EXP-009 Golden Flow Rewards

1 THE $1 REWARD ONBOARDED 118 USERS AT $1.36 EACH. FIX THE SECURE ACCOUNT FLOW TO UNLOCK QUALITY AND SCALE.

0

New users in 9 days

2.3x

Secured vs unsecured activity

126

Faced product issues while securing

$160

Spent of $1K (16%)

118 new users in 9 days. 76% of depositors were first-time senders. The incentive works.
Secured users are 2.3x more active. All bank transfers and card activity come from secured users.
Secure flow has product friction. 126 users (platform-wide) hit product errors (rate limiting, timeouts, key conflicts). 26 succeeded. Rate limiting alone affected 79 users.
Budget: $160 spent of $1K (16%). Cost per new user: $1.36.

Recommendation: Keep the reward. Fix the secure account infrastructure. Then use securing as a quality gate for payouts.

2 THE EXPERIMENT REWARDS BOTH SENDER AND RECEIVER $1 WHEN A CASH NOTE ONBOARDS A NEW USER

Mechanism

$1 depositor + $1 claimant on Golden Flow completion (claim plus wallet creation).

Budget

$100 to $1K. Started Mar 17. Increased to $1K on Mar 25. Still running.

Exclusions

WhatsApp bot wallet (caring_marmot_9046) excluded. Separate initiative.

Organic

80% organic. Ambassadors (ZAR's field team who onboard merchants and users) are 10% of depositors, 20% of notes.

So What

The reward's impact depends on the secure flow working. If users cannot secure their account, the reward fires but does not produce high-quality onboarding.

3 THE PAYOUT FIRES ON CLAIM, NOT ON SECURING. USERS GET $1 BEFORE THEY VERIFY.

Golden Flow mechanics: Sender creates a cash note and shares the link. A new user claims it, a wallet is created, and both sender and receiver get a $1 reward.
Secure screen appears after onboarding for Golden Flow users only. It is not part of the reward trigger.
Payout fires on claim, not on securing. The $1 reward has zero connection to the secure account step.

The reward is tied to onboarding, not security. The secure account screen is a separate product surface shown after wallet creation.

4 MOST CASH NOTES STAYED WITHIN THE SAME GROUP OF PEOPLE. THE NETWORK DID NOT GROW.

0

New users joined

0

Reached exclusively through other new users

118 new users joined during the experiment.
Only 4 of them were reached exclusively through other new users. The other 114 received at least one note from someone already on ZAR. The reward is not creating new paths. It is incentivizing existing ones.

Source: Metabase de-duplicated chain analysis

5 34% OF CLAIMANTS SENT MONEY BACK, MOSTLY THROUGH THE WAITLIST ONBOARDING FLOW

0

Sent back (of 118)

0

Cycled full amount+

0

Unique depositors involved

40 of 118 (34%) sent money back to their original depositor. Spread across 17 depositors.
35 of 40 cycled the full amount or more. Avg received $0.99, avg sent back $1.09.
Social reciprocity, not fraud. The waitlist hack cycling money back is designed product behavior.

Ping-pong is systemic but benign. Money cycles through 1-to-1 relationships as part of the waitlist onboarding flow.

6 RATE LIMITING BLOCKED 79 USERS FROM SECURING THEIR ACCOUNT Blocker

0

Rate limited

0

Total product errors (platform-wide)

23/118

Experiment-specific (19% secured)

Source: Amplitude product errors only, unique users. User errors (wrong phone, bad OTP) excluded. Users can hit multiple errors. Individual bars overlap.

These are product failures, not user mistakes. 126 errors are platform-wide (all users). 23 of 118 is experiment-specific. Rate limiting, key conflicts, and timeouts block users who are trying correctly.

7 29 NEW CLAIMANTS MET ALL CRITERIA BUT THE REWARD ENGINE NEVER FIRED Bug

0

Completed

0

Skipped (duplicate device)

0

No record

30 of 31 claimed after the experiment was live. Not a pre-launch edge case.
29 met all onboarding criteria but the reward engine never fired. Engineering investigation needed.
20 skipped users (per-user count). Primarily duplicate device detection.

Core flow bug. The system should create a payout record for every eligible claim. 31 users completed the flow and got nothing.

