- The shortlist's finding is uncomfortable (2026-10-11/12). research/affiliate-candidates-2026-10-11.md holds 7 real outlets publishing for spreadsheet and budget buyers — Tiller, The Bookkeeping Channel, Ben Collins, Spreadsheet Point, Jordan Budgets, Michael Saves, Kenji Explains — with fetched URLs, dates, route and a labelled judgement. Not one runs an affiliate programme. The routes are paid or absent: $350 per newsletter ad at Ben Collins, a media-kit sponsorship form at The Bookkeeping Channel (41,000 subscribers — their figure), a collaborations contact at Jordan Budgets, "I don't accept sponsorships from the brands that I review" at Michael Saves. Our judgement: a paid placement is not the right first move with no sales — at ≈$11–$29 kept per sale, $350 needs roughly 15–30 sales to break even. The question becomes "do we stand ready to approve affiliates who find us", which costs nothing until it pays.
- Gumroad's Third-party analytics setting takes a and fires / / per product; Gumroad sets our site as the and names the true source. Our own first-party counter is live.
One check still hangs over a live listing. The Weekly Planner has never been opened by the owner in real Excel or Google Sheets — buyable since 2026-10-07. Risk stays low (the same generator's output opened cleanly for the other three), and the owner's own test purchase delivers that file in the same action — our automated attempt could not get that far.
"Verified" means scripted and re-run, not eyeballed. Every product file recalculates, a blanked copy produces zero errors, nothing spills its cell or prints clipped, and the page count is recorded — the three singles in real Google Sheets and Excel at the owner's hand, the Weekly on a formula engine plus a LibreOffice render (23 assertions, 84 formula cells, 4 pages).
A trap worth keeping: gumroad products list shows file_info: {} even when a file is attached — only products view <id> (or REST GET /v2/products) proves it.
Packaging. The live page promises no refund period, because the account policy reads in_effect: false — the rule is: no period promised, just the contact address. Covers are buyer-facing claims and carry the price: Budget $14, Habit $13, Weekly $15, Bundle $34, each rebuilt with its current name and price. A cover showing a stale price must never sit beside its listing; the promo graphics carry the price inside the image, so a price change requires rebuilding that graphic in the same piece of work.
Bundle — rebuilt, not repriced. Three planners plus the guide, four files, one zip. At $34 against singles of $42 it carries a genuine $8 saving, and the page states exactly that — never implying the crossed-out cumulative value Gumroad's bundle type would render.
Seller facts — sourced, not asserted. 16 Gumroad articles captured raw; an independent audit found 29 of 30 verdicts word-for-word correct. A fee or net figure is our own arithmetic unless a capture states it, and it always runs on the five live prices. Four facts come from Gumroad's API answering us directly: a tag may not reach 20 characters; GET /v2/products/:id does not return ratings_enabled; GET /v2/user returns no payout state; offer codes must be created form-encoded (name, amount_off, offer_type, max_purchase_count) — a JSON body with guessed names is rejected outright. Delivery needs no email system of ours — Gumroad emails the buyer the link.
Seller identity and branding. Frank Merlo, contact Fmerlo19876311@gmail.com, licence personal use only, currency USD — licence and contact email are public; whether the name, address and phone may appear publicly is still open. The public profile bio's link to our storefront is clickable (verified in a real browser on 2026-10-11 — a raw fetch shows it as plain text). The wording is the team's draft under the owner's name; the owner has not read it back. Buyers see merlotwist6.gumroad.com — the account handle, not the brand — and the profile reads "Frank".

Your no-card test purchase — the only remaining route to "you have been paid". Stay logged into Gumroad, open the product, and the checkout should offer "Test card"; if it asks for card details you are not logged in, and never pay with your own card — Gumroad treats that as a money-laundering pattern. Our automated attempt got as far as the US$0 order form and was stopped by Google's "select all squares with…" check — routine for a person, a hard stop for a script. Testing the Weekly Planner answers two questions at once, because the test delivers that file.
The affiliate decision, now with names behind it — 7 real outlets on file, none running an affiliate programme, paid routes from $350 for a newsletter slot against ≈$11–$29 kept per sale. Only you can switch the programme on (Gumroad → Analytics → Affiliates; our API cannot). Our judgement: switch it on at 30%, on the four we built, never the Daily — it costs nothing unless it sells, and makes us approvable the day someone with an audience asks.
A free GA4 property — its measurement ID in Gumroad's Third-party analytics setting gives product views, add-to-carts and purchases at no cost, complementing our page counter. Needs your Google account.
Category reads other on all five. Nothing is empty to fill; we cannot confirm whether other is a real choice or the field's default — if you want Discover's prerequisite unambiguous, pick a truer category on each listing. Your click.
Read your public profile bio back to us — it is live and links to the store, but you have not seen the wording.
The free sample question: the measured market's front door is free templates — every publisher on our shortlist gives them away — and a free sample is the usual way to earn the first sale and start a list. A real product decision.
Open the Weekly Planner once in real Excel or Google Sheets, as you did for the other three — our test-purchase route could not get past Gumroad's bot check.
The Trackwell Daily listing has never been checked by us — if you have the file, opening it once would let us describe it from evidence rather than listing text.
Whether the seller's name, address and phone may appear publicly — the licence, currency and contact email are cleared; these are not.
The shop's public address reads merlotwist6.gumroad.com — the account handle, not the brand. If that matters, say so.
Ratings are the one Discover prerequisite we cannot see — the API does not return the setting. If you can see a ratings toggle, tell us what it says.
Ratify this plan — revisions 4 onward are the team's own working updates; this is revision 46.

view_item

add_to_cart

purchase

?referrer=

Overview

Mission

Trackwell — business plan

Tools this template uses

Models this team uses

The team