I hold all four in quantity. A bachelor's in information technology. A CISSP. Multiple MCSEs (earned across successive generations of Windows Server back when each track ran six or seven exams). The CompTIA stack. An ITIL foundation. My first was an MCP on Windows NT 4, which I passed as an early teenager. Underneath all of it, a career spent running the systems these certificates describe.
The degree taught me nothing I did not already know. The program was competency-based, built on the assumption that you demonstrate what you already have, and the assumption held.
So here is a question I have never seen answered well. If the degree taught me nothing, why did I bother, and why did it change anything for anyone evaluating me?
The answer is that all four are the same kind of thing: instruments for verifying something nobody can observe directly.
The problem all four exist to solve
An employer cannot observe capability before hiring. This is the entire situation, and every hiring practice that exists is a response to it.
Capability is only directly visible through work over time. You find out whether someone can do the job by watching them do the job, which requires already having hired them, which is the decision you are trying to make. There is no way out of this loop. Interviews sample it, work trials approximate it, references report on it secondhand, but nobody gets to observe the thing itself before committing.
So the market produces instruments. A degree, a certificate, a record of experience, and a claimed skill are four different attempts to make an unobservable property legible to someone who has to decide without being able to check.
That is what they have in common. All four are verification instruments for the same hidden variable, issued by different parties, acquired at different costs, and wrong in different directions.
Seeing them as instruments explains most of the confusing behavior around them.
Four failure modes
Each one fails in a way the others do not, which is why organizations use all four despite each being individually weak.
A degree certifies persistence more reliably than knowledge. The thing a four-year degree actually demonstrates with high confidence is that you started a multi-year structured commitment and finished it. The knowledge content is real but highly variable, decays fast in technical fields, and in my case was zero because the program was competency-based and I already had the competencies. The instrument still worked. It was issued by a third party with something to lose, it is expensive to fake, and it cleared filters that were closed to me before. Its failure mode is that the trait it measures only correlates with capability, and the correlation is loose enough that everyone knows people it got wrong in both directions.
A certificate certifies a bounded body of knowledge at a moment in time. It is the most precisely scoped of the four. You know exactly what a CISSP holder was asked about, because the domains are published. That precision is also its weakness. My own CISSP took about a week of preparation, because the domains describe work I had been doing for most of my career and the exam was a formality over knowledge I already held. Someone with none of that background can sit the same exam after a boot camp. The credential reads identically in both cases, because the exam measures what you can answer on the day and has no way to see what produced the answer. Its failure mode is that it certifies exam performance and lets the reader infer capability, which is a much larger claim than the exam supports.
The certificates that held their value have since been modified to address exactly this, and the modification deserves its own treatment.
Experience is measured in a unit that does not mean anything. Years is the worst unit in professional evaluation, and the entire market runs on it. Ten years can be one year of learning repeated ten times, or it can be ten genuinely different years in which the work changed under you and you changed with it. The résumé renders both as a date range. Its failure mode is that the unit conflates duration with accumulation, and duration is the part that is easy to verify.
A skill is the closest to the real thing and the least verifiable. A skill claim is an assertion about capability itself. That directness makes it the most informative of the four and the least trusted. It is self-issued. There is no third party with anything at stake. Its failure mode is that the person making the claim is the person who benefits from it, and the confidence with which someone claims a skill is uncorrelated with whether they have it.
All four are timestamps
The instruments share a property that matters more than any of their individual weaknesses. Every one of them attests to acquisition. None of them attests to possession.
A degree records that you learned something in a particular year. A certificate records what you could answer on a particular day. A record of experience records what you were doing between two dates. A skill claim carries no date at all, which makes it the worst of the four on this dimension, because the reader supplies a present tense that the claimant never actually asserted.
The asset underneath the attestation keeps moving after the attestation is issued, and it moves in three ways that have nothing to do with each other.
It erodes. You were good at something, you stopped doing it, and you are worse at it now. The rate is unknowable from the outside and frequently from the inside, because competence decays without announcing itself and confidence does not decay alongside it.
It drifts. Your knowledge stays intact and the thing it describes gets replaced underneath you. Twelve years of genuine excellence at on-premises Exchange is a real asset, and a meaningful fraction of it does not convert to Microsoft 365 administration, because the operational surface moved. What transfers is the structural part, the understanding of mail flow and directory dependencies and what failure looks like, and the structural part is exactly what no instrument records.
