How to judge a paid lead source: defects, judgments, samples
One kind of problem is worth reporting on a single lead. The other kind means nothing until you have five in a row, and knowing which is which is most of the skill.
Every problem with a bought lead falls into one of two categories, and they have opposite reporting rules. One kind is provable from a single example and should be raised the hour you see it. The other kind cannot be established from one example at all, and raising it on one wastes everybody's time including yours.
Sorting a complaint into the right category before you make it is most of the skill in buying leads, and almost nobody is taught it.
Defects: one is enough
A defect is checkable against the record itself. You do not need a second opinion, a sample, or a week of data, because the lead either has a phone number or it does not.
Defects. Provable from the record.
- No phone number
- A number that is malformed or plainly invalid
- A duplicate of a lead you already had
- The wrong product
- The wrong state
Name the lead and give the email, so the record can be found. One is enough.
Judgments. Not provable from one lead.
- The leads feel unqualified
- They seem to want a different service
- Something does not fit and you cannot name it
- The intent seems weaker than you expected
Five in a row with the same shape can be traced to a campaign. One cannot.
Why a judgment needs a run
A single odd lead is not evidence of anything. Some proportion of any batch will have answered a form in a way that does not survive contact with a phone call, and that is true of every source in every market. It is priced in, on both sides.
Five consecutive leads with the same problem is a different object entirely. It is a pattern, and a pattern is traceable: it can usually be tied back to one campaign, one placement or one piece of creative, and that thing can be switched off. That is why the number matters. It is not a threshold for being allowed to complain, it is the point at which the complaint contains enough information to act on.
The practical version
Keep a short note of the leads that felt wrong and in what way. If you reach five with the same shape, you are no longer reporting a feeling, you are reporting a pattern, and you can say what it is: five in a row asking about home services rather than a refinance, or five in a row from the same state with the same mismatch.
That is a message a vendor can do something with within a day. "The leads seem bad lately" is not.
Why delivery is spiky, and why that is not a fault
The second most common misreading of a paid source is expecting it to arrive evenly. An order gets spread across a month and the spend is paced to match, which sets an average. It does not set a daily rate, because the advertising underneath it does not work that way.
A longer gap usually has a specific and boring cause. A piece of creative stops performing, it gets switched off, and the next one takes a few days to produce and a few more to find its footing. During that stretch the daily number can be nothing at all while the underlying order is entirely healthy.
The useful habit is to judge the fortnight rather than the Tuesday, and to ask about a gap rather than reading it as a verdict.
What a pause actually costs
Pausing feels free and is not. A running campaign accumulates delivery optimisation that a pause discards, so restarting is closer to a relaunch than to lifting a needle off a record. A short pause can cost a disproportionate share of a month.
Which makes the calculation simple in both directions. A holiday when nobody will be calling is worth a pause, because leads arriving at an empty desk are worse than no leads. A slow week is not, because pausing in reaction to it makes the next week slower too.
When the verdict is actually due
Later than most people take it. Across our own book the stretch from a delivered lead to a funded loan runs roughly three weeks to three months, depending on the product and the borrower.
Three weeks to three months is our own observation across our own book, not an industry benchmark, and it is a range because it genuinely varies by product. What the public data does and does not support at each stage of this funnel is set out in mortgage lead conversion rates.So a verdict taken at the end of month one is mostly a verdict on your own follow-up, because the outcomes have not happened yet. What is measurable that early is conversations: how many people you actually reached, and what they said when you did.
That is also the more useful thing to measure, because it is the half you control. If the conversations are happening and going nowhere, that is a real signal about the source. If the conversations are not happening at all, the first thing to rule out is a filtered text or a flagged calling number, both of which look exactly like a bad batch from your side of the desk.
Four things to record, or the judgment is not worth making
- Attempts per lead. Without it, "unreachable" is an impression.
- Time from delivery to first call. The one number most likely to be the actual problem.
- Which number you called from. So a labelled number shows up as a pattern rather than as bad luck.
- Why a lead was closed out. In words, not a status code.
With those four, a complaint about lead quality and a complaint about your own process are distinguishable. Without them, they look identical from the outside, and the vendor cannot tell them apart either.
The short version
- Defects are provable from one lead. Report them the hour you see one.
- Judgments are not. Wait for about five in a row, then report the pattern rather than the feeling.
- Zero days and double days are the same fortnight. Judge the fortnight.
- A pause discards optimisation. Use one for a holiday, not for a slow week.
- The verdict is due after the sales cycle. Before that, measure conversations.
The calling process the judgment depends on is in the follow-up schedule. Buying questions rather than operating ones are on our FAQ.
Questions this page answers
By whether one example is enough to prove it. A missing or malformed phone number, a duplicate, or a lead in the wrong state or the wrong product is checkable against the record itself, so a single one is a fact and worth reporting immediately. Anything that rests on judgment, such as the leads feeling unqualified or seeming to want a different service, cannot be established from one example. That kind needs a run of about five in a row before it means anything.
One, if it is a defect: a missing phone number, an invalid number, a duplicate, the wrong product or the wrong state. Around five in a row, if it is a judgment about quality or intent. The reason is not politeness, it is that five consecutive leads with the same problem can usually be traced back to one campaign or one source and switched off, whereas a single odd lead gives nobody anything to act on.
Yes, and expecting a flat daily rate is the most common way to misread a paid source. Delivery arrives in bursts because the advertising underneath it does. Zero days sit next to days at twice the average, and a longer gap usually means creative was switched off and its replacement is still being tested rather than that anything is broken. Judge the fortnight, not the Tuesday.
Usually more than people expect. A live campaign accumulates delivery optimisation that a pause discards, so restarting is closer to a relaunch than to lifting a needle. A short pause can therefore cost a disproportionate share of the month. Pauses are worth it for a genuine emergency or a holiday when nobody will be calling, and are rarely worth it as a reaction to a slow week.
Longer than one month, because the sales cycle is longer than one month. Across our own book the stretch from a delivered lead to a funded loan runs roughly three weeks to three months depending on the product and the borrower, so a verdict taken at week two is a verdict on your own follow-up rather than on the source. What is measurable early is conversations: how many people you reached and what they said.
Attempts per lead, time from delivery to first call, which number you called from, and the reason a lead was closed out. Without those four, a complaint about lead quality and a complaint about your own dialer look identical from the outside, and the vendor has no way to tell them apart either. With them, the conversation is short in both directions.
Buying questions rather than research ones are answered on the FAQ, and anything that is not there gets asked on a call.
More in Resources
The trigger lead ban: what H.R. 2808 stopped in 2026
What the Homebuyers Privacy Protection Act stops, who is still allowed to call, and what changed on 4 March 2026.
TipsThe mortgage lead follow-up schedule: 20 days, then a year
Which days to call, at what hours, when to change the number, and what happens to a lead on day 21 rather than day 1.
AnalysisWhy mortgage leads do not answer, and what is broken
A filtered text and a flagged calling number produce the same evidence as a dead batch. Both are invisible from your desk, and both are fixable in a week.
StatisticsWhat a mortgage lead looks like: credit, home value, income
Our own thirty day snapshot: average credit score, home value and household income on delivered leads, and the share our screen turns away.
TipsThe loan officer tech stack: CRM, dialer, local presence
CRMs, dialers and local presence numbers, and where each one earns its keep once leads are landing.