A common way to test a new vendor is to order a big batch because the price drops with volume, dial it for a week, and decide by how the invoice looks. By the time the results arrive the budget is spent and the answer is muddy. If you buy leads for a call center, the test is where the money is won or lost.
We have supplied call centers since 2005, so we have seen cheap tests and expensive ones. Below is the 14-day protocol we would run as the buyer, from vendor questions to a scale-or-stop decision. Run it on us too: our fresh verified files average a 42% contact rate against an 18% industry norm for recycled data, and a test shows what they do on your floor.
Get the answers in writing before you spend anything. Source: does each record come from a person asking for help, or from a compiled list? Age: does the date mean when the consumer inquired or when the record was last checked? Consent: what did the person agree to, and can I see a sample record with its documentation? Every phone record we deliver is DNC-screened with consent documentation. Verification: what checks run, and what share pass? Ours is SNS verification, and 85% of records pass.
Exclusivity: one buyer or several? Bad records: what counts as one, how long do I have to report it, and what is the remedy? Ours is a 100% file accuracy guarantee with a replacement policy for verified bad records. Test terms: what is the smallest order, and will you run my suppression file before delivery? A vendor that dodges these before you pay will dodge them after.
Days 1 and 2 are paperwork, not dialing. Write the spec: category, states, format (live transfer, real-time exclusive, targeted data list or aged 30-90 day), the fields you need on every record, exclusivity and delivery method. Our services page shows what we offer. Then write the pass line: the contact rate, qualification rate, cost per conversation and cost per sale that would make you buy more. Base the levels on your current vendor's results and on what a sale is worth to you, and write them down before any leads arrive, so a good-looking batch cannot move the goalposts.
Common mistake: no suppression list. Build one now from existing customers, your internal do-not-call list and records you already own. Without it you pay for people you already have and may dial people you should not.
Order a batch small enough that a bad result is a lesson, not a loss, and large enough that your agents can work every record inside the test window. Load your suppression file before the batch reaches the dialer, and tag each record with the vendor and batch so results trace back. Then control what you can: the same script, agents, dialer settings and calling hours you use for your baseline. If you test two vendors at once, send equal batches to equal teams, or the result tells you about your teams, not your vendors.
Common mistake: ordering too large. A big first order is tempting when volume pricing is on offer, but it hides the problem: if the file is weak, you find out after paying for all of it. Volume discounts are for the second order.
Give every record the same treatment: the same attempts, spacing and times of day, with dispositions logged in the same categories. Record the minutes from delivery to first dial too, because slow dialing makes any vendor look bad. Track four numbers per batch. Contact rate: the share of records where a live person answers. Qualification rate: the share of those contacts who meet your criteria. Cost per conversation: total spend divided by live conversations. Cost per sale: total spend divided by sales closed. Count spend as the lead cost plus the agent time spent on it. Log every bad record with its reason (disconnected, wrong person, duplicate, never asked); you need that list on days 11 and 12.
Common mistake: judging on price per lead. A cheaper lead that contacts and qualifies less can cost more per sale, because every dial, conversation and sale is paid for with the same agent hours. Compare vendors on cost per sale, and let price per lead break ties.
Days 11 and 12: test the promise. Report your bad records the way the vendor's policy says, with the proof it asks for, and watch what happens: how fast they answer, whether you get a replacement or a credit, and whether the replacement is clean. Check delivery too: did the file arrive when promised, in the promised format, with the promised fields, and without duplicates against earlier files or your suppression list?
Days 13 and 14: decide. Put your four numbers beside the pass line from days 1 and 2. If you passed, scale in stages, with the same spec and the same tracking at each stage. If you came close, run one more small batch with one named change. If you failed, stop, write down why, and send the numbers to the vendor; a good one will want to see them. Selling Medicare? Plan the test to end before the annual enrollment period (October 15 to December 7) starts, and see our Medicare leads page for what we offer.
Big enough that your agents can work every record inside the test window and a few lucky or unlucky calls cannot decide the result, small enough that a bad batch is cheap. The right number depends on your floor and category, so tell your vendor your capacity and ask what size they recommend.
It depends on the age of the file, how well it was verified and how fast you dial. Our fresh verified files average a 42% contact rate against an 18% industry norm for recycled data. Treat those as reference points and let your own numbers decide.
Treat it as an answer. A vendor confident in its data should accept a small controlled batch with clear terms for bad records. If the terms are vague or the minimum is large, ask why, and be ready to walk away.
If you want to run this protocol on us, send your category, states and daily dial capacity through our contact page and we will build a test batch with you.