Upwork strategy¶
Living strategy folder for Jonny's Upwork JSS-building approach. Data-backed, sources cited, assumptions flagged.
Last updated: 2026-05-26
How to read this folder¶
Start with playbook.md if you want the operational rules to bid by today.
Start with todo.md if you want to know what to BUILD next.
Everything else is supporting evidence.
Files¶
Operational (read these to act)¶
playbook.md- rules to follow on every bid. Headline, timing, proposal structure, sign-off, reject list.packages.md- the 3-tier productised offer ($499 / $1,499 / $2,999), full spec, add-ons, what we never do.niche-portfolio.md- the 7 spec landing pages to build (verticals × tiers, build order, prerequisites).todo.md- dependency-ordered work plan across 5 phases. Mirrors the GitHub issues at https://github.com/madebyhustle/jobhunt/issues.tests.md- active A/B tests we're running on real bids.
Research / evidence (the why behind the rules)¶
proposal-research-2026.md- the underlying research (Upwork algorithm, proposal length, video, timing) with sources tiered by reliability.lp-deep-mine.md- what 100 real LP Upwork briefs ask for (verticals, tools, goals, features, sophistication signals).lp-deep-mine-sample.md- 30 actual brief samples for hand-scanning.brief-mining-report.md- older broader analysis (328 briefs, looser keyword filter, includes app-dev contamination - kept for reference butlp-deep-mine.mdis more accurate).brief-mining-data.csv- one row per brief from the broader mine.top-freelancers-analysis.md- 28 top-tier landing-page freelancer profiles (Western + global), rates, earnings, headlines, badges.top-freelancers-data.csv- one row per top freelancer.myths.md- claims I've debunked or downgraded based on real data. Read this before re-introducing any strategy that "sounds right".
Raw data captures (gitignored or local-only)¶
manual-search-page-*.html- manually-saved Upwork search pages (Cloudflare won't let us scrape directly).fl-debug-*.html- debug captures from failed automation attempts.search-page-1.html- first search page captured via cURL
Source quality note¶
Most freelance-tooling sites (GigRadar, SnipeWork, Getmany) sell proposal optimisation services and have a commercial bias toward "you need to bid more / faster / with our tool". Their underlying data is real but their conclusions skew. The Upwork-official source (upwork.com/resources) is the cleanest signal. Where multiple biased sources agree directionally, the signal is stronger.
Standing rules (from memory)¶
- Prove with data or flag as assumption. Every claim in these docs must cite a URL or be marked
[ASSUMPTION]. Source:memory/rules_prove_or_flag.md. - Supply-side vs demand-side data must be distinguished. What freelancers headline (supply) is a different question from what clients ask for (demand). For positioning, demand matters more. Source:
myths.md. - Push back on Jonny when his logic looks off. Don't capitulate just to agree.
- Test before committing. Untested tactics get a 5-10 bid trial before being added to the playbook.
- Upwork tone: plain Derry-bloke register, no consultant flourishes. Source:
memory/upwork_tone.md. - Currency: all prices in USD (Upwork's currency). Upwork takes a flat 10% service fee on freelancer earnings.
What's in the parent project¶
The Upwork strategy lives in strategy/upwork/. Related project files:
config/portfolio.md- inventory of what we have / don't have / reject; read by the scorer.config/criteria.md- scoring rules read by the scorer.references/landing-page-conversion-design.md- the conversion-design reference doc (system prompt for Lovable).references/landing-page-copy.md- copy reference (Phase 0.2, in progress).references/trust-signals.md- trust-signal reference (Phase 0.3, not started).references/client-research-method.md- client-research methodology (Phase 0.4, not started).references/vertical-patterns.md- per-vertical patterns (Phase 0.5, not started).log/applications.md- log of every bid + outcome.