Were the weakest LinkedIn ties best for finding a new job?

The message that changes a job search may come from someone you have not spoken to in months.

That is the promise of a weak tie. Close contacts often know the same people and hear the same news you do. A looser connection can reach into another company, industry or city and bring back information your usual circle never saw.

LinkedIn turned that familiar idea into an unusually large experiment. The answer was not simply more strangers, more jobs. It was more interesting: weaker ties helped, but what counted as weak changed where the advantage appeared.

An algorithm opened different doors

Between 2015 and 2019, LinkedIn tested versions of its People You May Know system. Some versions were more likely to recommend people with many mutual connections; others surfaced people with fewer. Members still chose whether to send or accept an invitation. The experiment changed the doors they were shown.

Researchers later analyzed those random differences. The 2015 wave involved more than four million members and 19 million new connections. The worldwide 2019 wave involved more than 16 million members, about two billion new connections, more than 70 million job applications and roughly 600,000 events classified as job transmissions.

Scale is the study’s great strength. Random assignment also helped separate the effect of newly created ties from the ambition, occupation and sociability of the people creating them—differences that blur ordinary network data.

Weaker worked. Weakest depended on the measure

There was no single winning category.

When the researchers defined strength by mutual connections, the pattern formed an inverted U. Moderately weak ties produced the largest marginal increase in the job-transmission estimate; the very weakest and strongest ties produced less.

When they defined strength by later message intensity, the lowest-intensity ties produced the largest estimate. The middle of the distribution was comparatively flat.

Those measures capture different relationships. Someone who connects you to a distant professional circle is not automatically the person you message least. Weak tie sounds like one thing. The experiment found at least two.

A job transmission was a clue, not a verified referral

LinkedIn could not see every conversation or hiring decision. It used a specific platform proxy: one member reported working at a company before another later reported joining it; the two had connected at least a year before the move, and the reported start dates were at least a year apart.

That sequence is consistent with opportunity travelling through the tie. It cannot show who shared a vacancy, supplied a referral or secured the hire. Nor does it reveal salary, job quality or whether the move worked out.

The industry pattern was also uneven. In the shorter 2019 application window, weak ties produced more applications to more digitally intensive industries, while stronger ties did relatively better in some less digital industries. Different work may ask different things of a relationship: novel information in one setting, deeper trust in another.

A bridge is not a contact score

These were LinkedIn recommendations, not randomly assigned friendships. The results apply to marginal new ties created through particular product tests, among people who saw and acted on them. The statistical interpretation also depends on the recommendation variants affecting jobs through the ties they created rather than another route.

Two authors were current or former LinkedIn employees and reported a significant financial interest in Microsoft. Aggregated data and code for the correlational analyses are public; individual member data are restricted.

The human lesson survives those boundaries. Close relationships and loose connections do different work. One carries depth. The other can carry distance.

For opportunity, the useful question is not how many people sit in your network. It is whether any relationship reaches somewhere your existing circle does not.