How we build
The making is part of the product.
You can read the principles we say we follow on the principles page. This page is for the people who want to know what we actually do — week to week, month to month — so they can decide whether the principles are real or just well-written.
Our weekly cadence
At least one worker conversation
30 minutes with a real grants manager or development director. Customer, prospect, or friend-of-friend — direct conversation, not chat. Notes go into our customer-conversations folder, in the repo, public to our team.
At least one observed work session
Pure observation. 15-30 minutes watching a worker do the actual grant work — Loom recordings count. We're trying to see what they do, not what they say they do.
The 5-gate process
Every feature passes through: name 3+ workers it came from, write a measurable claim, get a prototype reaction, ship to 3 users for 2 weeks, then it's marketing-eligible. If a feature can't clear the gate, it doesn't ship.
One marketing-copy audit
One page or section of our marketing site, graded against the five authenticity tests. If it fails, we rewrite it or retire it.
Persona refresh + self-audit
Personas updated from the last 20 conversations. Old ones retired with notes. Plus a public self-audit of the product against the HCD framework — including what we got wrong.
Customer council
3-5 of our most engaged customers meet with us for 60 minutes. They tell us what's working, what isn't, what we're missing. They get $100 in credit + decision-influence. They are not focus-group fodder; they are co-designers.
The artifacts we publish
We try to publish the work, not just the product. If a customer wants to know how we make decisions, they can read the same documents we work from.
- HCD-FRAMEWORK.md
The internal working reference: 5 principles, 8 anti-patterns, weekly/monthly/quarterly cadence, 5 marketing authenticity tests.
- HCD-SELF-AUDIT.md
Our most recent quarterly grade against the framework. The May 2026 audit found 5 anti-pattern violations in our own product. We named them publicly.
- customer-conversations/
The interview protocol, the per-session note template, and the working persona documents (v0 drafts pending interview validation). Every product decision traces back through this folder.
- Pull request template
Every code change goes through this checklist. The HCD per-feature gate is enforced at merge time, not at marketing-launch time.
- Research desk
Three white papers on the market, on nonprofit burnout, and on HCD itself — published with full sources, reusable, citable.
What we've gotten wrong
We don't hide our corrections — we publish them. This list grows over time. Every entry is a thing we either claimed before we should have, built before we validated, or framed in a way that didn't respect our users. Dates show when we caught and corrected it.
- May 2026
Unverifiable "~4 hours/week reclaimed" claim
What was wrong: Our Solution copy said the average team reclaims ~4 hours/week per fundraiser. The number was extrapolated from a generic Lokalise SaaS study, not measured on actual customers.
The fix: Removed the claim. Replaced with "We're early in measuring time-saved with real customers — we'll publish what we find." Restores Test 5 (verifiable-claim) compliance.
- May 2026
Strikethrough-tools vignette (dignity failure)
What was wrong: Our Solution section had a list of tools the user already uses ("Grant tracker spreadsheet, Sticky notes, Drive folder...") with strikethrough styling. Read as "what you've been doing is wrong."
The fix: Removed the strikethrough. Reframed as "You've built this out of necessity. It works, but it costs you time and attention every week." Restores Test 4 (dignity) compliance.
- May 2026
AI-drafted reports advertised before built
What was wrong: The Premium tier on our pricing page listed AI-drafted reports + monthly opportunity refresh as if they were shipped. Neither was.
The fix: Added "Coming Q3 2026" labels. We commit to shipping by end of Q3 or honestly extending the date.
- May 2026
Hero demo card showed the buyer view, not the worker view
What was wrong: Our landing page demo card opened with three metric tiles ("Active grants 12 / Funded $184K / Tasks due 4"). That's an ED dashboard, not a grants-manager dashboard. The Hero text claimed worker-first; the visual contradicted it.
The fix: Rebuilt the card to lead with the next thing to do, then the upcoming week, then funder context. Aggregates removed from the lead. Same principle applied to the in-product dashboard.
- May 2026
Brand rename: "Coach House Grants" → "Bothy"
What was wrong: We launched as Coach House Grants. Two problems: the "Grants" in the name locked us into a grants-only positioning when the actual product is fund development (grants + donor stewardship + funder relationships). And we did not check trademark space first — "Coach House Grants" was generic enough to never be defensible if a similar player emerged.
The fix: Renamed to "Bothy" (a bothy is a basic working shelter — the metaphor matches our worker-first positioning + sustains the architectural family with the Coach House parent studio). Cleared trademark space first this time. Updated all product surface, operating docs, and SETUP. White papers keep their original publication-date bylines as historical record.
Hold us to this
If you see us drift — claim a feature we haven't shipped, advertise a benefit we haven't measured, or design for the buyer at the worker's expense — tell us. We'd rather get the email from a customer than learn it from a thread on LinkedIn three months later.
Send corrections to hello@bothy.so. They get tracked. They get acted on. They get published if material.
Next quarterly audit
August 2026. We'll publish the grade — including failures — within 7 days of the audit. If we slip the date, that's a correction on this page too.