Stardance: funnel, cost, and weighted-project forecast
compared with Summer of Making 2025 and Flavortown

1Summary

A weighted project in the unified YSWS DB is approved hours divided by 10, exactly, in all three programs. Everything below is therefore an accounting of hours: hours tracked, hours shipped, hours certified, hours reviewed, hours landed, and dollars per hour landed.

At the same elapsed point, Stardance has produced more than either comparison program and landed less. On day 102 Summer of Making had 3,768 shipped projects and 1,627 approved; Flavortown had 2,217 shipped and 921 approved; Stardance has 5,355 shipped and 982 approved. Both earlier programs had converted about 46 percent of shipped hours into approved hours by day 102; Stardance is at 10.6 percent. The difference is not student output and not review capacity upstream. It is the unified-pass backlog: 2,506 projects that are ship-certified and through YSWS review, with adjusted hours already attached.

The unified pass does not change hours. Every Stardance row in the unified DB carries an override written by the Stardance app at sync time with the YSWS reviewer's adjustment. The 79 percent of logged hours that survive is decided inside Stardance's YSWS review. The model therefore treats the unified pass as a 100 percent pass-through and places the adjustment at the review stage.

On that basis the base case is 45,762 approved hours, about 4,900 approved projects and 4,576 weighted projects, with a range of 3,958 to 5,579. Stardance has spent $457k to date against $265k and $406k final for the earlier programs, and unlike them it spent most of it early: $115k before launch and $152k in July on hardware. Projecting $585k final spend, the base case costs $12.78 per weighted hour, about twice Summer of Making and slightly above Flavortown. The cost outcome is almost entirely a function of the unified pass: the same $585k buys $64 per hour at the current approval pace and $13 per hour if the backlog is cleared.

2What each stage means

The funnel stages are the same for all three programs because the three apps share one codebase lineage (Summer of Making, then Flavortown, then Stardance). Table 1 defines each stage in the order a student passes through it.

3Same-point comparison at day 102

Stardance started on 30 May 2026 and was 102 days in at the data pull. Table 2 places Summer of Making and Flavortown at their own day 102, counting only events timestamped before that point. Project counts exclude deleted projects in all three programs.

3.1What the earlier programs landed after day 102

A naive scaling says Stardance at 982 would finish around 1,700 to 2,050 projects, because the earlier programs were 48 to 58 percent done at day 102. That understates Stardance: it already has more ship-certified projects (3,471) than Summer of Making had at its end (3,421). The forecast in Section 8 therefore works from the pipeline.

4Full funnel

4.1Why Stardance converts differently per signup and per hour

• Signup mix. 45,587 of 60,359 signups are new to Hack Club (GPU raffle, AMD, NASA and GitHub campaigns). 12 percent of signups tracked an hour, against 13 percent for Summer of Making and 20 percent for Flavortown on the same basis.
• Less of the tracked time has shipped, so far. 49.5 percent of tracked hours sit on shipped projects, against 67 and 61 percent for the earlier programs at their day 102 and 79 and 83 percent at their end. 6,077 draft projects hold 70,085 hours; 31,192 of those hours are on 1,604 projects active in the last 14 days.
• Ship certification returns more, and returns are final. 1,595 shipped projects (30 percent) sit in needs_changes with 16,253 hours. Of 1,158 projects returned before 10 August, 38 (3.3 percent) later certified.
• YSWS review trims hours. 79 percent of originally logged hours survive review on landed projects. If review stopped adjusting, the forecast rises by about 1,000 weighted projects.
• The unified pass is the bottleneck. Sections 6 and 9.

5Where the approved count comes from

The count of 982 is what the unified DB holds under the program name Stardance. Table 6 lists every count between "shipped in Stardance" and "row in the unified DB", with the table it comes from.

All 981 matched Stardance rows have override_hours_spent set and equal to hours_spent. The justification text is generated by the app: "The user logged 10h 21min on hackatime. (This was adjusted to 7h 51min after review.) They had 32 commits, which compared to the original 10.4 logged..." The adjustment comes from the Stardance YSWS review; the unified pass copies it. The app's approved_minutes column is empty in the warehouse, so adjusted hours are visible only once a row lands, which is why the 79 percent ratio is measured on landed projects.

Two things this rules out. There is no pool of pending Stardance rows waiting for a date: zero Stardance rows have a null approved date. And Stardance projects are not leaking to other program names in volume: about 50 ship-certified Stardance projects were approved under Macondo, Horizons, Boba Drops, Stasis or beest, and Stardance-shipping users hold roughly 1,000 approvals under other programs for other projects, which is expected for a summer umbrella and is not double counting.

