The 90-Day Lead Vendor Audit

Most agents evaluate a lead vendor on a feeling formed in the first two weeks. The leads felt good, or they felt terrible, and the renewal decision follows the feeling.

Two weeks is not enough data to distinguish a good source from a lucky one, and “felt good” is measuring the wrong thing anyway. A vendor whose leads convert well and lapse at 40% is more expensive than one whose leads convert modestly and stay on the books — and you cannot see that difference until month four.

Ninety days is the minimum honest window, because that is how long it takes for the first cohort’s policies to issue, draft twice, and start telling you the truth. Here are the nine tests to run at the end of it.

Test 1 — Contact rate, by cohort

Right-party contacts ÷ unique leads delivered, measured per weekly cohort, with at least six dial attempts across different days and dayparts per lead.

Vendors quote this generously. Measure your own, and measure it per cohort rather than in aggregate, because aggregate hides the pattern that matters: a vendor whose week 1 leads contacted at 34% and whose week 9 leads contacted at 19% is degrading, and the average conceals it entirely.

Benchmark ranges by lead type are in connect-rate benchmarks. Before blaming the vendor, rule out your own caller ID with a fresh-number test — Spam Likely explains why that has to come first.

Test 2 — Wrong-number and disconnect rate

Track disconnected, wrong-party and clearly-invalid records as a percentage of delivered leads.

A rate above roughly 12% on fresh leads means the vendor is not validating. Above 25% and you are buying a list, not leads. This also directly damages your caller-ID reputation, so a dirty vendor costs you on two lines at once.

Test 3 — Issue rate, not close rate

Applications submitted is the number agents brag about. Applications issued is the number that pays.

Issued ÷ applications, per vendor. Final expense declines and postpones are common enough that a 15-point spread between sources is normal, and it usually reflects who the vendor is actually generating leads from — a source that produces uninsurable prospects can look excellent on close rate and terrible on income.

Test 4 — Cost per issued policy

Total spend on this vendor ÷ policies issued from their leads.

This is the primary metric. Everything above is diagnostic; this is the verdict. Run it in the cost per issued policy calculator, which is client-side and takes nothing.

A $12 lead with a 1.8% issue rate costs $667 per issued policy. A $28 lead with a 5.5% issue rate costs $509. The expensive lead is cheaper, and cost-per-lead comparison gets this exactly backwards every time.

Test 5 — 13-month persistency by source

The test everybody skips, because it requires waiting.

Policies still in force at month 13 ÷ policies issued, per vendor. You will not have this at 90 days for your first cohort — start tracking it now so that your next renewal decision has it.

This is where lead sources separate hardest. Two vendors with identical cost per issued policy can differ by 15 points of persistency, which is a 15% difference in money you keep. The mechanics of why, and what moves it, are in persistency and chargebacks. Your CRM has to be able to report chargebacks by lead source for this to be possible at all — see commission and chargeback tracking.

Cost per issued policy that survives 13 months is the only number that ranks vendors correctly. Every other metric on this page exists to explain it.

Test 6 — Exclusivity, verified rather than accepted

Exclusive leads cost a multiple of shared ones. The premium is a promise, and promises are testable.

Three tests:

1. Ask the prospect. On live conversations, ask plainly whether other agents have called about the same request. Log it as a field. Fifty conversations is enough to see a pattern.
2. Check delivery timing. Exclusive leads delivered eleven minutes after generation behave differently from ones delivered ninety minutes later. A delay is where a “sold once” lead gets sold again first.
3. Read the contract definition. “Exclusive” sometimes means exclusive to you, sometimes exclusive within your territory, sometimes exclusive for 72 hours. These are three different products at one price.

Full economics in exclusive vs shared final expense leads.

Test 7 — Consent records, requested in writing

For every lead, the vendor should be able to produce: the exact consent language shown, the source URL, timestamp, IP address, and the form or recording itself.

Request the full record for ten specific leads by name. Not a sample the vendor chooses — ten you pick.

If they cannot produce it within a few business days, that is your answer. You are the one who gets sued, not the vendor, and an indemnity clause from a company with no assets is a document, not a defence. What a record has to contain to survive summary judgment is in do bought leads count as TCPA consent, which sits in the compliance section and carries no product links by policy.

Test 8 — Replacement policy, tested

Every vendor has one. Test whether it functions.

Submit ten replacement requests for genuinely bad records. Measure: how long until credited, what percentage were approved, and what quality the replacements were. A replacement policy that returns worse leads than the originals is a cost-shifting mechanism, not a guarantee.

Test 9 — Delivery consistency

Plot leads delivered per day across the ninety days.

Lumpy delivery — 4 leads Monday, 31 Thursday — wrecks dial planning and answer rates, because 31 leads in a day means most get contacted late, and speed-to-lead is the strongest single conversion variable in the category. Volume you cannot work on time is volume you did not buy.

The renewal decision

Build one table. One row per vendor, one column per test.

Renew if:

• Cost per issued policy is at or below your target, and
• Consent records were produced on request, and
• Contact rate is stable rather than declining across cohorts

Renegotiate — do not simply cancel — if cost per issued policy is 10–25% over target but consent records are clean and delivery is consistent. Vendors move on price for accounts that measure carefully and say so, because those accounts are the ones that stay.

Cancel if consent records could not be produced, full stop. That one is not a price negotiation.

Doing this without heroics

The nine tests need exactly one thing your system must support: lead source recorded on every record, carried through to policy status and chargeback. If that field is missing or unreliable, none of this can be computed and you are back to feelings.

If your source data is a mess, fix that before the next contract rather than after — the field map in switching CRMs is written so a spreadsheet built to it supports every test above.

Related: aged final expense leads, direct mail and live transfer leads for how each type behaves under these tests.