Lies, Damn Lies, and Statistics: The Talent Data Dilemma
Look, here’s the thing. Everyone in talent acquisition thinks they have analytics figured out. And sure, most recruiting organizations have some snazzy looking dashboards, filters for days, and the unwavering belief that if you stare at the numbers long enough, clarity will happen.
We treat analytics like one of those Magic Eye books everyone was obsessed with, back before smartphones diverted our collective attention — empirical evidence is all we need to make a business case, even if the underlying data is incomplete, inaccurate, or just completely made up. The numbers will reveal the narrative, and that’s really what talent analytics are all about, right?
Yeah, not so much.
February has been one of those months where research and reality finally collided. Governments revised their employment numbers, adding additional clarity to just how strapped the global labor economy actually is. Economists, naturally, contradicted the headlines with hot takes and tons of speculative correlaries.
The judiciary reminded everyone that algorithms can’t be held legally accountable, and finally, a ton of TA leaders realized that their “insights” are really just activity logs with nicer fonts and a white label Looker instance.
If you’re looking for talent, or looking for a job, this should all start sounding pretty familiar. That’s why this month’s Recruiting Unfiltered takes a closer look at talent analytics that actually, you know, analyze talent.
This is a crazy concept, obviously, that’s sometimes uncomfortable, often inconvenient, but reflective enough of recruiting realities to drive informed decision making — and improved talent acquisition outcomes.
We know this sounds pretty boring, but don’t worry. You don’t need to know how to do math to make the importance of talent analytics add up.
Required Reading: Five Talent Analytics Stories That Matter
1. The Jobs Numbers You Already Quoted Got Rewritten
This month, the Bureau of Labor Statistics revised its 2025 employment data, which, as it turned out, proved to be a statistically significant adjustment, to say the least. Although, given that the BLS didn’t even publish official employment statistics for a couple of months last year, this should come as no surprise.
While the BLS numbers have recently become a charged, weirdly partisan issue, no matter what your ideology might be, the final numbers weren’t great (unless you’re an anarchist, obviously).
Payroll growth in 2025 was adjusted down by more than 400,000 jobs — a rounding error that’s about equal to the total population of Tulsa, Sacramento or Fresno, and just as scenic.
Job openings kept sliding; voluntary turnover and natural attrition barely moved. But this is only nonfarm data, so there’s a chance that the agricultural sector made up the difference — although workforce participation’s precipitous slide likely made the numbers look slightly better than the flaming dumpster fire that is the current job market.
Full story: Bureau of Labor Statistics 2025 Revised JOLTS Report
Why it matters:
Most workforce planning was predicated on overly optimistic forecasts and unrealistic projections that ultimately didn’t hold up. If your talent analytics felt a little off, that probably wasn’t a user error or a screwed up pivot table or Excel macro. Blame this one on bad assumptions — kind of like, “Yeah, Kid Rock should totally headline an alternative halftime show.”
That’s it. That’s the problem: Assuming.
2. Interview Scheduling Doesn’t Have to Suck.
Scheduling interviews has historically been one of the worst parts about being a recruiter, and one of the biggest choke points in the hiring process. Endless email trails, scheduling across time zones, and coordinating hiring team availability, in a word, sucks.
The good news is, it doesn’t have to. Joveo recently ran a roundup of the top interview scheduling tools for 2026, and, in their words, it’s less a G2 Crowd style vendor catalog and more of a practical look at how modern talent teams are leveraging this tech for use cases that go way beyond calendar management.
The post breaks down the relative strengths and challenges of each top platform, and which are worthy of consideration if you’re evaluating interview scheduling solutions that help real recruiters solve real workflow challenges — not just some white labeled Calendly clone that’s heavy on shiny features, but light on impact.
Full article: The Top Interview Scheduling Tools for 2026 (Joveo)
Why it matters:
Scheduling automation saves more than time and sanity. These tools actually prevent candidate attrition, increase hiring manager responsiveness and save 2-3 hours a day, on average, that TA pros waste on interview coordination and logistics.
Automating this part of the hiring process, unlike so many other AI use cases, actually delivers on the promise of enabling more efficient, scalable and optimized hiring processes, with less costs and fewer mistakes.
