How to Test a Lead Source Before You Commit

A structured trial that produces a decision rather than an anecdote: how many leads, over how long, measured on what, and when to stop.

Figures in this article describe the wider market and are drawn from the third-party sources listed at the end. They are not Lead Foundry results, and nothing here is a projection of what any individual broker will achieve.

The short version

A broker tries a new supplier, buys fifteen leads, converts none, and concludes the supplier is poor. A different broker buys fifteen, converts two, and concludes the supplier is excellent. Both conclusions rest on a sample too small to distinguish either from chance.

This is how to design a trial that actually produces a decision, using measurements that show up in weeks rather than in the quarter the settlements land.

Why do most lead trials fail as experiments?

Three failures do most of the damage, and none of them are about the supplier. The trial is too small to separate signal from variance, the measurement waits for settled business, and the working process changes halfway through because the broker gets discouraged.

The last one is the most common and the most invisible. A broker who works the first five leads through a full cadence and the next ten with a single call each has not tested a supplier. They have tested their own enthusiasm.

The discipline that makes it a test: Work every lead in the trial through the identical cadence regardless of how the first few go. The moment the process varies with your mood, the result measures your mood.

How many leads do you need in a trial?

Enough that the metric you are measuring first has room to move. Contact rate is measurable on a few dozen records because most leads produce a contact outcome within a fortnight. Conversion to settled business is not, because the base rate is low enough that small samples are dominated by luck.

What you are measuringRoughly how many leadsHow long until you know
Contact rate30 to 50Two to three weeks
Appointment rate50 to 80Four to six weeks
Quality of fit and intent30 to 50Two to three weeks, from call notes
Conversion to settled businessWell over 100A full sales cycle, often a quarter or more

The practical implication is that a first trial is a data-integrity test, not a profitability test. You are establishing whether the records describe reachable people who wanted what the page offered. Profitability comes from a second, larger phase once the first has passed.

What should you measure, in order?

Measure in the order the funnel produces the information, because each stage tells you something different about where a problem lives.

That last one is a control, not a supplier metric. If your median time to first attempt during the trial was four hours, the trial measured a four-hour process. Given how sharply contact odds decay, comparing that against a supplier you once worked in five minutes is not a comparison at all.

What a trial can and cannot tell you at 40 leads (Strength of signal at a 40-lead sample)
Whether the phone numbers reach real peopleClear
Whether consumers expected the callClear
Whether the intent is genuineMostly clear
Whether appointments will hold upWeak signal
Whether it will be profitableNo signal

A small trial answers data-integrity questions well and profitability questions not at all. Buying a second batch to answer a question the first batch could never have answered is the most common way brokers spend money on nothing.

What are the stopping rules?

Write these before the first lead arrives, and hold to them. Rules written afterwards are rationalisations.

SignalWhat it meansAction
Repeated disconnected or wrong numbersVerification is not realStop immediately and claim replacements
Consumers do not recognise the enquiryThe landing page is misleading or the lead is oldStop and ask for the source URL
Consumers report other brokers callingThe lead is shared or was resoldStop and check the exclusivity term
Contact rate holds but appointments do notData is fine, your first call needs workContinue and fix the script
Everything holds through six attemptsThe source is worth a second, larger phaseScale to a measurable volume

Note the fourth row. It is the only outcome where the correct response is to keep buying while changing your own behaviour, and it is the one brokers most often misdiagnose as a supplier problem.

What should you agree with the supplier first?

A trial with no agreed replacement terms is not a trial, it is a purchase. Settle these before any money moves.

How Lead Foundry answers this: You set the criteria a lead has to meet before you start. Anything that fails them is replaced, and you can see the source brand on every record, so a query about where an enquiry came from is answered by looking rather than by discussion.

How many leads should I buy to test a supplier?

Between 30 and 50 for a first trial, which is enough to judge contact rate, data accuracy and whether consumers expected the call. It is not enough to judge profitability, because conversion to settled business has a low enough base rate that samples under 100 are dominated by chance. Treat the first trial as a data-integrity test and run a larger second phase for the commercial question.

What is the first thing to measure when testing leads?

Contact rate through a fixed six-attempt cadence. It is the earliest signal, it is the least affected by your sales skill, and it directly tests whether the verification the supplier claims is real. If people are answering and recognising the enquiry, the data is sound and any remaining problem is in your process.

How long should a lead trial run?

Two to three weeks for contact and accuracy signals, four to six weeks if you want a meaningful appointment rate. Anything shorter cuts off the follow-up cadence partway, which biases the result against the supplier, since more than half of eventual contacts happen after the first attempt.

Should I test two lead suppliers at once?

Yes, if you have the calling capacity to work both through the identical cadence, and only then. Running them sequentially introduces every seasonal and personal variable that changed between the two periods. Running them simultaneously with different levels of effort is worse than not testing at all.

Sources