6Stardance pipeline and weekly dynamics

Student output is steady: 400 to 440 newly shipped projects and 11,000 to 12,300 devlog hours per week through August. Ship certification kept pace at 1,200 to 1,500 decisions per week in the last three full weeks, and YSWS reviews reached 555 and 644 per week. Unified-pass approvals peaked at 335 in the week of 24 August, then fell to 36 and 16. The reviewed-but-not-landed backlog grows by roughly 500 projects a week.

7Cost

Spend comes from the YSWS true-spend ledger (HCB data pulled 10 September 2026). True spend is money that left the program's HCB org tree for the outside world; transfers into personal author funds, intra-tree transfers and returns to HQ are excluded. Table 9 shows the composition. The unified DB's own total_spend figures ($355k, $412k, $469k) are gross outflows and overstate Summer of Making in particular, because $87k of its outflow went into personal author funds.

7.1Cost per weighted hour

Because weighted projects are approved hours over ten, cost per weighted hour is the comparable unit; cost per weighted project is ten times it. The earlier programs spent most of their money after day 102 on prize fulfilment, so their day-102 cost per hour is low and rises to the final figure. Stardance bought its hardware up front and its day-102 figure of $62 reflects $457k spent against 7,340 hours landed. What happens to that number depends on how much of the pipeline lands and how much more is spent on fulfilment.

8Forecast

8.1Model

Each pipeline stage is carried forward to approved hours with the rates below, then divided by ten. The unified pass is a 100 percent pass-through.

• Unified pass: 100 percent of projects, hours unchanged. Spot-check eliminations after landing have run at 6.4 percent of rows and are not deducted.
• YSWS-review hour adjustment: 79.3 percent of originally logged hours survive (7,312 of 9,220 on landed projects). Applied to every stage not yet reviewed, and to stage B because its adjusted hours are not visible in the warehouse until the row lands.
• Ship-certification pass: 74.6 percent of decided shipped hours (47,690 certified of 63,943) and 68.4 percent of projects (3,461 of 5,056). Applied to stage E and new ships.
• Hours at ship vs. current hours: 0.80 (18,343 / 22,845 on stage B); 0.90 for late ships that have little time to accrue.
• needs_changes resubmission: 3.3 percent.
• New ships, 9 to 30 September: two weeks at the last-four-week average (404 projects, 5,805 hours) plus a final week at 2.3x projects and 3.0x hours, between the Summer of Making (2.7x / 3.7x) and Flavortown (1.9x / 2.7x) final-week multipliers: about 1,740 projects and 29,000 hours. Cross-check: the last three weeks contributed 27.5 percent (SoM) and 33.5 percent (Flavortown) of all shipped hours, which applied to Stardance implies 26,000 to 35,000.

8.2Base case

8.3Sensitivity

9Throughput constraint

The forecast assumes the pipeline drains. Unified-pass approvals in the last four weeks were 255, 335, 36 and 16. Summer of Making and Flavortown both kept approving for three weeks past program end, so the comparable deadline for Stardance is about 21 October, six weeks from the data pull.

Because the pass is a check for pipeline bugs rather than a review, the practical question is whether it can be batched. The 2,173 projects in stage B are already reviewed with adjusted hours attached; landing them as a batch would move Stardance from 734 to about 2,060 weighted immediately and from $62 to $22 per weighted hour with no new spend.

10Final analysis

Output. Measured at the stages students control, Stardance is the largest of the three programs by a wide margin: 60,326 accounts against 24,180 and 12,283; 139,683 tracked hours on surviving projects at day 102 against 89,167 and 51,472; 5,355 shipped projects against 3,768 and 2,217. Its weakness at those stages is depth per signup, not volume: the raffle and partner campaigns brought in 45,587 people new to Hack Club, and 12 percent of signups tracked an hour. In absolute terms it still has more than twice Flavortown's engaged users.

Conversion. Stardance converts shipped work into certified work less well than the earlier programs (65 percent of shipped projects certify, against 79 and 83 percent) and its YSWS review trims hours harder (79 percent survive, against 98 percent in Flavortown). Those two effects together cost it roughly a quarter of its shipped hours relative to Flavortown's pipeline. They are real, but they are second-order next to the unified pass, where 2,506 projects and about 25,000 reviewed hours are waiting.

Forecast. With the unified pass as a pass-through, Stardance lands about 4,576 weighted projects, above Summer of Making's 4,381 and Flavortown's 3,595, from roughly 4,900 approved projects. The range of 3,958 to 5,579 is driven by two inputs: whether the end-of-program shipping surge materialises (the earlier programs shipped 22 to 25 percent of their projects in the last three weeks) and whether YSWS review keeps adjusting hours at 79 percent.