Of course, it’s not AI — but that’s not a bad thing, considering that this category has minimal risk associated with leveraging artificial intelligence. The rewards are totally worth it: less scheduling friction, more candidate engagement, and more time to focus on engaging and evaluating candidates.
This post proves that no matter what size, industry or market you’re hiring for, and no matter how big your TA team or budget might be, there’s a scheduling automation tool that’s right for you. In a profession where time is money and speed determines success, this might be the most underrated tool in any TA tech stack.
3. The Global Talent Shortage Narrative Still Doesn’t Add Up
Distributed workforces and multinational companies exist in over 190 countries (at least, according to the number of markets EORs claim to operate in, which makes you wonder which 3 UN recognized states are green field opportunities for payroll and compliance outsourcing), which makes talent analytics like the EPCOT of talent data.
To get the full picture, you’ve got to look at the entire world, too. Fortunately, the International Labour Organization (the extra “u” is for “Euro”) released its benchmark Employment and Social Trends 2026 report last month, and it’s about as optimistic about the future as climatologists, the Book of Revelation, or keynote presentations about how AI is changing everything.
The numbers are existential crisis level — the global jobs gap now sits at a whopping 400 million people who want to work, but either can’t access the formal economy or find opportunities to segue from “subsistence farming” or “gray market actors” to “active workforce participants.”
To put that into perspective, that’s around 3 times the number of Super Bowl viewers — which probably doesn’t help much outside the US, so it’s like 1 out of 12 people watching the World Cup.
Full report: Employment and Social Trends 2026 (International Labour Organization)
Why it matters:
When employees wonder where all the candidates are or complain that there’s no talent, they mean there’s no talent that meets their laundry list of qualifications — like living within driving distance to the job site, being open to compensation that’s within the tightly predetermined salary range, or having that most important of soft skills: a high threshold for professional pain.
400 million workers looking for a job — any job, presumably, except for staffing agency CSR — isn’t a talent shortage. It’s a talent imbalance and, increasingly, a labor crisis with profound global implications. The good news is, most of these workers live in developing countries, which means that demographically, those jobs will open eventually — first world workers just have to retire first.
Unless, you know, AI starts actually replacing jobs instead of providing a convenient excuse for RIFing them. Speaking of…
4. The AI Job Apocalypse Narrative Is Also Overstated
If you’ve been following the AI-and-jobs discourse for the past year, you’d be forgiven for thinking every job out there is in danger of imminent replacement by a handful of large language models and one very confident McKinsey associate.
The good news is, well, good news for workers and worrywarts everywhere. Oxford Economics released a short, but devastating analysis this month with the very sexy, peer review ready title, Evidence of an AI driven shakeup of job markets is patchy. This is economist-spreak, “everyone calm the hell down, already.”
This report is interesting in that it’s decidedly, and refreshingly, unsexy (see: title). It looks at real labor market data, choosing longitudinal data for its job market analysis instead of, you know, vibes or LinkedIn posts by people who think automation and algorithms are synonymous with AI (they’re not, unless you’re a product marketer these days).
The headline is kind of an understatement, the kind of stiff upper lip restraint you’d expect from Oxford Economics. It’s pretty restrained, actually, in revealing what can only be considered a plot twist: AI isn’t actually replacing workers at a structural or systemic scale.
Job displacement due to AI certainly exists, but it’s a fractional percentage of mostly administrative or support roles (sorry, recruitment coordinators). In the last year, Oxford provides a high end estimate of around 60,000 total jobs globally were formally linked to AI displacement.
That’s about as many workers as Amazon lays off every quarter — certainly nowhere near the several hundred thousand jobs lost in the US alone due to old fashioned economic considerations like lowered demand, overhiring or corporate austerity.
This raises an awkward question: if AI is so transformative, then why aren’t productivity numbers exploding?
Output per worker hasn’t spiked; rather, productivity growth globally demonstrably slowed in 2025, which is pretty much what you’d expect in a cyclical downturn — not what you’d expect to see if companies suddenly replaced human workers en masse with hyper-efficient machines or advanced robotics.
The conclusion is clear: if AI were driving large scale job displacement, the productivity signals would be obvious. Instead, the impact of AI on the future of work is still a ton of imperceptible noise on a branded “thought leadership” deck.