8 120 OF 158 DEPOSITORS WERE FIRST-TIME SENDERS

0

New depositors (76%)

0

Repeat depositors

0

Total during experiment

76% first-time senders. The reward activated a new sender pool, not just existing behavior.
38 repeat depositors active in both pre (Mar 8-16) and during (Mar 17-25) periods.
120 new depositors replaced 50 who stopped sending.

The reward activated new behavior. Three quarters of senders during the experiment had never sent a cash note before.

9 SECURED USERS HAVE 2.3X MORE TRANSACTIONS THAN UNSECURED

Secured (23 users)

Secured email or phone

2.3

Avg txns

26%

Zero activity

6 bank transfers. All from this group.
2 card activity. All from this group.

Unsecured (95 users)

No email or phone secured

1.0

Avg txns

36%

Zero activity

0 bank transfers. None in this group.
0 card activity. None in this group.

Securing is a quality signal, not just a security step. Every bank transfer and card transaction came from a secured user.

10 66% OF NEW USERS TOOK A MEANINGFUL ACTION WITHIN 7 DAYS

0

Meaningful actions (of 118)

66%

Engagement rate

0

Zero meaningful (34%)

Of the 118 new claimants, 78 did something real beyond just claiming (sent a note, bank transfer, card spend, customer exchange).
34% did nothing meaningful. 40 users with zero meaningful transactions.
Meaningful excludes reward receipts, reclaims, and merchant-side transactions.

Two thirds of claimants engaged. The reward is not just creating accounts. It is creating users who take real product actions.

11 96% OF NEW USERS ARE GOLDEN FLOW. ONLY 5 CAME THROUGH THE WAITLIST HACK.

0

Golden Flow (96%)

0

Waitlist Hack (4%)

Metric	Golden Flow (113)	Waitlist Hack (5)
Secured	20 (18%)	3 (60%)
Got payout	64 (57%)	3 (60%)
Avg txns	1.3	1.2
Became depositor	73 (65%)	2 (40%)
Zero activity	38 (34%)	2 (40%)

Hack cohort is too small (5) for meaningful comparison. But 60% secured vs 18% is notable. Waitlist hack users may be more motivated to complete account security.

Source: Metabase golden_flow_rewards_payouts + users

12 75 OF 118 NEW USERS SENT THEIR OWN CASH NOTE WITHIN 6 MINUTES

0

Sent their own note (of 118)

6 min

Median time to first send

75 of 118 new users (64%) sent a cash note after claiming. Most did it within 6 minutes. 89% within the first hour.
68 of those 75 had at least one note claimed by someone. 7 sent notes that nobody claimed.

The loop is fast. Most claimants who become senders do so within 6 minutes of their first claim. The bottleneck is not speed.

13 REWARD USERS MATCH BASELINE RETENTION BY DAY 7. THE FIRST-DAY DIP IS TRANSIENT.

Pre-experiment baseline: claimants from Mar 1-16 using the same Amplitude start event.
Pre Day 1: 45%, Day 7: 28%. During Day 1: 34%, Day 7: 27%. Reward users are less sticky early but normalize by week's end.
Day 1 gap is 11 percentage points. Day 7 converges to within 1 percentage point. Weekly retention is comparable.

Source: Amplitude retention chart (any active event after claim day)

14 REPEAT DEPOSITORS SENT 27% MORE NOTES AT LOWER VALUES. MEDIAN NOTE DROPPED FROM $5 TO $1.

38 depositors were active before and during the experiment. Pre period: Mar 8-16. During: Mar 17-25.

Metric	Pre (Mar 8-16)	During (Mar 17-25)	Change
Notes	160	203	+27%
Median note value	$5	$1	-80%
Unique claimants	88	109	+24%
New users reached	55	70	+27%

More people reached at lower per-note values. Repeat depositors are sending cheaper notes to more people. Median dropped from $5 to $1, expanding reach.

15 FIX RATE LIMITING AND THE NO-PAYOUT BUG. THEN CONSIDER GATING PAYOUTS ON SECURING.

Priority	Action	Impact
P0	Fix rate limiting. 79 users faced this issue, top failure cause.	79 users
P0	Fix the 31 no-payout bug. 29 met all criteria but the reward engine never fired. Engineering investigation needed.	26% of claimants
P1	Evaluate gating payouts on securing. The secure screen already tells Golden Flow users they earn $1 for securing. The screen is only shown to Golden Flow users (account created same day as claim).	Quality gate

Secured users generate all the high-value activity. Fixing the secure flow is not just an infrastructure fix. It is the path to higher-quality onboarding.