7 patterns from 10 real CVs — and why capable people are so often mis-read
Marketing claims seem unreliable. Technical claims read as unremarkable. Nobody wins.
Quick context: I’m currently building ReLoop.me — a tool that reads professional capability in non-linear careers. It has 2 sides: ‘Profile Intel’ reads CVs beyond keyword matching, perfect timelines, AI inflation, personal claims, blurs or inconsistencies, while ‘Role Intel’ reads job descriptions for what they truly ask for, what the role actually needs, and what they’ll likely offer.
The tool is still in private beta, and I’m still giving away free CV and JD reviews in exchange for feedback while I continue updating, tweaking, and calibrating it before the paid public beta, which should be ready soon.
And I’ve got some very interesting insights to share already.
The TL;DR version:
9 of 10 CVs showed up as likely to be misread or put in the wrong pile.
ReLoop.me’s premise held up across most profiles. We’ll see if it holds at scale.Only 27% of CV claims are actually demonstrated. And 49% are just plausible.
But that’s also because most CV formats keep it tight. Right?7 of 10 profiles claim leadership. 2 can evidence it. But 1 never claimed it at all.
CV inflation, overclaims, skills in the blindspot - the tool sees them all. Me proud.
But first, THE BASELINE:
Disclaimer: I selected a very small sample of anon data from private beta testers.
So they’d be as diverse as possible, but an easy number (10) for stats & my own attention span.
Just saying - these were not selected randomly, but I wasn’t very picky either.
10 real people. 3 women · 7 men.
(1 friend · 1 ex-colleague · 3 online connections · 5 complete strangers.)Geography:
Romania ×6 (Bucharest ×3 · Cluj · Timișoara · Sibiu)
Rest of the world ×4 (Spain · Switzerland · UK · Canada)Sectors:
B2B SaaS, Fintech, Cybersecurity, Automotive, Crypto, Marketing, Non-profitSeniority:
Individual Contributor ×3 · Senior IC ×1 · Manager ×1 ·
Senior Manager ×3 · Director ×2CV completeness: Substantial ×2 · Moderate ×5 · Limited ×3
Transferability: High ×3 · Medium ×7 · Low ×0
Profile types: Specialist ×3 · Operator ×3 · Builder ×3 · Connector ×1
Nothing about this group is similar. That is the whole point.
The one thing they have in common:
they wanted to see if they’re being (mis)read as professionals, and they tested ReLoop.me.
BTW - Every figure above and below is from ReLoop.me's own framework and features.
I didn’t run the CVs through an ATS, so these are not measured ATS rejection rates. They are just patterns that complicate or influence automated or human CV screening.
Here’s what the bigger picture looks like:
1. Everybody borrows the company’s numbers
Triple X ARR. Thousands of users. Billions of.. money. $Exponential → Growth.
Most outcome claims in a CV borrow corporate public numbers.
And while (of course) companies get their big numbers thanks to the people actually doing the work, the top corporate outcomes are usually a team result, even when pushed by just one person.
On the other end, those who don’t include any results or metrics at all just list their previous responsibilities instead. Even AI knows that it’s best to:
lead with what you personally caused; let the company’s number sit behind it as context.
But who keeps track of that, really? Well…. maybe we should.
2. The system doesn't know which pile you go in
ReLoop.me is designed to read capability where other tools - or people - only see gaps, blurs and inconsistencies. Which means the tool itself is built to catch what ATS systems might flag. So I didn’t run the CVs through an ATS, but rather built ReLoop.me to show the specific shapes and biases that make a profile harder to get.
Turns out, only 1 in 10 CVs lands where it should.
But even that 1 profile got labelled with Medium inflation, because 5/7 of its claims were only partially demonstrated. Aaand the cleanest profile in the whole batch — with 0 unsupported claims was still considered ATS risky due to other triggers.
So…. being screenable and being honest are a bit unrelated.
The system doesn’t filter for truth — it filters for shape.
Here are just a couple of triggers that get people mis-read or put in the wrong pile:
single employer → lack of breadth One high-severity misread (coordination figures reading as seniority) on top of a single-employer arc, => breadth penalty
multi-employer → the profile keeps landing in the wrong pile A Director title reads as unproven, plus a multi-context career pattern => the profile gets mis-routed.
concurrent roles → date conflict, or fabrication A fractional label reads as a gap, a multi-company history reads as instability, => ATS is unreliable.
advisory → gap
sector-grouped CV → unrelated jobs
Stay in one company and you’re narrow. Diversify sectors and you’re unreliable.
