You accept collateral from clients at one haircut and post it to venues at another. That difference is a spread like any other — and concentration, wrong-way exposure and cheapest-to-deliver all move it. Safetifi prices both sides from the same asset registry the risk engine uses, so an asset cannot be worth one thing on the margin page and another on the collateral page.
For every eligible asset, Safetifi shows the haircut you grant your client against the haircut you receive from the venue holding it, and the spread between them. A negative spread means you are financing the difference yourself, which is invisible in a one-sided haircut table.
Eligibility is a policy per client, not a single global list. Allow-list or deny-list modes are supported, and a client with no policy inherits the registry default. Concentration limits and a collateral wrong-way test run against the same book, so collateral that correlates with the exposure it secures is flagged rather than counted at face value.
When several assets satisfy the same requirement, the cheapest one to deliver is rarely the one already posted. Safetifi ranks eligible assets by what they actually cost to pledge, and models a substitution before you move anything — so the effect on LTV, concentration and the haircut spread is known in advance rather than discovered afterwards.
A liquidation horizon has two parts, and most systems model only one. Before a single order can be placed there is mobilisation — the custodian withdrawal SLA, the exchange transfer, the multi-sig ceremony. A 24-hour horizon against a T+1 custodian is already spent, and it reads on screen as a comfortable 24 hours. Safetifi records where each client’s collateral is held, with the mobilisation time of that location, and treats anything it cannot reach inside the horizon as stranded rather than usable.
Loan-to-value is tracked continuously against call and liquidation levels with headroom shown in currency, not just as a ratio. Margin calls are the point where collateral management stops reporting and starts acting, and they carry an owner and a deadline.
Deciding which assets a firm accepts as collateral and at what haircut, tracking their value continuously against the exposure they secure, and managing calls, substitutions and concentration — while also tracking the haircuts the firm itself is granted on the collateral it posts to venues.
It is the difference between the haircut you grant a client on an asset and the haircut you receive on that same asset from the counterparty or venue holding it. A positive spread is margin of safety you are paid for; a negative one means you are absorbing the difference.
Collateral is valued from the same asset registry and live marks the risk engine uses, including FX and stablecoin pricing, so a de-pegged or thinly traded asset is not carried at par on the collateral page while the risk page knows better.
Because a liquidation horizon has to cover mobilisation as well as execution. Assets behind a 24-hour custodian cannot be sold inside a 24-hour horizon, so they are stranded rather than usable — and the credit cap they were supporting shrinks accordingly. The same assets at the same prices on a one-hour venue account produce a materially larger usable figure.
It is counted as mobilisable and reported as a named assumption. Excluding unplaced collateral would zero almost every cap on the first day — defensible, and useless. Naming the unknown creates pressure to record it instead.
A pilot connects one channel against a slice of your live book and reports what it finds — client by client, product by product.