Cost. Stardance will be the most expensive of the three in absolute terms, at roughly $585k against $265k and $406k, because it bought hardware up front and ran a large print and marketing campaign. Per weighted hour it lands near $12.78 in the base case, twice Summer of Making and about 13 percent above Flavortown. That figure is entirely contingent on the pipeline draining: the money is already spent, so every project that does not land raises the unit cost. At the current pace the same spend yields $64 per hour, ten times Summer of Making.

What matters now. The single lever is unified-pass throughput. Clearing the 2,173 already-reviewed projects as a batch would land about 1,320 weighted projects at no marginal cost; sustaining 650 approvals a week through 21 October lands the rest. Two smaller levers exist upstream: the 1,595 needs_changes projects (16,253 hours) almost never return, and the 1,604 active draft projects with 31,192 hours are the pool the final-week surge has to draw from. Nothing in the spend data changes this ordering; the cost question is decided by the same bottleneck as the count question.

11Methodology and data notes

11.1Sources

• Hack Club data warehouse, airtable_unified_ysws_projects_db.approved_projects joined to approved_projects__ysws_name; fields hours_spent, override_hours_spent, override_hours_spent_justification, ysws_weighted_project_contribution, approved_at, code_url, record_id. ysws_programs for signups, dates and program totals. eliminated_projects for spot-check removals.
• summer_of_making_2025: users, email_signups, projects (is_deleted), devlogs, ship_events, ship_certifications (judgement, updated_at), ysws_review_submissions.
• flavortown: users, funnel_events, user_identities, projects (deleted_at), posts, post_devlogs, post_ship_events (certification_status, hours).
• stardance: users, materialized_all_signups, projects (ship_status, duration_seconds, deleted_at), posts, post_devlogs, post_ship_events (certification_status, hours_at_ship, hours_at_payout), certification_ship_reviews (decided_at), certification_ysws_reviews (original_minutes, approved_minutes, reviewed_at, returned_at, airtable_synced_at, in_unified_db, summary_justification), project_memberships.
• Spend: ysws-true-spend.duckdb (YSWS true spend, HCB data pulled 10 September 2026 16:33 UTC), tables programs, program_orgs and spend_transactions with counted_as_spend = true. Category labels from the warehouse's ysws_spend_ledger.

11.2Definitions

• Day 102 cut: program start_date from ysws_programs plus 102.6 days (Stardance's elapsed time at the pull): Summer of Making 26 September 2025 14:00, Flavortown 31 March 2026 14:00. Each stage counts only events timestamped at or before the cut; hours are devlogs created at or before the cut. Spend at day 102 is cumulative through elapsed week 14.
• Tracked hours: devlog duration_seconds on non-deleted devlogs; "surviving projects" further excludes projects deleted in the app.
• Shipped project: non-deleted project with at least one ship event. Shipped hours are the project's devlog hours.
• Ship-certified: approved judgement (Summer of Making), approved ship event (Flavortown, Stardance).
• Approved, weighted: rows in approved_projects with the program's ysws_name; weighted = sum of ysws_weighted_project_contribution = hours_spent / 10.
• True spend: the true-spend ledger's definition: external spend plus card grants to makers plus internal payments for real costs, excluding personal author funds, intra-tree transfers, returns to HQ and wash transactions.

11.3Caveats

• Repo-URL matching leaves 2 percent of Stardance, 8 percent of Flavortown and 14 percent of Summer of Making unified rows unmatched. Unmatched rows are in all unified totals and absent only from per-stage hour ratios.
• Flavortown has no certification timestamp, so its ship-certified count cannot be cut at day 102.
• Stardance's approved_minutes is empty in the warehouse, so the YSWS-review adjustment is inferred from landed rows. The sensitivity table brackets 73 to 100 percent.
• The final-week surge is the least certain input. Stage C averages 27 original hours per project and is carried at the same rates as stage B.
• Remaining Stardance spend is a run-rate estimate. Stardust balances not yet redeemed in the shop are a liability the ledger does not show; if redemptions cluster after program end, spend runs toward the $650k case.
• Approved dates in the unified DB sometimes precede the YSWS review date for the same project, so weekly approval counts by approved_at may not reflect when the pass happened.

Model: model_v2.py; page generator: build_site.py; spend file: ysws-true-spend.duckdb; all in /root/projects/stardance-forecast/. Revision 1 (9 September) assumed an 85 percent unified pass with hour cuts and forecast 3,860; revision 2 (9 September) introduced the day-102 comparison and the pass-through model; revision 3 adds cost, stage definitions and the final analysis. A PDF of this page is available.