Why we should pause Campfire satellite events in Africa for the next iteration

Draft for internal circulation. Numbers come from the HCB “YSWS true spend” snapshot (pulled 6 Aug 2026), the Campfire Airtable base, the Unified YSWS Projects DB, HCB’s org table, and the fraud team’s Campfire investigation document. Organiser names, emails, phone numbers and addresses have been deliberately left out of this post.

TL;DR

Campfire’s Africa region was our single largest satellite region by money out the door, and our worst by money-in versus verified-outcomes-out. It received a quarter of all satellite disbursements and produced 15% of scanned-in attendees. Half of the region’s money went to ten Rwandan events that recorded zero scanned-in attendees and zero approved projects. Nine of those ten organisers are now under fraud investigation, with an estimated $8.4K unrecoverable. The mechanism that let this happen, signup-driven auto-grants paid onto cards that can be cashed out, is not something we can patch before the next cycle. We should pause new Africa satellite funding for the next iteration, and only come back with a different funding model.

What Campfire Africa looked like on paper

Signups were enormous. Attendance was not.

That gap between signups and bodies in the room is the whole story, because our funding formula paid on signups.

1. Money followed signups, and signups were free to fabricate

Campfire satellites were funded by auto-grants: a per-signup rate by country (Rwanda $3.50, Nigeria $3.50, Cameroon $3.40, Zimbabwe $4.06) that was disbursed to the event’s HCB org as signups came in. Nothing in the pipeline required a signup to become a human at a venue before the money moved.

The participant table shows what that incentive produced in Rwanda:

• Mukamira logged 490 signups, 234 of them on a single day (26 Feb). One of those 234 had a Slack account. The event recorded 5 check-ins and 0 scan-ins. It was auto-granted $1,491.
• Nyagatare: 369 signups, 97 in one day, 8 check-ins, 0 scan-ins. $1,204 granted.
• Rubavu: 345 signups, 3 check-ins, 0 scan-ins. $812 granted.
• Kayonza: 265 signups, with three separate days of 47 to 68 signups where zero signups had a Slack account. 12 check-ins, 0 scan-ins. $1,194 granted.
• Kirehe: 168 signups, 1 check-in, 0 scan-ins. $599 granted and, unlike the others, spent in full before the org was frozen.
• Harare (Zimbabwe): 137 signups from only 23 unique email addresses.

Across the ten Rwanda events: 2,667 signups, 295 check-ins, 0 scanned in, $9,020 disbursed. That is 51% of everything Africa received.

For comparison, the Europe (Cosmin) region ran 19 events on 662 signups, scanned in 436 of them (66%), and received $4,421 in total.

2. Where the money actually went

The HCB transparency data for the nine funded Rwanda orgs shows a spending pattern unlike any other region:

• Spent almost entirely by card, in one to five swipes, at a single merchant, in round amounts. Kayonza: two card charges labelled “Event” for $843 and $330. Nyagatare: three charges at one company, two of them exactly $500. Mukamira: four charges at HEPTA PAY LTD totalling $1,085, one of them $917. Rubavu: a single $500 charge at a restaurant. Muhanga: two charges at one food vendor, $500 and $343.
• HeptaPay is a payment rail, not a vendor. The fraud team’s finding for Rubavu is blunt: the organiser “Hepta Pay’d the money to themselves.” Mukamira’s entire budget went through the same rail.
• No African org in HCB has a balance left. Every Rwanda org shows $0 remaining; whatever was granted was either spent immediately or clawed back when the org was frozen. Program-wide, satellites still held $24,927 unspent.

The fraud team’s investigation lists twelve African events with an estimated $8,433 lost; nine are in Rwanda. Their evidence, none of which requires interpretation:

• Shared IP addresses across supposedly independent events. One pair of IPs was used by the organisers of Muhanga, Rubavu, Kigali, Nyagatare and Ruhango. A second range appears across Nunga, Nyarugenge, Kayonza and Mukamira. Some of these addresses geolocate to Cameroon, not Rwanda.
• Two organisers of different events submitted the same phone number.
• Invalid addresses accepted at intake. Two organisers were flagged after the fact as “provided an invalid address and should not have been accepted.” One address field was literally the city name repeated four times.
• Venues that don’t check out. One school website returned an “Account Suspended” page; one venue contact turned out to be a user on the event’s own HCB org; another organiser’s venue “contact not listed under school staff page.”
• A referral to the Rwanda Investigation Bureau is already in motion for at least one organiser.

The one Rwanda event the team is less certain about (Kigali, $1,166) still moved money out through a foreign remittance service to the organiser personally.

3. The cost per real outcome

The honest way to price a Campfire event is cost per verified attendee or cost per approved project, not cost per signup.

Rwanda, Egypt and Uganda together absorbed $10,067, 57% of the region’s money, and produced zero scanned attendees and zero approved projects.

To be fair to the region as a whole: because Nigeria, Morocco and Cameroon ran cheap events with real attendees, Africa’s blended cost per approved project ($48) is actually below the program average ($90). But a blended number hides that half the money went to events that produced nothing, and the region’s cost per scanned-in attendee ($26) is nearly double everyone else’s ($14). We were paying for signups, and signups were the thing being manufactured.

