Name-only search vs. group search

Across 69 paired searches for a person associated with a property but without an address of their own, a traditional skip trace returned nothing at all 65% of the time. Searching the same two people together, as a group, returned nothing 4% of the time. In every pair where the traditional search found the right person, group search found them too; the reverse never happened once.

The question

An investor has an address for Person A and a name for Person B, someone connected to A but without a known address of their own. Both methods get the same three facts: A's name, A's address, B's name. A traditional query can only use two of the three. Does giving a method access to the third fact (A's name) change the outcome, and by how much?

How we tested it

We decomposed 28 historical production jobs under one tenant into every (A, B) pair where A was documented at a searched property and B was any other named person on that same job. That produced 70 pairs across 22 properties, which is the entire population this rule can produce from the available history, not a sample drawn from a larger pool. One pair was an exact duplicate entry of the same two names on the same job, so it is counted once in scoring: 69 unique observations.

Both arms ran on the same 69 pairs, which makes this a paired design:

• Traditional skip trace: B's name plus A's property address. A's name is not used.
• Group search: A search of A and B together, with A's property as the input address.

"Delivered" means B's record was returned as entry #1 for the traditional arm, or affirmatively matched for group search. Ground truth is B's identified match from the original production job, a standard this platform's own method produced, which is a disclosed asymmetry (see below). 19 of the 69 pairs had no such ground truth, because B was never found in the original run; those 19 are included in the "returned nothing" count, which needs no ground truth, and excluded from the "delivered" count, which does.

Results

Metric	Traditional skip trace	Group search
Returned nothing at all	45/69 (65%), CI 45–86%	3/69 (4%), CI 0–12%
Returned something	24/69 (35%)	66/69 (96%)
Best match delivered (n=50)	23 (46%), CI 20–67%	37 (74%), CI 57–88%

Paired difference on delivered (group search minus traditional): +28 points (95% CI +13 to +46). The interval excludes zero.

Strict dominance, counted across all 50 scoreable pairs: group search delivered where traditional returned nothing in 14 pairs; both delivered in 23; neither delivered in 13; traditional delivered where group search did not in 0 pairs.

Segmented to the heir case specifically, where B has no documented tie to A's property (n=44): traditional returned nothing 77% of the time, group search 0% of the time.

What this does and does not show

Group search's "delivered" figure is scored against this platform's own production identification of B, not an external or field-verified source. That is a real asymmetry: group search is being measured, in part, against a standard its own method produced. This is why the "returned nothing" numbers (65% vs. 4%) carry the most weight in this study. They require no ground truth at all and are externally checkable on their own terms.

19 of the 69 pairs have no ground truth, because B was never located in the original production run. Those 19 are excluded from the "best match delivered" metric and are not evidence of group search's success; they remain in the "returned nothing" metric, where no ground truth is needed.

At 69 observations across only 22 properties, several of the individual rate intervals are wide. Traditional's "returned nothing" rate, for instance, has a confidence interval spanning 45–86%. The paired difference and the strict-dominance count are the two figures robust enough at this sample size to lean on; treat any single rate with more caution.

This is one tenant's property book, predominantly Texas. The claim is about the query an investor can actually construct from three known facts, scored by one standard, on one dataset.

Why it matters if you skip trace

This is the pattern investors describe as "dialing blind": the query comes back empty two-thirds of the time, not wrong, just silent, and you're left guessing whether to try another spelling or give up on the lead. The traditional method isn't lying to you when it fails here; it simply has nowhere to put the one fact (the other person's name) that would have let it answer.

Group search uses that third fact by design, which is the entire difference measured here. If your current workflow runs into that same wall of empty results on heir or associate leads, run a group search on the next one and compare.

Data and reproducibility

• 69 unique observations (70 pairs, one exact duplicate) across 22 properties, 28 source jobs.
• Measured July 19, 2026, under a pre-registered blueprint committed before any result was observed.
• 80 total data-provider calls against a 500-call ceiling (70 for the traditional arm; 10 for group search, the rest served from cache). Zero errors, zero timeouts.
• This study has since been superseded in scale, not in kind, by a larger follow-up on 171 paired cases; see Finding the heirs a skip trace can't.
• The full blueprint and results file are available on request.

Cite this: GroupSkip Research (2026). Name-only search vs. group search: a 69-pair A/B test. groupskip.com/research/name-only-vs-group-search.

---

Related: Finding the heirs a skip trace can't · Why the first record a vendor returns is often wrong · How these studies were run