Enrollment yield is the only number most admissions teams report. The problem: it lags 6–12 months behind the work that caused it. By the time yield moves, the agent has been live for two seasons and you've already made staffing decisions on bad data.

Here are the seven KPIs that actually let you manage week-to-week.

Quick answer

The 7 KPIs to track on an AI admissions funnel: inquiry-to-first-reply time, qualified-lead rate, application-start rate, application-completion rate, advisor-call show rate, deposit conversion rate, and net enrollment yield. Yield alone lags 6–12 months behind the work that drives it; the other six let you see what's working sooner. The most reliable leading indicator of yield is application-start rate, a sustained lift there in the first month usually means the agent is doing its job.

⚠️ A note on the target ranges below. The percentage and timing targets in each KPI section are educated estimates drawn from typical admissions funnels, not from a published industry benchmark. Treat them as starting expectations to compare your own data against, not as universal standards.

KPI 1: Inquiry-to-first-reply time

What it measures: seconds between inquiry submission and the agent's first substantive reply.

Target: <60 seconds (top quartile <10 seconds).

Why it matters: It's the earliest leading indicator in the funnel. Every other KPI degrades when this one creeps up.

KPI 2: Qualified-lead rate

What it measures: percentage of inquiries that complete the agent's qualification flow (typically 1–3 questions of program fit, intake, geography).

Target: 30–45% for undergraduate programs; 25–35% for high-tuition programs.

Why it matters: Distinguishes traffic quality (top of funnel) from agent effectiveness (middle of funnel).

KPI 3: Application-start rate

What it measures: percentage of qualified leads who initiate an application within 30 days.

Target: 15–20% of all inquiries start an application.

Why it matters: This is the earliest reliable predictor of enrollment yield 6 months out. A sustained lift here in the first month is usually the cleanest signal that an agent change worked.

KPI 4: Application-completion rate

What it measures: percentage of started applications that get fully submitted.

Target: 55–70% (varies by program; lower for selective).

Why it matters: Catches drop-off in the document collection stage. AI agents that send timely follow-ups bump this number 10–20 points.

KPI 5: Advisor-call show rate

What it measures: percentage of booked advisor calls that the prospect actually attends.

Target: 65–80%.

Why it matters: Tells you whether the agent is qualifying for real intent or just booking. A high book rate with low show rate = agent is over-promising.

KPI 6: Deposit conversion rate

What it measures: percentage of submitted applications that result in a paid deposit.

Target: 25–45%.

Why it matters: First real revenue signal. Predicts enrollment yield with about 90% accuracy.

KPI 7: Net enrollment yield

What it measures: percentage of admitted students who actually enroll.

Target: wildly variable by program (5% for highly selective, 60% for open-access).

Why it matters: This is the number leadership cares about most. But you only learn it months after the work is done, too late to act on. Use the other 6 KPIs for decisions, this one for reporting.

The 2 leading indicators that predict yield 90 days out

If you only track two of these, track these:

1. Application-start rate week-over-week: Moves first when the funnel improves or breaks.

   
   

2. Advisor-call show rate: Moves first when qualification quality changes.

   
   

Together these two predict enrollment yield 6 months out with ~85% reliability in with the data gathered.

A/B testing agent changes without breaking reporting

Three rules:

1. Test one variable at a time (greeting, qualifier order, handoff timing).

   
   

2. Run for at least 200 conversations per arm before calling a winner, less than that is noise.

   
   

3. Lock the metric in advance: If you're testing greetings, the metric is reply rate at message 2, not yield 6 months later. Match the metric to the variable's expected impact.

   
   

The biggest agent-side wins we see in 2026: rewording the first qualifier (lifts qualified-lead rate 8–15%), adding a delay before the first proactive follow-up (lifts reply rate at the cost of speed), and tightening advisor-handoff criteria (lifts show rate 20+ points).

Frequently Asked Questions

How long until I see KPI movement after deploying an agent?

Each KPI moves on its own clock. First-reply time drops within hours. Qualified-lead rate stabilizes in 1–2 weeks. Application-start rate takes 3–6 weeks. Enrollment yield won't move for 6+ months. If a KPI isn't moving on schedule, treat it as a diagnostic, each non-movement points to a specific broken piece.

What if my CRM can't join WhatsApp conversation IDs to applications?

Three options. (1) Add a "WhatsApp Conversation ID" field to your CRM contact record, your agent vendor populates it via webhook. (2) Use the prospect's phone number as the join key (less precise but workable). (3) Manual weekly export-and-match as a stopgap. Without one of these, none of the other KPIs are attributable.

Should I track cost per stage?

Yes. Track four costs in funnel order: cost per inquiry, per qualified lead, per booked advisor call, per enrollment. Rising cost per inquiry with stable cost per enrollment = your acquisition channel is getting expensive. Stable cost per inquiry with rising cost per enrollment = funnel leakage.

Are there any vanity metrics to avoid?

Three to skip. Messages-per-conversation: high counts often mean confusion, not engagement. Bot-uptime: meaningless if no one's chatting. Total-conversations: grows with traffic, says nothing about quality. Track conversion rates at each funnel stage instead.