Tacnode™
← The lineup

The Frozen Score

The nightly job scored this customer low risk at 2 a.m., and nothing has rescored them since.

Nightly job scores every customer. This one is low risk.

What’s true
Low
The feature store
Low
reads
Decision
Approve a large transfer to a new payee?
waiting

Illustrative figures.

Where this shows up

Account takeoverMost of the risk sits in the minutes after a credential change.
Instant paymentsThe money can’t be pulled back, and the score is from last night.
Buy now, pay laterRisk shifts with what’s in the basket.
Marketplace trustSeller and buyer risk read from last night’s features.

Why the feature store gets it wrong

02:00LOWthe score never changes in here12:10email12:11phone12:12new payee12:14transferAPPROVED ON A 10-HOUR-OLD SCORE

Feature stores compute features ahead of time on a schedule, which keeps serving fast. But each feature is only as recent as its last run, and they all inherit the same delay whether they can afford it or not.

By the time the decision ran, its context was outdated. We call that a context gap.

Score the customer on what they just did.

email changednew payeelimit raisedTACNODErecent changespayees · velocityper-feature freshnessTRANSFER CHECKHIGHHELDhold written back

Tacnode keeps the signals a transfer check reads current at the moment it runs, including recent profile changes, new payees and velocity. You decide per feature how fresh it needs to be: some are computed on demand, others sit in a view on their own refresh schedule. The hold is written back too, so the next check knows about it.

What comes inSource data arrives via change streams, and decision state is written to Tacnode.
What staysYour models don’t change. Tacnode supplies what they read.
What the decision getsThe check sees what the customer did a minute ago.
If an engineer asks how

Per-feature freshness control: on-demand against the latest committed data for the features that matter, incrementally refreshed views on a schedule you set for the rest.

Questions to put to your team

  1. How old is the oldest feature your transfer check reads?
  2. Which of those features can’t wait for the nightly job?
  3. What does your system do in the minutes after a credential change?

Is your transfer check telling this lie?

Bring it to a demo, and we’ll trace what it reads and how old that data is by the time it runs.

Book a demo →
← The Linked AccountsThe Stale Segment →