Abstract double-entry ledger — two columns of figures aligning, a thread of light linking a debit to its credit
EngineeringMay 28, 20269 min

Inside the instant settlement rail

How a charge becomes spendable balance in 1.2 seconds — the pre-funding model, the risk math, and what we gave up to get there.

By Akseli Virtanen · CTO, co-founder

A card charge clears networks, acquirers, and clearing windows before the money is really yours — usually a day or two later. Ours lands as spendable balance in 1.2 seconds. This is how, and what it costs us.

When a customer taps Pay, two very different things happen. The first is fast: the card network authorizes the charge in tens of milliseconds. The second is slow: the actual money crawls through clearing files, batch windows, and bank cut-off times, arriving in the merchant’s account one to two business days later — longer over weekends.

Most platforms paper over the gap with language. The dashboard says “paid,” the books say “receivable,” and somewhere a finance person maintains the spreadsheet that reconciles the two. We decided the gap itself was the product problem worth solving.

Why settlement is slow

Card settlement is slow for institutional reasons, not technical ones. Networks clear in daily batch cycles. Acquirers net their positions against each other once per cycle. Each hop has a cut-off time, and missing one means waiting for the next. The honest version of “instant settlement” is not making the networks faster. It is deciding the network delay is our problem instead of yours.

The pre-funding model

When your charge authorizes, we already know — with high confidence — that the money will arrive in our settlement account in T+1 or T+2. So we credit your Lumina balance immediately from our own pre-funded float, and we collect from the network when the batch lands.

In effect, we extend every merchant an interest-free, fully-collateralized loan with a tenor of about a day, thousands of times per second. The float requirement scales with volume, which is why “just pre-fund it” is easy to say and expensive to do: the float is capital that earns nothing while it waits.

Pricing the risk

The risk in the gap is narrow but real: the network could claw back the batch, the charge could be reversed before clearing, or we could have authorized a charge that the issuer later disputes. Our risk engine scores every charge at authorization time on reversal probability, per-merchant exposure, and concentration limits.

Charges that score clean — about 97% of volume — settle instantly. The rest settle on the network schedule, and the API tells you which path each charge took and why, in the settlement.mode field. No silent downgrades.

What we gave up

Three things, deliberately: capital efficiency, since the float is a standing cost; universal coverage, since the 3% of charges that settle slow are the exotic ones; and simplicity of explanation. “Your money, immediately, minus a precisely-stated risk model” is a longer sentence than “T+2.” We think it is the better sentence.

More reading