Suspicious account review

Classify account evidence against your team’s abuse policies and route cases for investigation or additional evidence.

Step 01

Define and Align

Review synthetic examples and correct the model’s fields. Explain the evidence and policy distinctions your reviewers use.

Output schema
RouteChoose one

The next review step supported by the supplied account evidence.

  • investigate
  • no_flag
  • request_evidence
SignalsChoose any

Observed policy signals; use insufficient_evidence when the required facts are unavailable.

  • repeated_unsolicited_messages
  • continued_after_notice
  • insufficient_evidence
Missing evidenceList of values

Specific records or context needed before the review policy can be applied.

ExplanationText

The evidence and policy conditions supporting the review route.

app.sutro.sh/functions/suspicious-account-review
Collecting feedback · 1/3
Apply the supplied account-abuse policy to the documented activity and select the next review step.
Example 1 of 3

A new account on a shared network

Model input
Review policy
Synthetic account-review policy. Send an account for investigation when verified logs show repeated unsolicited messages continuing after a restriction notice. New-account age or a shared office network alone is not an abuse trigger. Request evidence when reports cannot be checked because message content, consent context, or restriction history is missing. Base the route on observed behavior; this function does not suspend accounts.
Account summary
A one-day-old account uses a shared office network. No restriction notices are recorded.
Activity evidence
Reviewed messages are replies to existing customer requests. The supplied records show no unsolicited outreach or continued messaging after a notice.
Model output
Route
SignalsMultiple values
Missing evidenceMultiple values
Explanation

Step 02

Model Sweep

Compare models on the same expert annotations and choose the quality–cost tradeoff for your workflow.

Compare models on expert agreement and cost. The interactive chart uses illustrative measurements.

Illustrative sweep · select Value or Best to use that model in your function.

Step 03

Deployment

Call Sutro Functions like any other software. Swap the model as soon as something more performant is released.

Selected modelWaiting for sweep results.

Real-time

import sutro as so

input_data = {
    "review_policy": "Synthetic account-review policy. Send an account for investigation when verified logs show repeated unsolicited messages continuing after a restriction notice. New-account age or a shared office network alone is not an abuse trigger. Request evidence when reports cannot be checked because message content, consent context, or restriction history is missing. Base the route on observed behavior; this function does not suspend accounts.",
    "account_summary": "A one-day-old account uses a shared office network. No restriction notices are recorded.",
    "activity_evidence": "Reviewed messages are replies to existing customer requests. The supplied records show no unsolicited outreach or continued messaging after a notice."
}

result = so.run_function(
    "suspicious-account-review",
    input_data,
)

Batch

import sutro as so

job_id = so.batch_run_function(
    name="suspicious-account-review",
    data="suspicious_account_review.csv",
)
results = so.await_job_completion(job_id)

Apply your team’s review policy

Use the behaviors and evidence your team actually reviews. Teach what meets an investigation threshold and what requires more context, while retaining the reasons behind each route. These synthetic examples focus on documented activity and review decisions.

Ready to get started?

Book time with the team

Join the self-serve waitlist

What are your bottlenecks? (Select any that apply)