So, why does every layoff suddenly get attributed to AI? Well, because shareholders really don’t like to hear “we overhired, the market softened, and EBITA is down.” This is a way better investor narrative, even if it’s pretty much as fake as your average ICO.
Full report: Evidence of an AI-driven shakeup of job markets is patchy (Oxford Economics)
Why it matters:
For TA leaders and recruitment practitioners, this causality matters more than headlines suggest. If layoffs are cyclical (and evidence suggests they are), then they come back when the economy recovers and demand strengthens. If layoffs are structural, those gigs are gone forever.
Oxford – the institution which brought you penicillin, the internet and Steven Hawking’s voice emulator – reports that most of what we’re seeing right now mirrors what’s historically seen in every economic downturn, with AI functioning more as a narrative cover than a root cause of job elimination.
That doesn’t mean AI isn’t making an impact — it’s pretty obvious that it’s clearly changing how we work, where we work and what we work on.
Tasks are changing; job responsibilities are shifting; certain skills are becoming less valuable, while others are suddenly a golden ticket to six figure jobs. That’s a reorganization, not a revolution. And like every reorg, AI is simply changing hiring priorities and shifting skill requirements to create alignment and efficiency.
In other words, this isn’t Skynet. It’s like Excel, but with better branding and a cleaner UI/UX.
The danger for employers is mistaking technology for a strategy. If you assume AI is the workforce equivalent of Krishna, the destroyer of worlds, then you stop building pipelines in the short term rather than retrenching, and reloading, for the inevitable rebound. It’ll happen, eventually – and TA should be ready.
More often than not, though, recruiting teams and functions have to essentially get rebuilt from the ground up, which is why hiring teams miss the window of opportunity provided every time the economic pendulum swings back upwards.
Unless, of course, AI actually does start replacing jobs at scale – in which case, you’ll probably find out at your fully automated exit interview.
5. Bias Is Still There. We Just Made It Faster.
There’s a persistent fantasy in HR tech that bias is a human problem and algorithms are the fix. You know the pitch. Networks are neutral; math doesn’t discriminate. And, of course, data is totally objective.
2024 research from the University of Washington takes about three pages to ruin that fantasy forever. This study examined how resume screening models evaluate candidates with equivalent qualifications, but with names generally connoted with specific race or gender identities.
The result was exactly what anyone in this business would probably expect. LLMs favored traditionally white names (think: Karen, Stu, Brittany or Alaistair) at disproportionately high rates, even when everything else about the candidate’s experience or qualifications were the same.
A follow up study from the same research team at UW just dropped, and it should make everyone in talent acquisition feel a little awkward and mildly uncomfortable. The study suggested that when human reviewers are presented with AI generated recommendations, such as stack rankings, they follow them roughly 90% of the time — even when bias isn’t readily apparent, or when those humans voice concerns, like, “Gee, this seems pretty antithetical to our whole DEIB initiative.”
The presence of algorithmic bias — an unconscious bias that’s pervasive in hiring — seems to have already reached ubiquity in our race for AI supremacy. And the more we choose to blindly trust the judgement of LLMs instead of our own intuition and experience, the more quickly this bias will scale.
Unconscious biases used to be minimized with employee training; now, they’re being entrenched with employee training data. This isn’t to say that all models are malicious, nor are all recruiters mindless lemmings looking for a shortcut (just the overwhelming majority). It’s just when systems deliver, humans defer. This makes for quicker decisions and frictionless hiring processes.
It also makes for a major ethical dilemma — and potential compliance pothole — for TA leaders everywhere.
Full Research: AI Bias in Resume Screening (University of Washington)
Why it matters:
The bias conversation isn’t disappearing along with DEIB budgets; it’s just shifting — and beyond abstract ethical frameworks and transparency statements, the recruiting repercussions are becoming increasingly high stakes, from both an outcome and a compliance perspective.
Algorithmic bias, unlike most unconscious biases, is measurable, auditable system usage and behaviors. Evidence lives in user logs, candidate ranking schema, assessment knockout thresholds and systems of record — the same sort of documentation scope as OFCCP requirements.
If your talent analytics stack can’t audit screening outcomes by proxy variables, and trace decision paths from application initiation to rejection notifications or welcome letters, you’re increasing your company’s risk profile, not managing or mitigating it. While TA is busy placing bets on tools or technology, they’re also playing with house money. That’s as high stakes as gambling gets.
