Should you fire your lead-gen vendor?

If the same ads have carried most of your spend for two months with nothing meaningful launched beside them, the account is coasting. On Meta, creative is the biggest lever a vendor controls. When creative stops moving, active management has stopped, whatever the monthly report says. It is rarely laziness. Winning ads feel safe, every new test risks this month's average, and the pipeline dies one skipped test at a time.

Verify it yourself: open Ads Manager, filter to active ads, sort by amount spent over the last 30 days, and check the creation date on the top five. If they are all 60+ days old and nothing launched in the past month has taken meaningful spend, you have your answer. Ten minutes.

If every test in the account started as your idea, you are paying for execution and doing the strategy yourself. The strategist you met during the sales process is not the person in your account each week, and the person in your account is measured on keeping the numbers stable and the client calm. Executing your idea transfers the risk to you; if it fails, it was your idea.

Verify it yourself: search your shared channel or email thread for "test," "idea" and "what if" over the last 30 days. Count how many started with them and how many with you. Worse than one of theirs for every two of yours and the strategy is running the wrong way. Fifteen minutes.

The leads in the vendor's report should reconcile to contacts in your CRM and bookings on your calendar. If the report has never shown the platform's count next to yours, the check has never been done, and the report describes what the platform thinks of itself rather than your business. The cause is structural, not dishonest: reports get built from the platform's exports, because that is what exports. Why the counts can never match.

Verify it yourself: take the last full month. Put Meta's lead count next to your CRM's contacts from those ads and your calendar's bookings. If the report has never shown those three side by side, this signal holds. Ten minutes, three tabs.

A vendor who cannot split leads into booked, attended and closed cannot tell you whether they are filling your calendar or filling your CRM with names. The blended number is always the more flattering one, and a report built on it will improve while the calendar stays flat. This is the signal to act on first, because it moves real money: the cheapest leads to buy are the ones least likely to book.

Verify it yourself: open the last vendor report and look for booked, attended and closed by source. Then open your calendar for the same month and count. If the report does not contain the split, the vendor is not steering by it. Ten minutes.

A managed account should have a testing agenda you never asked for: a new angle, a new audience, a new page, a shorter form, a faster follow-up. If nothing ships unless you push, you have a caretaker, not a growth partner. The incentive math explains it: tests cost hours and risk this month's reported number, holding steady costs nothing and keeps the scoreboard green.

Verify it yourself: open the last three monthly reports or call recaps and highlight every test the vendor proposed unprompted. Zero to one across three months is your answer. Note how many "tests" were your own messages coming back as plans.

You should own all five and hold admin access to each today, not "on request." If any of it sits inside the vendor's Business Manager, their CRM or their booking tool, a switching cost is being built into the relationship, whether anyone meant it or not. It starts as a convenience in month one and becomes leverage by month twelve: leaving means walking away from pixel history, contacts, booking data and sometimes the pages themselves.

Verify it yourself: open Business Manager settings and check who owns the ad account and the dataset, and what your role is. Then check whose login the CRM and the booking calendar run under. If ownership sits with the vendor, or you cannot see the screen at all, this signal holds and it is also your first fix. Ten minutes.

Account-level cost per lead tells you very little until you see cost per booked and attended appointment beside it. A loose form, a broad audience and a lead-optimised campaign will drive cost per lead down for years while the calendar stays flat, because Meta gets very good at finding people who submit forms, which is not the same population as people who book an estimate and turn up. No villain is required. A vendor graded on leads optimises for what the grade rewards.

Verify it yourself: divide last month's spend by leads, then by bookings, then by attended appointments. If cost per lead fell over the last quarter while cost per attended appointment rose or held, volume is carrying the number. Fifteen minutes.

Ask what the account can afford to pay for a booked appointment. The answer should be instant and specific, and it should come from your revenue per client, your margin, your close rate and your show rate. If it is not, every budget decision so far has been a guess. A target without an allowable under it is not a target; it is a preference. Most vendors never ask for those four inputs, not from neglect, but because the number they are graded on works fine without them.

Verify it yourself: work out your allowable with the Allowable Cost per Appointment Calculator. Then search the vendor's reports and proposal for that number or anything like it. If it appears nowhere, the account has been steered without it. Fifteen minutes.

If responses only speed up after you escalate, you have already learned your real priority level. This is capacity math, not character: vendors stack accounts, and attention flows to the largest, the loudest and the newest. If you are none of the three, you get the queue. The gap between what the sales call promised and what the shared channel delivers is the signal.

Verify it yourself: time the responses to your last ten non-urgent messages and compare them to the first month of the relationship. Then list everything you have had to ask for twice. If the list is long and the response times have doubled, you are not imagining it. Fifteen minutes.