Three months into the search, the spreadsheet has 140 rows. Each one is a role you were qualified for, an application you tailored, and in most cases a silence that followed. You have had three first interviews. Last week you learned that the job you most wanted went to someone a former colleague of the hiring manager had mentioned over coffee. That person never appeared in the applicant queue at all.

It is tempting to read this as bad luck, or as a reason to send 140 more. The data on how hires are made suggests a different reading. The route a candidate arrives by is the largest measured difference in the hiring funnel. A job search run mainly through applications is concentrated in the route with the lowest odds per attempt. The alternative is to treat the search the way you would treat any decision made by several people you do not control: as a stakeholder problem.

The cold application has the lowest odds, and they are falling

Gem, a recruiting-software company, publishes benchmarks drawn from its customers’ hiring data: more than 165 million applicants and 1.2 million hires between mid-2021 and mid-2025. In that data about 0.5% of applicants are hired, and recruiters are handling 93% more applications than in 2021 on teams that are 14% smaller [1]. Ashby, another recruiting platform, finds the offer rate for inbound applicants fell from 7 in 1,000 to 2 in 1,000 between late 2021 and late 2024, as inbound volume roughly tripled [2].

Official figures point the same way. In March 2026, people unemployed for 27 weeks or longer made up 25.4% of the unemployed in the United States, up from 21.3% a year earlier [3]. And in the New York Fed’s consumer survey for December 2025, the average perceived chance of finding a job after losing one fell to 43.1%, the lowest reading in the survey’s history [4].

The honest version of this needs one qualification. Applications still fill roughly half of all jobs. In Gem’s data, job boards and company marketing generate about 90% of applications and about half of hires [1], and on Ashby’s platform inbound applicants reached 52% of hires in mid-2025, a four-year high [5]. The application works. It is simply the route where each attempt is least likely to succeed: volume is highest there, and the employer knows least about each person in the queue.

Same market, very different odds by door

Now compare the other routes in the same datasets. Gem reports that referred candidates convert to hires at 11 times the rate of inbound applicants, and internal candidates at 32 times. Candidates a recruiter approached directly are nearly eight times as likely to be hired, producing 11% of hires from 2.6% of applications [1]. On Ashby’s platform, 40% of referred candidates move from application to interview, against about 3% of inbound applicants. Internal applicants are under 1% of all applications, and about 42% of them get an interview [2].

Two cautions apply. These figures come from companies that sell recruiting software, and their customers lean toward technology and fast-growing firms. And the multipliers compare different people, not the same person entering by different doors: referred candidates may simply be stronger on average. What the data establish is where hiring attention goes. They do not promise that a referral will multiply your own odds by eleven.

Two further findings show how much hiring now runs on people an employer already knows. Nearly half of the hires recruiters source, 46%, are “rediscovered” candidates already sitting in the employer’s own database, such as past applicants, up from 26% in 2021 [1]. And in an older New York Fed survey, employed people who were not looking for work received about a quarter of all job offers, and more than half of the unsolicited ones, while the unemployed sent 40% of the applications and received about 16% of the offers [6]. Offers flow toward people who are already visible to someone.

What a referral actually transfers

Why would a name passed along by an employee count for so much? The best answer comes from three field experiments on an online labor market, which found that a referral carries positive information about a worker’s performance and persistence that is not contained in anything an employer can observe on the profile [7]. A referral is a transfer of credibility. Someone the employer already trusts is saying something about you that your résumé cannot.

The advantage is concentrated at the gate. In personnel data from one large US corporation, referred candidates were more likely to be hired and stayed longer, but their initial wage advantage faded over time [8]. Across nine large firms, referred workers were substantially less likely to quit, while their productivity was similar to everyone else’s on most measures [9]. The referral gets you looked at properly. After that, your work is judged like anyone else’s.

That has a practical consequence for how you ask. If a referral works by transferring information, the person vouching for you needs information worth transferring. “Can you put in a word for me?” gives them nothing to say. Three specific outcomes, the role you want and why you fit it, and a short paragraph they can forward give them a case they can make in their own voice.

Which contacts: weaker than you think, but not the weakest

The largest experiment on this question ran on LinkedIn: randomized changes to connection recommendations across more than 20 million people over five years, during which two billion new ties and 600,000 new jobs were created. It found an inverted-U relationship. Weaker ties led to more job moves than strong ones, but only up to a point [10]. The most effective contacts were moderately weak ones, people with whom a member shared about ten mutual connections [11]. The pattern held most strongly in more digital industries; in less digital ones, strong ties did better [12].

A study of about six million Facebook users adds a useful correction. Most people who were helped into a job were helped by a weak tie, because weak ties are so numerous. But any single strong tie was significantly more valuable than any single weak one [13]. The two findings are compatible. Your close contacts are individually your best advocates, and there are not many of them. The acquaintances a step or two away are individually less likely to help and collectively reach far more openings.

The LinkedIn experiment measured newly formed connections, not requests for referrals, so it does not prove that asking an acquaintance works. It does suggest where to look: former colleagues from two jobs ago, people who left your company for one you are targeting, the manager of a team you once worked alongside. These are people who know enough of your work to say something true and move in circles you do not.