The courts, after all, are catching up — and case law, while limited, will soon follow. Lawsuits pending class action should provide some codification and clarification, but the uncomfortable truth is simple.
AI isn’t making hiring decisions more effective; it’s just automating professional bias and personal preferences at scale. Automation can’t fix broken selection logic or flawed decision matrixes — that part of “responsible AI” has nothing to do with AI at all.
Spotlight: Joveo 2026 Recruiting Benchmark Report
Joveo unveiled its 2026 Recruiting Benchmarks Report earlier this month, and in case you missed it, you should definitely fix that — and we’re not just saying that because, well, this is their newsletter. It’s similar to the annual report Appcast releases, only it’s got more utility and applicability for TA leaders for one big reason.
It doesn’t try to turn all recruitment advertising and talent attraction baselines and benchmarks into an industry and market agnostic monolith. The truth about the average recruiter, or the average employer, is that they don’t actually exist.
When it comes to benchmark data, the often drastic differences between locations, industries and job functions can’t be averaged away or aggregated. Fortunately, Joveo has the receipts across 16 occupation categories and every US state — the kind of data that’s prescriptive for recruiters, not one that’s pitchable for an external PR agency (ahem).
Here’s the full report, if you want to do a deep dive into the data. And you definitely should.
Read the Report:
Joveo 2026 Recruiting Benchmarks Report
Across 16 occupation categories and every US state, the data shows something most TA dashboards still refuse to admit. Two labor markets are running at the same time. Not sequentially. Not cyclically.
TL;DR: Top Takeaways from the Joveo 2026 Recruiting Benchmark Report
It shouldn’t be any surprise that the data shows that there are two distinct labor markets developing in the US, which bifurcates hiring into two general categories: white collar, tech adjacent jobs and the essential, frontline work that’s often referred to as “high volume” or “hourly” roles.
Applications for salaried knowledge work and experienced, exempt openings grew more than 9 times over the past three years alone. Frontline and non-exempt roles, conversely, saw minimal movement — application volume is about the same today as it was in the post-pandemic hiring frenzy.
In the same economy, at the same time, often at the same employers, hiring results and recruiting benchmarks were drastically different — a divide that’s only accelerating as the job market craters and professional aspirations give way to personal pragmatism.
This is why “average” isn’t really applicable to talent analytics. Volume and quality are conflated; quality is a subjective concept, and almost impossible to standardize and systematically measure with any degree of accuracy.
The Law of Averages Should Be Illegal: What Joveo Benchmarks Really Tell You
Applicant volume isn’t the same as demand.
Nine times more applications usually means nine times more automated rejections, not a 9x increase in qualified applicants overnight. Without screening infrastructure, volume becomes a liability, rather than a competitive advantage.
Two Different Labor Markets Require Two Very Different Strategies
The glut of experienced, professional knowledge workers and, say, the paucity of healthcare workers or skilled tradespeople look more or less identical when they’re living inside your ATS. But they’re very different markets, with very distinct challenges. Treating them the same way is how recruiting organizations fail the fastest.
There’s No Such Thing As “Overqualified” Anymore.
As applications surge, employers continue to add to their existing laundry list of preferred qualifications, requiring even more niche expertise or industry experience, even as salaries stagnate. This increases screening costs, drains pipelines and talent pools, and, often, raises candidate expectations for compensation or job title. Stick with the basics, or you’re likely going to get burned.
CPC spikes tell you where attention actually is.
The spike in cost per clicks for recruitment advertising that the report highlighted in December 2025 wasn’t seasonal, or even some random statistical outlier. It’s what happens when there are more candidates, less advertisers and a finite amount of job ad inventory.
Those companies that posted the most job openings in Q4 paid a premium per click — but those clicks also converted into actual hires and filled reqs. Employers who decided to wait until Q1 to reopen roles and reactivate their recruitment advertising face much tougher competition to attract the same talent. Even if the CPC costs are marginally lower, so too are the hiring outcomes.
Location Isn’t A Variable, It’s a Strategy.
From a talent acquisition perspective, the same role with the same title and identical responsibilities that’s open in Dallas versus, say, San Francisco couldn’t be any more different. National averages (like JOLTS reports or BLS data) are pretty much useless except as clickbait or a conversation starter.