And the market revalues it independently of both. A skill can be sharp, current, and unwanted. Demand moves on its own schedule and does not consult your portfolio.
None of these three are visible in any of the four instruments, because all four were issued at a moment and the subject kept moving afterward. This is the gap that everything else in the post sits on top of, and it explains why the certifying bodies eventually did something about it.
What an expiration date certifies
Somewhere in the last fifteen years the serious certifications stopped being permanent.
A CISSP now requires 120 continuing education credits across a three-year cycle plus an annual maintenance fee. CompTIA moved its mid-level exams from good-for-life to a three-year renewal window with continuing education attached. The cloud and networking vendors run similar cycles. Across the market, the certificates that kept their standing are the ones that added a maintenance requirement, and the correlation is tight enough to be causal.
This is the certifying bodies responding to the degradation of their own instrument, and the response is a reasonable one. A credential that certifies a moment in 2011 says progressively less every year about what its holder can do in 2026. Attaching a renewal cycle converts a point measurement into a series of them, and a series is harder to fake than a single reading. The credential now asserts that you passed an exam and that you have stayed close enough to the field to keep the credential current, which is a meaningfully larger claim.
It also introduces a new failure mode, because continuing education credits measure attendance. Vendor webinars generate them. Conference sessions generate them. A person can accumulate a full cycle of credits without encountering a single idea that changed how they work, and the renewed credential is indistinguishable from one earned by somebody who spent three years going deeper. The maintenance requirement improved the signal against cramming and did nothing at all against coasting.
The expiration rules describe the issuer's policy decisions and very little else. My MCSE on Server 2008 has no expiration and never will. My MCITP from the same generation, covering the same product family, from the same vendor, expired on schedule. The technology underneath both went obsolete at identical speed. Only one of the credentials was structured to notice.
The other function of a renewal cycle is supply management. A certification that never expires accumulates holders monotonically, and a credential held by everyone who ever passed the exam carries less information every year purely through dilution. Expiration prunes the population. People lapse, people decline to pay, people leave the field, and the pool of active holders stays closer to the pool of practicing professionals. That is genuinely good for the signal and it is also a recurring revenue line, and both of those things are true at once without either one being the scandal.
What this means in practice is that a current certification and a lapsed one are different instruments, and the difference is not the knowledge. A lapsed CISSP held by someone who has run security programs continuously since they earned it tells you more than a current one held by someone who has been collecting webinar credits. The renewal date is evidence about administrative diligence, and it gets read as evidence about competence.
Why the stack works
Line up the failure modes and they do not overlap, which is the whole reason a hiring process reaches for all four.
The degree is hard to fake and weakly informative. The certificate is precisely informative and easy to cram. Experience is expensive to acquire and reported in a meaningless unit. The skill claim is directly relevant and entirely unverified.
Each instrument covers a gap the others leave. A certificate plus a long record is a different proposition than either alone, because cramming a body of knowledge is cheap and cramming a decade is not. A degree plus a specific skill claim is different from a degree alone, because the degree supplies a third-party issuer and the skill supplies the actual relevance.
The system accreted, which is why it is inelegant and why everyone involved complains about it. It persists because the instruments triangulate, and triangulation is the only available response to a variable you cannot measure directly.
The argument that credentials are worthless
I have some version of this conversation several times a year, usually with someone on the network or systems side, a decade or two into a career, genuinely good at the work. The position is that certificates are revenue for the certifying bodies, degrees are overhead, and time spent actually running production is the only thing that counts. It is usually delivered with an edge, because the person has watched credentialed people fail at things they can do in their sleep.
Most of the supporting evidence is correct. Certifying bodies do run a business. Degrees are expensive and slow. Certificate holders who cannot do the work are real, common, and infuriating to work alongside. Nothing in this post contradicts any of that.
The conclusion is the part that goes wrong, and the error is specific: they concentrated their entire portfolio in one instrument.
That was a legitimate choice and it paid real dividends. While other people were studying, they were on call, and on call is where the knowledge comes from. Nobody should apologize for that trade, and a lot of the best operators I have worked with made it.
But the instrument they concentrated in is the one with the weakest verification properties of the four. Experience carries no third-party issuer, which means nobody outside the employer can attest to it. Its unit is years, which conflates duration with accumulation. Its content erodes with the technology stack it was acquired on. And it is the instrument that most requires a conversation to convey, which means it only works after someone has decided to have a conversation with you.
That last one is where the position turns from a defensible preference into a career problem.