The perfectly linear, small progression career timeline almost nobody has anymore.
But one pattern was clear: overlapping roles trigger biases in 100% of cases.
3 profiles with overlaps. All 3 flagged. Failure rate: 3/3.
We all probably suspected this already, but now it’s obvious. Eh..moving on..
3. Almost everyone is a leader these days
Except for the real ones. Funny how that really pops up, unexpected.
In the sample, 7 of 10 profiles make or imply a leadership claim.
But only 2 profiles really evidence it — and one of those didn’t even claim it.
“team of 13” / “team of 18” → reads like internal team coordination, not management of actual team
“hiring, coaching, scaling small teams” → no headcount, no outcome. Partial
“built and led cross-functional teams” → don’t we all? Claim unsupported
“board chair, 60–70 volunteers” → aka leading dozens of volunteers for real, yet never labelled as leadership on the CV. Demonstrated, but never claimed.
“7 → 35, three EMs, six teams” → a team grown, explained, demonstrated.
Reading between the lines is so different from reading for keyword matches.
And I didn’t even write this rule down in the framework for AI to catch specifically.
4. The real blind spot for undervalued capability
So - ReLoop.me is calibrated to find capability — but it turns out undervalued capability is just as obvious for it to find. It found some in every single profile.
Turns out, the blind spot for undervalued capability sits on either
‘standards and compliance’, or ‘influence without authority’ side.
Both are invisible in a CV because neither has a title or a headcount attached.
These are among the hardest capabilities to hire for and the easiest to screen out, because they live in the in-between and the professional operating system of people.
What that actually looks like in practice:
An IC who had quietly created the quality standards an entire team adopted.
A director whose rarest skill was fixing cross-organizational problems without any formal authority to do so — spotting the systemic issue, building the proof, convening the people, shipping the fix.
A simple coordinator whose real capability was influence-based leadership, although they had not a single direct report on paper.
A standard CV would look odd if it read “this person does things without being told to.”
But ReLoop.me sees that across a CV’s timeline, if the real capability is there.
5. Certifications > Formal Education
Across the 10 profiles, ReLoop.me flagged formal degrees as either “relevant”, neutral, or noise for the overall profile intel and the bigger picture.
Guess how degrees in Geology, Law, Architecture, or Arts went.
But turns out:
Certifications matter a lot more.
The recent, role-specific certifications contributed more directly to the professional read than the original degree in several profiles. Not a finding per se, and maybe not that relevant in such a small sample, but that’s something to know, maybe.
6. The pay gap isn’t about seniority.
It’s about TRANSFERABILITY.
So - since Profile Intel is designed to also estimate a market value for each profile/CV,
I assumed the gap between local pay and European pay would narrow as people got more senior. Nope. Turns out the pay gap doesn’t hide in Seniority at all.
Sure, seniors earn better than juniors, but when comparing each Romania-based profile’s local market value against the EU/remote benchmark for the same seniority, the gap was something like this:
IT support → +60%
QA engineering → +40%
Go-to-market → +30%
Product specialist → approx +10%
Automotive safety and compliance → approx +5%
So the biggest gap is on generic skills easy to transfer remotely. Support, QA, marketing. But the narrowest gap is on work that is very regulated, very niche, and locally scarce. And that specific type of work is already paid well in Romania, close to European rates actually.
So the “Go work remote’’ or ‘go work Abroad’ trend needs some context:
If your professional value is deep, in a niche, you’ll already get most of the premium pay at home. The remote move is far less of a pay upgrade.
Not news, I know. But worth saying that 3/4 people in this batch working outside of RO are actually Romanian. And for them, no European anchor showed up at all in their profile intels. Because they’re already paid at EU level . But they got to find out their current market value in Ro. Aka one of the reasons they left, maybe.
7. Most CVs are plausible but unproven
In this sample of 10,
49% of what people wrote on their CV is plausible but unproven.
Partial claims showed up in 10 out of 10 profiles.