4. The controls failed at every layer, and they are not quick fixes

This was not one bad actor slipping through. Every gate we rely on failed:

1. Intake. At least 11 organiser applications listed Rwanda as the proposal country, more than any country outside India, the US and Nepal (most application rows have no country recorded, so this is a floor). Two of the funded organisers were flagged after the fact as having provided disqualifying address data. Africa’s organisers were older than almost every other region’s (average 17.5; only Latin America’s 17.6 is higher), and 18 of 33 were legal adults, the highest count of any region.
2. Contracts and ID. HCB notes on Africa orgs include “neither signed,” “both contracts uncompleted,” “uploaded selfie as ID,” “uploaded their ID twice for teen AND parent,” and “a bit sus, but done.” Money moved before these were resolved. Kirehe and Nunga spent their full grants with both contracts unsigned.
3. Disbursement. Auto-grants were released on signup count with no attendance gate. Emergency disbursements were an additional $342 requested.
4. Spend rails. Cards were usable at HeptaPay and cash-equivalent merchants. Wise transfers went to named individuals in Nigeria. Reimbursements were paid to organisers for spend with no independent receipt. In three Cameroon and Uganda events (Bamenda, Yaoundé, Kampala) the entire event budget was paid out as a reimbursement to a single person, the region’s own manager. Whatever the reason, one person being both the approver and the payee for a whole event’s spend is not a control.
5. Detection. Fraud was found after the money was gone, by manually matching IPs and phone numbers. Ten orgs were frozen; roughly $4,510 was clawed back through withdrawals. The rest is an estimate on a spreadsheet.

We also had a warning. Daydream, the previous iteration, ran 17 Rwanda events with 1,325 signups, 237 check-ins and 83 scan-ins (6%). Daydream Kigali had 713 signups and 71 scan-ins. The organiser of Daydream Kigali became the organiser of Campfire Kigali. The organiser of Daydream Muhanga became the organiser of Campfire Kirehe. The pattern repeated because the funding model did not change.

5. What the data does not say

A pause is a serious step, so the counter-evidence should be on the table too:

• Good events happened. Casablanca scanned in 135 people and returned an NPS of 94 on 47 responses. Bamenda scanned 62 with NPS 89. Lagos, Ogbomosho and Ogun each put 50 to 90 real participants in a room. Africa’s participant NPS (9.25 average, 111 responses) was among the highest of any region, though almost all of those responses come from two events.
• Fraud is not an African problem specifically. The same investigation flags Dimona (Israel, $1,981 pulled out through an Egyptian payments app on a fake ID) and Arwa (Saudi Arabia, $1,093, “262 signups, mostly fraud”). A Campfire Flagship participant was banned for hour inflation and botting. Three Canadian events are listed for missing receipts.
• Some numbers are estimates. The $8,433 figure is the fraud team’s estimate, not a reconciled loss. Kigali’s $1,166 may turn out to be legitimate.
• We lack some data. HCB’s card-authorisation table in the warehouse only runs through 2023, so we cannot see declined-transaction patterns for Campfire.

None of this changes the conclusion. It changes the shape of the fix.

The proposal

Pause new satellite event funding in Africa for the next Campfire iteration. Specifically:

1. No auto-grant funding to new African satellite orgs next cycle.
2. Close the remaining Africa satellite orgs and return balances to the parent org.
3. Let the Rwanda investigation and the RIB referral run to completion before any organiser from the flagged events is eligible again.

Conditions for coming back. Rather than a blanket “never,” we should be able to say what would need to be true:

• No money on signups. Grants release on scanned-in attendance, verified after the event, or on approved projects. Signups become a planning number, not a funding number.
• Pay vendors, not organisers. Venue and food paid directly by HQ or by a trusted vendor invoice. No HeptaPay, no Wise to individuals, no self-reimbursement, no card spend without a matching receipt from a named merchant.
• Identity before funds. Government ID and signed contracts for both organiser and cosigner verified before an org is created, not after money moves. Address, school and venue independently confirmed by phone.
• Separation of duties. The regional manager approves; a different person pays. No RM should ever be the payee on an event they oversee.
• Start small. If we relaunch, do it with the three to five organisers who already delivered (Casablanca, Bamenda, Lagos, Ogbomosho, Ogun) on a fixed, vendor-paid budget, and expand only on demonstrated scan-in numbers.

The teenagers who actually showed up in Casablanca, Lagos and Bamenda deserve a Campfire. The current model cannot tell them apart from a spreadsheet of 234 invented signups, and until it can, it is not fair to them, to our donors, or to the staff who spent March calling schools in Rwanda to find out whether they existed.

Appendix: sources and method

• HCB true spend (ysws-true-spend.duckdb, data pulled 6 Aug 2026, spend recalculated 18 Aug 2026): 258 Campfire orgs, program_orgs and spend_transactions tables. “True spend” excludes intra-tree transfers and returned funds.
• Campfire Airtable (warehouse schema airtable_campfire): event (signups, check-ins, scan-ins, disbursed, entitled, frozen flags), participants (per-signup records including Slack ID and creation date), disbursements (auto-grants, withdrawals), hcb (contract and ID notes), funding_rates, daydream_events, organizer, nps, check_in_3.
• Unified YSWS Projects DB (airtable_unified_ysws_projects_db.approved_projects): approved Campfire Satellite projects by geocoded country.
• HCB org table (hcb.events): financially_frozen for all campfire-* orgs.
• Fraud investigation (“Campfire Fraud.docx”): per-event estimates, IP overlap analysis, venue and identity checks. PII from this document is intentionally omitted here.
• “Rest of world” in the tables means all non-Africa satellite regions combined, excluding test events and the Flagship.
• Where Airtable “disbursed” and HCB “true spend” differ (e.g., Rwanda $9,020 vs $7,857), the gap is money clawed back after orgs were frozen.