Designing a recruiting strategy based on the averages of aggregate data is basically an exercise in futility. The way to interpret recruiting benchmarks isn’t complicated — it’s just a bit nuanced, and a little less convenient than a single data point.
The truth is, though, you can’t apply a single hiring model or TA strategy in such a fragmented market; dashboards are even less meaningful when they’re only showing national (or global) data or blend together both sides of the increasingly bifurcated labor market.
But Wait, There’s More:
Go beyond benchmarks and explore occupation and state-level data in real time, all the time with Joveo Interactive Insights. For real. https://www.joveo.com/interactive-insights/
Meet Us In March (Because Sitting Still Isn’t a Strategy)
It’s that time of the year again where the weather — and event season — are heating up. If you’re going to one of the many recruiting related events or conferences cramming this month’s calendar, there’s a good chance Joveo will be there, too.
We’re going to be doing a lot of closed door conversations and intimate meetups with talent leaders to learn more about what’s actually working, what’s quietly failing, and what’s new, and next for enterprise recruiting leaders.
Here’s the rundown of where we’ll be getting our geek on.
RogueHire Basecamp | March 4–6
We’ll be joining an expert panel digging into the state of programmatic hiring and the evolving dynamics around Indeed. The conversation should be practical, occasionally skeptical, and refreshingly light on buzzwords or BS.
In other words, it’s going to be useful and actually worth the time OOTO.
TA & M by Consero | March 15–17
Joveo takes the stage for AI-Driven Hiring in 2026: Redefining Talent Acquisition at this practitioner focused executive event to share a balanced look at how automation is actually influencing outcomes, where hype is lagging behind reality, and how TA leaders should be thinking about scale, signal, and accountability right now.
UNLEASH America | March 17–19
We’re going big for one of the biggest shows of the year. That’s right. Platinum sponsorship (and Vegas), baby. Because subtlety is overrated, we’ll be in our 10×20 booth (331, stop by, say hi) and hosting a boardroom session, AI in the Real World: 5 Must-Have AI Agents for Talent Acquisition.
The booth and panel, though, are really just pretext for the two exclusive events we’re hosting, including a private dinner and what will almost certainly be the most over-subscribed RSVP list of the week.
You’ve been warned.
SIA Exec Forum North America | March 23–26, Austin
Look for us in Booth 507. While everything isn’t bigger in Texas (our booth is slightly smaller than at Unleash), we’ll be at the biggest event of the year for staffing leaders and executives. The conversations here are way less theoretical, way more practical, and inspired more by cash than case studies or connections.
If you’re going to be deep in the heart of Texas, we hope to see you there, y’all.
March is packed, and so are our bags. If you want to talk about what the numbers are really saying, what hiring strategies are actually working and which strategies employers are using to drive better results and competitive advantage, well, you know where to find us.
We’ll even buy you a top shelf drink or two. Promise.
What Actually Matters Right Now
If you are running talent acquisition in 2026, here’s our list for what matters the most when it comes to more meaningful metrics and actionable analytics:
Funnel conversion rates by role type.
Qualified applicant yield per dollar spent.
Time in stage by hiring manager.
Offer acceptance variance by compensation band.
Source quality over source volume.
The teams that treat talent like a demand generation engine outperform those who treat it like a requisition fulfillment function. Talent analytics are the bridge between recruiting related spend and successful hiring outcomes. Without the right data, you are guessing; with it, you’re actually accountable.
Which is pretty much the entire point.
What’s next for talent analytics? Well, you probably guessed that the future is going to be less pretty dashboards, and more operational discipline. There will be fewer vanity or volume metrics, and many more uncomfortable conversations and shifts in strategy and spend.
If you ended up in TA, it’s probably not because you’re some sort of math prodigy, but the good news is that hiring data isn’t really that complicated. “Best practices” tend to favor complexity, but instead, baseline metrics and talent analytics provide the rarest thing of all in this business — clarity.
And there’s no bigger competitive advantage when it comes to hiring success than keeping it simple.
We’ll have to keep that in mind for next month’s newsletter.
Read the 2026 Recruiting Benchmarks Report Press Release (and full report)
