For risk teams

DojiPad feeds your risk framework and replaces none of it.

• First line: traders and desk heads get the plan check and DJ before the order.
• Second line: risk and compliance see limit breaches, the desk overview and breach predictions.
• Third line: internal audit can read the audit log and rebuild it as of any date, and can't change anything.

DojiPad sees the plans traders state, so which decisions need a plan is a desk policy. How your team checks coverage today is under What DojiPad doesn't do yet.

You give DojiPad the document that sets out your firm's limits. DojiPad proposes each limit, ready to be set up in the system. Each limit's threshold, and what it is measured against, must come from the document's own text. If they do not, the proposal says the limit cannot be expressed. It never shows a guessed number. DojiPad does not guess.

Each limit, once set up, is either switched on or not. When a limit changes, the old version is kept, and each limit finding records which version it was checked against. Every exported PDF carries a stamp. The stamp lists which limits are monitored and which are declared not monitored. So a gap in coverage is stated openly, never left unsaid.

Futures are sized in contracts, using a verified record of contract multipliers. A plan that cannot be measured this way is refused, not estimated.

One group from the list of limit breaches, as the product renders it, with simulated data. What was breached comes first, then by how much, then who.

The desk overview has one row per trader. It shows open flags, the largest breach (as a multiple of the limit) and the last plan. The open-flag count is the same number as in the list of limit breaches. It means the same thing on every screen. Checking plans against what was actually traded needs the execution feed, the record of your filled trades. Until that is connected, those columns say plainly that they are unavailable. They are not estimated. Once the execution feed is connected, orders without a stated plan show on the desk overview.

Analytics summarizes the list of limit breaches, the briefs and the breach predictions. It shows open flags now and how often traders reshape a plan. It shows how many briefs traders write and how long acknowledgements take. A table per trader shows how closely each trader's predictions matched what happened. Limit breaches are also shown by type of limit. Every card states the time period it covers. Analytics only reads: it records nothing and defines nothing of its own.

This screen compares realized results with whether the plan was followed, by behavior pattern and by cost. It cannot link results to their causes until the execution feed is connected. The screen says so, and shows nothing rather than misleading figures.

A prediction is made each time a plan is committed. It is settled when the trader's next committed plan is checked against your limits.

Breach prediction is measured in two standard ways.

• Calibration compares the predicted rate with the actual rate over settled plans, for each trader. A plan counts as settled once the trader's next committed plan has been checked. The screen shows the two rates side by side. So you can see whether the number claims more than the record supports.
• Brier scores, a standard accuracy score for probability forecasts, measure how far the predictions were from what happened. They are kept in the detail view, next to the figures they are compared with, never shown on their own.

• No figure for a trader with fewer than three prior plans.
• No tiers until at least five traders have been scored.
• The firm's overall rate and the tier cut-off lines are shown at the top of the screen.

Each trader sees their own prediction and its trend, in their own view. They see nothing about any other trader.