Run the search as a stakeholder map

In practice this means spending the first week of a search on a map before a single application goes out.

  • List the employers, then the people. For each target organization, who do you know inside, who knows someone inside, and who has been approached by its recruiters? Mark each as a strong, moderate or weak tie.
  • Check the internal door first. Internal candidates convert at the highest rate of any source, partly because some internal roles are opened with a particular person in mind. Even so, before looking outside, find out which roles at your current employer you could credibly move into, and who decides.
  • Write down what each person can say. What have they seen of your work first-hand? What could they credibly claim? Where the honest answer is “not much,” they are a source of information, not a referrer.
  • Give each referrer a case. Three outcomes with numbers, the role, and a forwardable paragraph. Make it easy to repeat accurately.
  • Stay findable. With nearly half of sourced hires coming from employers’ existing databases, an earlier application or a final-round finish is an asset. Keep your profile current and answer recruiters even when the timing is wrong.
  • Send fewer, better applications. A meta-analysis of 378 samples found that the sheer intensity of a search was only modestly related to interviews and offers and did not predict the quality of the job obtained, while the quality of the search and the discipline with which it was managed predicted both [14].

Once you are in a process, the same thinking applies inside it: a panel interview is a small stakeholder map of its own.

Who the referral system leaves out

A system that runs on who knows whom does not treat everyone alike. In the single-firm study above, most referrals took place between people similar in age, gender, ethnicity, education and staff level [15]. In a Payscale survey of about 53,000 US workers, white women were 12% less likely than white men to have received a job referral, men of color 26% less likely, and women of color 35% less likely, after controlling for industry, job level and other factors [16]. And the higher the role, the more it depends on the network: Ashby notes that only about 35% of executive hires come through inbound applications [5].

If your network is thin where you want to go, that is a constraint to plan around. Two things follow from the evidence. First, the contacts that matter most are moderately weak ones, which can be built deliberately and fairly quickly: through professional communities, former colleagues’ colleagues, and people doing the job you want who are willing to spend twenty minutes describing it. Second, the application still fills about half of all roles. A small number of well-chosen applications, each backed by evidence, belongs in every search.

Persona Map keeps the two things this approach depends on in one private workspace: a map of the people who can vouch for you, and a record of the outcomes they would be vouching for. But the method needs nothing more than a page and some honesty about who knows your work. The hiring decision will be made by a few people with too many candidates and too little time. Most of them will start with the names someone they trust has already mentioned. Your search is the work of becoming one of those names.

Sources

  1. Gem. “Key Takeaways from the 2026 Recruiting Benchmarks Report.” December 2025. Link
  2. Ashby. “Are Referred Candidates More Likely to Get Hired?” Talent Trends Report, 2025. Link
  3. US Bureau of Labor Statistics. The Economics Daily: duration of unemployment, March 2026. Link
  4. Federal Reserve Bank of New York. Survey of Consumer Expectations, December 2025 results. January 8, 2026. Link
  5. Ashby. “How Many Hires Really Come from Inbound?” Talent Trends Report, 2025. Link
  6. Federal Reserve Bank of New York, Liberty Street Economics. “How Do People Find Jobs?” April 2017. Link
  7. Amanda Pallais & Emily Glassberg Sands. “Why the Referential Treatment? Evidence from Field Experiments on Referrals.” NBER Working Paper 21357, 2015. Link
  8. Meta Brown, Elizabeth Setren & Giorgio Topa. “Do Informal Referrals Lead to Better Matches? Evidence from a Firm’s Employee Referral System.” Federal Reserve Bank of New York Staff Report 568. Link
  9. Stephen Burks, Bo Cowgill, Mitchell Hoffman & Michael Housman. “The Value of Hiring through Employee Referrals.” Quarterly Journal of Economics 130(2), 2015. Link
  10. Karthik Rajkumar, Guillaume Saint-Jacques, Iavor Bojinov, Erik Brynjolfsson & Sinan Aral. “A Causal Test of the Strength of Weak Ties.” Science 377(6612), 2022. Link
  11. MIT Sloan School of Management. Summary of the LinkedIn weak-ties study, 2022. Link
  12. MIT News. “The Power of Weak Ties in Gaining New Employment.” September 15, 2022. Link
  13. Laura Gee, Jason Jones & Moira Burke. “Social Networks and Labor Markets: How Strong Ties Relate to Job Finding on Facebook’s Social Network.” Journal of Labor Economics 35(2), 2017. Link
  14. Edwin van Hooft, John Kammeyer-Mueller, Connie Wanberg, Ruth Kanfer & Gokce Basbug. “Job Search and Employment Success: A Quantitative Review and Future Research Agenda.” Journal of Applied Psychology 106(5), 2021. Link
  15. Meta Brown, Elizabeth Setren & Giorgio Topa. Federal Reserve Bank of New York Staff Report 568, full paper. Link
  16. Payscale. Analysis of job referrals by gender and race, from a 2017 survey of about 53,000 US workers. Link