The OpenRate Power Score
The $100 million mistake every stablecoin card makes — and the signal that fixes it. We can predict, from a wallet's first few days, whether it will become one of the network's highest-value users.
Executive summary
Every stablecoin card issuer faces the same problem: thousands of wallets onboard each month — some become loyal, high-volume users worth thousands in lifetime payment volume, and others vanish after a single transaction. Until now there was no reliable way to tell them apart at onboarding.
Analyzing 66,807 mature wallets, OpenRate found the strongest predictors of a wallet's future value are already visible within its first few days. The Power Score predicts future power users with a held-out test AUC of 0.842(0.792 using behavioural signals alone). This isn't a statistic — it's a predictive intelligence system that lets issuers identify valuable users before they become valuable.
The signals, visualized
Power-user rate by early signal. Base rate = 10%; brand-coloured bars beat it.
First transaction size → future power user
Bigger opening spend, higher odds. Base rate = 10%.
Speed to 2nd transaction → future power user
The strongest behavioural signal. Same-day return wins.
First card → future power user
The onboarding door shapes the outcome.
Model score decile → actual power-user rate
Held-out test set. Dashed line = 10% base rate.
The second transaction matters more than almost anything
Wallets that return the same day become power users 15.1% of the time; wallets that never return, essentially never (0.1%). Engineering a meaningful second transaction may be the single highest-return onboarding improvement available.
| Time to 2nd transaction | Wallets | Power-user rate | Lift |
|---|---|---|---|
| Same day | 31,709 | 15.1% | 1.51× |
| 1–3 days | 10,289 | 11.0% | 1.10× |
| 3–7 days | 6,224 | 6.9% | 0.69× |
| 7–30 days | 8,297 | 3.2% | 0.32× |
| >30 days | 3,871 | 1.8% | 0.18× |
| Never (one-and-done) | 6,417 | 0.1% | 0.01× |
First transaction size is the strongest financial signal
A wallet whose first purchase exceeds $2,000 is 17× more likely to become a power user than one starting under $5. Partly mechanical (size contributes to lifetime volume) — which is why the behavioural-only model still reaches 0.79 AUC.
| First transaction | Wallets | Power-user rate | Lift |
|---|---|---|---|
| <$5 | 25,839 | 4.1% | 0.41× |
| $5–25 | 24,017 | 8.7% | 0.87× |
| $25–100 | 10,161 | 16.7% | 1.67× |
| $100–500 | 5,195 | 21.5% | 2.15× |
| $500–2k | 1,212 | 39.0% | 3.90× |
| >$2k | 383 | 68.9% | 6.89× |
The model — which combination predicts
| Feature (standardized) | Coefficient | Reading |
|---|---|---|
| First transaction amount (log) | +0.665 | Bigger opening spend → power user |
| Slow / never 2nd transaction | −0.598 | Slow return → strong negative |
| First card = Fiat24 | −0.583 | Micro-payment door → negative |
| Started Sunday / Monday | +0.372 | Early-week start → positive |
| First card = Gnosis Pay | +0.329 | Positive |
| Same-day 2nd transaction | +0.313 | Fast return → positive |
| First card = EtherFi | +0.262 | Positive |
| First card = Plasma | +0.122 | Positive |
- • Opens with a ≥$100 first transaction
- • Returns the same day
- • Enters via Plasma / Gnosis / EtherFi
- • Starts on Sunday or Monday
- • Sub-$5 first tap
- • No second transaction within a week
- • Enters via Fiat24
- • Never establishes an early habit
Future value isn't random. It leaves measurable fingerprints almost immediately.
Why this matters for issuers
- • Prioritize high-potential wallets for premium onboarding (40% hit rate in the top decile).
- • Reduce acquisition waste on a base that is 25% one-and-done.
- • Measure onboarding quality instead of raw user counts.
- • Compare channels by predicted lifetime value, not installs.
Methodology & limitations
Cohort: 66,807 wallets first seen Jan–Apr 2026 (tenure-controlled so every wallet had time to mature). Outcome: top 10% by lifetime settled volume (threshold $5,912). The model sees only first-days signals; trained on ⅔, scored on a held-out ⅓. Treasury wallets and automated activity excluded.
Limitations:the mature cohort covers four early card programs — the model will be retrained as newer programs mature. Custodial ecosystems whose spend isn't observable on-chain are excluded. Geographic features were omitted (low regional-attribution coverage today). First-card, first-chain and first-token are collinear and treated as one signal.
The OpenRate Power Score is the first in a family of predictive metrics for stablecoin issuers, payment networks and investors. See all OpenRate reports →