Predictive Signals: know who to call first

Every loan officer and recruiter has the same problem on Monday morning: a long list of people they could call, and time for a few of them. Predictive Signals reorders that list for you. Each signal looks at the loan officer, company, loan and transaction history Model Match already tracks, and hands back the people to reach first, with the reasons in plain English.

Three signals, three questions#

Recruiting Radar ranks the loan officers active in a state by how strongly their history and their company’s recent movement point to a move to a new company in the next six months. You can also look up one loan officer’s priority.

Client Retention Risk takes one loan officer’s past clients, the properties they financed between six months and ten years ago, and puts first the ones most at risk of taking their next refinance or equity loan to someone else.

Relationship Pulse takes the buyer’s agents a loan officer has closed with and puts first the relationships most worth working toward another closed deal in the next six months.

Scores are recomputed on a regular schedule, so the lists move as the market does.

Why it matters#

For recruiters, the best time to start a conversation is before a loan officer is on the market. Recruiting Radar surfaces the people whose situation has changed, such as a short stint at their current company or a wave of departures from their team, so outreach starts with the people most worth a call.

For loan officers, your past clients are some of the best repeat business you have, and that book erodes quietly. Client Retention Risk tells you whose next refinance is most at risk of going elsewhere, so you can call them first. Relationship Pulse does the same for your referral partners: it notices when an agent you usually close with every couple of months has gone quiet.

For everyone, the reasons do the explaining. You don’t have to trust a number. Each row says why it ranks where it does, in sentences you could repeat on the call.

Reading a signal#

Every row carries the same three things:

Ranked lists also include rank (1 is the highest priority), and every row has scoredAt, the time its score was computed.

Get started#

Recruiting Radar: loan officers to talk to in a state#

curl "https://api.modelmatch.com/v1/signals/lo-move?state=CO&limit=25" \
  -H "x-api-key: mm_your_key_here"

import { createClient } from "@model-match/api";

const client = createClient({ apiKey: process.env.MODEL_MATCH_API_KEY });

const result = await client.signals.listLoMove({
  query: { state: "CO", limit: 25 },
});

const rows = result.data.data; // highest recruiting priority first

{
  "data": [
    {
      "nmlsId": "1234567",
      "name": "Loan Officer A",
      "companyName": "Example Mortgage Co.",
      "companyNmlsId": "7654321",
      "city": "Denver",
      "state": "CO",
      "score": 0.42,
      "percentile": 99.1,
      "reasons": [
        "About 20% of their company's loan officers moved to another company in the last 6 months",
        "Joined their current company about 8 months ago"
      ],
      "scoredAt": "2026-11-10T00:00:00.000Z"
    }
  ],
  "cursor": "eyJleGFtcGxlIjoiY3Vyc29yIn0"
}

For one loan officer, call GET /v1/originators/{nmlsId}/signals/lo-move (client.signals.getOriginatorLoMove({ path: { nmlsId } })). It returns the same row under data.

Client Retention Risk: past clients to call first#

curl "https://api.modelmatch.com/v1/originators/1234567/signals/client-retention?limit=25" \
  -H "x-api-key: mm_your_key_here"

const result = await client.signals.getOriginatorClientRetention({
  path: { nmlsId: "1234567" },
  query: { limit: 25 },
});

const clients = result.data.data; // most at risk first

{
  "data": [
    {
      "rank": 1,
      "propertyId": "example-property-id",
      "address": "100 Example Ave",
      "city": "Denver",
      "state": "CO",
      "zip": "80202",
      "lastLoanDate": "2021-06-15",
      "lastLoanType": "FHA",
      "score": 0.63,
      "risk": 0.63,
      "percentile": 98.4,
      "reasons": [
        "Their last loan was a home purchase loan, and buyers often refinance elsewhere.",
        "This was their first loan with this loan officer."
      ],
      "scoredAt": "2026-11-10T00:00:00.000Z"
    }
  ],
  "cursor": "eyJleGFtcGxlIjoiY3Vyc29yIn0"
}

risk carries the same value as score here: higher means more at risk of going elsewhere.

Relationship Pulse: agents to reconnect with#

curl "https://api.modelmatch.com/v1/originators/1234567/signals/agent-pulse?limit=10" \
  -H "x-api-key: mm_your_key_here"

const result = await client.signals.getOriginatorAgentPulse({
  path: { nmlsId: "1234567" },
  query: { limit: 10 },
});

const agents = result.data.data; // highest priority first, up to 50

{
  "data": [
    {
      "rank": 1,
      "agentId": "example-agent-id",
      "agentName": "Agent A",
      "agentLinked": true,
      "score": 0.37,
      "percentile": 99.3,
      "reasons": [
        "You usually close a deal together every 2 months. It has been 4 months since the last one.",
        "This is one of your top 3 agents by deals closed together this past year."
      ],
      "scoredAt": "2026-11-10T00:00:00.000Z"
    }
  ]
}

When agentLinked is true, pass agentId to the agent routes to pull the full profile. When it is false, the id only identifies the agent within this list.

Good to know#