Claims: 27% Demonstrated · 49% Partial · 23% Unsupported (n=81)
Claim = a statement about ownership, responsibilities, outcomes, leadership, or expertise identified in the CV. ReLoop.me classifies each claim as demonstrated, partially demonstrated, or unsupported based only on the evidence present in the submitted document. That’s the best it can do before an actual test/interview, really.
So, if half of what we write in a CV can neither be confirmed nor denied,
the point isn’t that “people lie.” It’s just very hard to tell what’s true in a CV.
Maybe the CV format itself is the actual problem.
There’s simply no place in a CV where you can explain the context, the scale, the ownership, or the before-and-after. The capability may be real. It just isn’t legible.
And now for the interesting finds:
Non-linear and multi-context profiles were the easiest to mis-file
— aka the founding premise behind ReLoop.me, confirmed on live reads and real profiles other than mine. And yes, I know that’s just proof, not actual validation.“Distinctive” skills are truly rare — only 2 found across all 10 profiles.
Both on Directors with quantified outcomes if you’re curious. ReLoop.me knows to surface something, only if indeed truly rare and distinctive in a profile.The 3 marketers in the sample produced 74% of the unsupported claims. Ha!
And out of 10 profiles, the 3 marketers had 14 of the 19 Unsupported claims (!).
And the only High Inflation profile in the sample was also a marketer.
All while a technical profile got flagged as under-read by ReLoop.me (that can happen too!)
But that’s only partly a marketing problem, and I need to defend it:
Folks, marketing outcomes are collective by nature (leads, traffic, ARR).
And a… QA framework or a shipped feature is more of an individual outcome, no?
Yet funny enough, out in the real hiring market,
marketing claims get mis-read as unreliable, tech claims get mis-read as unremarkable.
For the most inflated CV in the sample, ReLoop.me refused to give a pay range, because it couldn’t support the estimate with evidence. So it published only the minimum instead, and asked for human verification. Haha.
Turns out ReLoop.me really does have a spine. I’m so very proud.
And the old Romanian saying ‘cum ti-l cresti asa il ai (aka you get what you raise)’ applies to AI models as well, it seems.
And now time for some out-loud conclusions:
Big Thank you to everybody featured here, anonymously. You really helped.
To be perfectly honest, I worked primarily from the personalised reports ReLoop.me generated rather than reviewing the original CVs.
Privacy issues aside, I also wanted to see if the profile intel reports were enough for me to understand somebody’s professional journey, without looking at their CVs. (plus I already know CVs are old stories, and knew ReLoop.me would evidence the good stuff). And it did - about 4 of these 10 profiles really helped me improve the product by triggering something I hadn’t seen, hadn’t included, or hadn’t fixed properly before.
So thank you, truly.
There really IS something wrong with the hiring market.
I was mostly expecting friends and connections from Bucharest to join the early beta trials. Instead I was very surprised to see people I didn’t know (from Sibiu, Iași, Timișoara, Cluj) reach out to me or to the ReLoop.me Linkedin page (which has less than 5 followers), and people from other countries as well. Like someone in Canada who really resonates with the “mis-read non-linear” messaging, even though ReLoop.me is built and designed especially for the EU.
Speaking of product design: I built ReLoop.me for me senior professionals, since being mis-read usually requires having something more to (mis)read in the first place. But, to my surprise, at least 2 or 3 juniors (with under 5 years of total experience) asked for a CV review anyway. So… yeah. Hope it helps.
I know, 10 reports is a very small sample to draw conclusions from.
I selected these 10 for the diversity of it (and because it was easier for me.) The acctual private beta batch is a bit bigger and still growing, and I might share some more comprehensive insights on profile intels in the coming weeks.
I also plan to do a 10 insights after the Role Intel findings, so stay tuned.
Meanwhile, don’t rely on these patterns being true.
They probably are, but I can’t tell for sure yet.
What I know is that the first ten live reads were consistent with the problem ReLoop.me was built around. So maybe the tool itself is a little self-biased, although I built many guardrails for AI to be able to contradict itself, internally, repeatedly.
Because what’s harder, and what I’m actually working on, is the calibration behind the system that reads human capability: how to read it all against the same scale, so that when you compare or correlate, the conclusions really mean something. Stay tuned.
And do reach out if you want to reloop.
___
Previous articles in this series:
