How can I test a skip-tracing provider's accuracy before buying a large list?

A labeled sample is a file of owner records where the correct phone number or email is already on hand. Closed deals and past conversations produce rows with a verified answer behind them, which is what separates a test from another purchase. One hundred fifty records separates providers that differ by more than a few points and stays small enough to score by hand.

Entity-owned parcels belong in the sample at close to their share of the target list. An LLC or trust row exercises a different part of a provider's matching than an individual owner does. A sample drawn only from individually owned single-family homes ranks providers correctly for that one shape of record and misleads on every other.

Vendor figures cannot be compared with one another, because each company publishes a different metric measured on a file nobody outside it has seen [1][8][9]. A match rate counts records that came back with something attached, while a right-party contact rate counts attempts that reached the named owner [6]. Two providers quoting those numbers side by side are answering different questions, so the ranking stays undecided until you measure both on one file.

The TCPA sets damages at $500 for each violation, which a court may raise to $1,500 where the violation was willful [14]. Investors comparing providers in a BiggerPockets thread report usable contact information on a fraction of a run, one poster putting it at 20% to 30% of 1,000 properties [15]. Those posts are reports rather than measurements, the same limitation the vendor pages carry.

The owner name as recorded goes on every row alongside the site address and the tax-roll mailing address. The county and APN follow, plus the date each field was pulled. PropertyRadar's import documentation names county and APN as the starting point for a high-confidence match, which is why the identifier set carries the result [2]. The same page requires a number and street plus a city or ZIP before an address match qualifies, and accepts phone or email matches only when they are exact [2].

That documentation also offers a low threshold that admits loose matches for review and a high threshold that excludes relevant records to protect accuracy [2]. Every matcher makes that trade without showing you the dial, so a sample missing identifiers tests your own file more than the provider.

Owners you have already reached skew toward the ones who were reachable, which flatters every provider in the test. A fifth of the sample held by LLCs or trusts is a reasonable floor when the target list runs entity owners at that rate. Entity rows change what a right-party conversation means, because a returned contact there is a signer with no verification of beneficial ownership behind it.

Match rate and precision move independently

Match rate counts records that came back with something attached, while precision counts how much of that belongs to the owner on the deed. Google's machine learning course defines precision as the proportion of positive classifications that were correct, computed as true positives divided by true positives plus false positives [5]. A returned contact file rarely receives that arithmetic, because it requires the answer key a labeled sample supplies.

Right-party contact is the one some vendors publish

REsimpli defines a contactable match rate as confirmed working phones or emails divided by total leads loaded, counting a live connect or an SMS reply [6]. It puts a workable contactable rate at 60% to 80% against a stated industry baseline of 30% to 40%, figures the company publishes with no study attached [6].

Reachability and recency decay after the export

A number that matched cleanly in March may be disconnected by June, which is why reachability belongs in its own column. Contacts under 12 months old divided by contacts returned gives the recency figure, and no vendor in this group publishes it.

One vendor refuses to quote a number at all

Skip Genie declines to publish an accuracy percentage [7]. The company tells buyers to ask any provider quoting one for the method behind the study, since such a figure varies with whatever variables were measured [7]. The variables it names are the ones a labeled sample fixes in place, which is the argument for scoring 150 rows yourself.

One frozen file goes to every candidate inside the same week. Providers refresh on different cycles against constant churn, with roughly 35 million U.S. numbers disconnected and released for reassignment each year [11]. Submitting a month apart hands the later provider a data advantage unrelated to its matching.

A provider returning four numbers on one owner has handed you four separate compliance decisions. The do-not-call status of the primary line says nothing about the other three, so screening has to cover every number in the return. Registry access rides on your own Subscription Account Number, which the FTC issues to the seller separately from any telemarketer's account [13].

Attempt budgets keep one stubborn record from consuming the test, and REsimpli puts the expected touch-to-connect ratio at 4 to 6 attempts [6]. Phone type decides the order inside that budget, because 78.7% of U.S. adults lived in wireless-only households during the second half of 2024 [12].

PropertyRadar publishes one number and labels where it came from, stating that customers have reported an accuracy rate of 80% or better [1]. The same page adds that no skip-tracing service reaches a 100% accurate match rate, and places a cold-calling match rate above 70% inside its very-high-quality band [1]. No right-party contact rate appears there, which means the figure cannot be set against one measured on conversations.

REISkip states that its users see 85% to 90% of their leads matched with phone numbers, a figure that counts phones and says nothing about who answers [8]. Flat pricing at $0.15 per skip appears on the same page, which makes REISkip cheap to include as a third candidate [8].

BatchData's product listing on the Datarade marketplace headlines a 74% right-party contact rate for its residential real estate data [9]. The marketplace labels that quality section as self-reported by the provider, and the same listing carries a 100% fill rate [9]. Fill rate counts how many cells arrive populated rather than how many owners answer, so those two figures measure different events.

None of those numbers share a denominator, which is why none of them can be ranked against another. A measured result from 150 rows in one county beats all of them, because it runs on one file and one set of definitions.

Cost per reachable owner divides total spend by the number of owners an attempt reached during the test. Include the platform fee where you are buying the platform for the trace, and leave it out where the subscription was already paying for list building you would keep anyway.

PropStream charges 12 cents per contact on Essentials and 10 cents on Pro and Elite, with the first traces bundled through PropStream Connect [10]. A 150-record sample therefore costs $18.00 on Essentials, sitting on top of a $99 monthly plan [10]. Skip Genie charges 17 cents per record with bulk lists as low as 13 cents, which brings the same sample in between $19.50 and $25.50 [7].

PropertyRadar meters contacts by plan rather than charging per record [3]. Solo at $119 a month includes 250 phone and 250 email contacts, Team at $249 includes 500 of each, and Business at $599 includes 2,500 of each [3]. A 150-record sample fits inside the Solo allowance, so the marginal trace cost is zero there and the plan fee carries the whole denominator [3].

Additional contacts past those allowances run 8 cents on Solo, 6 cents on Team, and 4 cents on Business [4]. One purchase covers every phone number and email address the platform holds for that person, so a record returning four numbers costs the same as a record returning one [4].

The arithmetic below is illustrative, using invented return counts to show how the division works. Suppose the 150-row sample comes back from PropStream Essentials with contacts on 104 records and 38 conversations that reached the owner of record. Trace spend of $18.00 divided by 38 gives 47 cents per reachable owner, and adding the $99 plan fee for a month bought only for this test raises it to $3.08 [10]. The same sample through REISkip at $22.50 with 34 owners reached lands at 66 cents with no platform fee attached [8], so the ranking flips depending on whether you keep a subscription anyway.

A bundled allowance and a per-record price are different purchases

A bundled allowance charges nothing extra until it runs out, so the platform fee carries the entire cost per reachable owner [3]. A flat per-record price exposes the trace cost cleanly, which means the ranking between the two reverses as soon as your monthly volume changes [8].

The winner holds only for the shape of record in your sample

A test built on absentee single-family owners in one county says little about entity-held parcels three states away. Split the tally into individually owned rows and entity-owned rows before signing anything larger, because a provider that wins on individuals and loses on entities is still the right buy for a list running 90% individual owners.

The result expires faster than the contract

Score the same labeled rows again before any large renewal, since the file a provider matched against in one quarter is a different file six months later [11]. Phone numbers turn over faster than mailing addresses, which is how a provider that wins one test loses the next without changing its matching.

A published match rate is not a measure of whether anyone answered. It counts records that came back with a phone number or an email attached, whether or not that contact belongs to the owner named on the deed [5]. Precision is the separate figure that answers the ownership question, and no vendor can compute it without the answer key a labeled sample supplies [5].

A match rate is also not a statement about your county or your owner profile, since vendors measure on files they assembled and never publish the composition of those files [7].

An accuracy claim is not a right-party contact rate, even when both arrive as percentages on one page [9]. PropertyRadar attributes its 80% or better figure to customer reports rather than an internal study, which tells you the denominator is customer-supplied [1].

A right-party contact on an entity-owned parcel is not a confirmed conversation with the beneficial owner. The person a trace returns there is a signer or an authorized representative, so score those rows apart from the individually owned ones.