Adding a new lead source is one of the riskiest routine decisions a buyer makes. A good source can lower acquisition costs for years. A bad one can waste budget, frustrate a sales team and, worse, expose the business to consent complaints that surface months after the leads were bought. The answer is not to avoid new sources, but to test them the same way every time: small, documented, measured with your own data, and scaled only on evidence. This guide sets out a test framework that works for networks and direct buyers alike.
Step 1: review compliance before anything else
Compliance review comes first because it is the only part of a test that cannot be fixed afterwards. Leads bought on defective consent remain defective however well they convert. Before agreeing a single lead, ask for:
- Live landing page URLs and mobile screenshots for every funnel that will feed your programme.
- The exact consent disclosure text, and confirmation that your company or brand is named in it.
- A description of the evidence stored per record: timestamp, IP, user agent, page URL, disclosure version and, where you require it, an independent consent certificate.
- Any vertical-specific disclosures, for example the CMS-required disclaimer in Medicare, or state-specific language in legal and home improvement.
- How opt-outs are captured and passed on.
If a publisher cannot produce these quickly, stop there. The answer tells you how the rest of the relationship will go.
Step 2: understand the traffic
Lead quality follows traffic quality. Ask which channels the traffic comes from (search, social, native, display, email, affiliates) and in what rough mix. Ask directly whether any traffic is incentivised, co-registration, sweepstakes or from sub-affiliates. None of these is automatically disqualifying, but each changes what you should expect, and you should know before you pay. A publisher running its own media to its own funnels can answer precisely; a reseller often cannot.
Step 3: write the filters down
Most early disputes in a test are really disagreements about filters that were never written down. Put them in the insertion order:
- States, counties or ZIP codes you can serve, and any you exclude.
- Hours of delivery and the time zone they refer to, including weekends and holidays.
- Qualifying criteria specific to the vertical, such as currently insured, minimum debt amount, homeowner only or case type.
- Duplicate rules: what counts as a duplicate, against which data set, and over what window.
- Exclusivity: exclusive to you, or shared with a stated maximum number of buyers.
Step 4: agree returns before the first lead
Returns are where goodwill is won or lost. Agree a short list of valid return reasons (typically: invalid or disconnected phone, wrong person, outside filters, duplicate, consumer did not request contact), the window in which returns can be made, and what evidence is needed. Be realistic: "did not buy" is not a return reason in most models, and a publisher that accepts any return will eventually price that into the lead.
Step 5: set caps and a sensible window
A test should be big enough to tell you something and small enough that failure is cheap. Set a daily cap that your team can work properly, and a total cap or end date. The window should be long enough to cover your sales cycle: for a same-day product such as an auto insurance quote that may be a couple of weeks, while for mortgage, education or legal intake it may need to run longer before you can judge downstream conversion. Agree in advance what happens when the test ends: a pause and review, not automatic scaling.
Step 6: measure with your own numbers
Decide before launch which metric will decide the test. It should be one your own systems produce, not one the publisher reports. Typical choices by model:
- Forms and posted leads: contact rate, quote or appointment rate, then sale or bind rate.
- Live transfers: transfer connection rate, average handle time and close rate.
- Inbound calls: billable call rate, qualification rate and close rate.
Track by source, sub-source and day if the publisher can pass that information. A source that looks average overall may contain one funnel that performs very well and one that does not.
Step 7: share feedback during the test
The most useful thing a buyer can do during a test is share outcome data back with the publisher. Even a simple weekly file of lead IDs with contacted, quoted and sold flags lets a publisher change targeting, questions and filters. Tests without feedback mostly measure the publisher's default settings; tests with feedback measure what the source can become.
Step 8: audit a sample
Pick a sample of records and check them end to end: does the consent evidence match the live page, was your name in the disclosure, does the timestamp make sense against the delivery time, and do the consumers you reach remember asking for contact? A short audit early in the test catches problems while they are cheap to fix.
Step 9: scale in steps
If the test meets the agreed threshold, raise caps gradually rather than all at once. Quality can change as a publisher buys more media to meet higher volume, and stepped increases make any change visible early. Keep the same filters, returns and reporting in place as volume grows.
A short checklist
- Live pages and consent text reviewed and approved.
- Traffic sources disclosed in writing.
- Filters, exclusivity and duplicate rules in the insertion order.
- Return reasons, window and evidence agreed.
- Daily and total caps set; test window covers the sales cycle.
- Decision metric chosen and measured in your own systems.
- Weekly feedback loop in place.
- Sample audit completed.
- Scaling plan agreed in steps.
How Growth Box runs tests
Growth Box is a new publisher, so every relationship starts with exactly this kind of capped test. We share funnels and consent text before launch, write filters and returns into the agreement, and ask for outcome data so we can tune traffic. We do not ask anyone to scale on our word. To start a test, tell us about your programme.
This article is general information about testing lead sources, not legal advice.