The filter does not take arguments
The case for experience over credentials is a good case, and the person making it almost never gets to make it to anyone.
Applications at any organization of size go through an applicant tracking system, and increasingly through a model that scores and ranks before a person opens anything. Neither one is evaluating whether you are good. They are matching structured tokens against a requisition: degree present or absent, certification names, years in a title, keyword density against the posting. A requisition that lists a bachelor's as required produces a hard filter, and a hard filter does not weigh your twenty years against the missing field. It removes you from the set and moves on.
The result is a category of applicant who is entirely qualified, frequently overqualified, and structurally invisible. They send fifty applications, hear nothing, and conclude that the market has gone insane or that ageism explains it. Some of that is real. A meaningful fraction of it is that they never reached a human, and the thing standing between them and a human costs a few months and a few thousand dollars to acquire.
This is the least satisfying part of the whole system and the easiest to fix. A degree that teaches you nothing still clears the field that says degree. A certificate you could have written the exam for still matches the token the screener is looking for. Paying for access is an unglamorous reason to acquire something, and it remains the correct reason when access is the thing being sold.
The automated layer is also getting thicker at exactly the moment the instruments are getting noisier, which is a genuinely bad combination. Models read structured signals well and unusual careers poorly. A human evaluator can be impressed by a shape they have never seen before. A scoring model has nothing to be impressed with, and the fields it can read are the four instruments.
The part that is breaking
Every one of these instruments degrades in proportion to how cheaply it can be acquired relative to the capability it claims to represent. This is Goodhart's law operating on the hiring market: an instrument that becomes a target stops carrying information.
The degradation used to happen slowly and unevenly. A certification would become popular, boot camps would spring up around it, the pass rate would climb, and employers would gradually stop treating it as meaningful. Degree inflation worked the same way over a longer cycle. Each instrument degraded on its own schedule, and the others held while it happened.
What is different now is that the cost of producing the appearance of all four collapsed at the same time.
Résumés that present a generic career as a coherent narrative of increasing responsibility can be generated in minutes. Exam preparation has become dramatically more efficient, which is a genuine good for learning and a genuine problem for certification as a signal. Skill sections are trivially expanded to match any posting. Even the interview, which was the fallback when the instruments failed, has been partially compromised by the fact that common interview questions now have polished answers available to anyone who looks for them an hour beforehand.
All four instruments are being optimized against simultaneously, by more people, at a lower cost than at any previous point. The triangulation still works in principle, but the individual measurements got noisier at the same time, which means the intersection is wider and less useful than it was.
What still carries information
The instruments were always proxies for the same underlying question, and when proxies degrade, evaluation moves back toward the thing itself.
The thing itself is depth, and depth has a signature that has not become cheaper to fake: the ability to keep answering as the questions get more specific.
Anyone can say they led a data platform migration. The follow-up questions are where it lives. What did the cutover plan look like. What did you do about the reconciliation gap. Which vendor claim turned out to be false and when did you find out. What did it cost, all in, counting the implementation hours and the person who keeps it running. What would you do differently. Name a specific decision you regret.
A person with real depth goes three or four levels down and then tells you exactly where their knowledge stops. A person without it produces more vocabulary, at a higher level of abstraction, with increasing fluency. The two are easy to distinguish if you know the domain and nearly impossible if you do not, which is why the quality of hiring now depends heavily on whether the people conducting it have depth themselves.
Good interviewers have always done this. What has changed is how much weight it carries, now that every instrument around it has gotten noisier at once.
What follows
For anyone assembling a set of these for themselves: acquire each instrument for what it actually does. A degree buys you past filters, and that is a legitimate reason to get one even if it teaches you nothing. A certificate is worth the time when it is scoped to something you need to know anyway, and close to worthless as a substitute for knowing it. Experience is the only one you cannot compress, which is precisely why it is the one to protect, by ensuring your years are actually different from each other. And the skill claim is the one that will be tested, so make claims you can defend four levels down and do not make any others.
For anyone evaluating: the four instruments are a filter. They tell you who to talk to, and they are worse at even that this year than they were last year. The hiring decision has to come out of the conversation.
The person with twenty years and no certificates chose a path, and the path bought them something real that the credentialed route does not reliably produce. What the path does not include is a mechanism for anyone outside the room to verify what they know. That is the whole gap, and closing it costs a fraction of what they already spent acquiring the thing the credential would attest to. Refusing to close it on principle is a defensible position that other people pay for on your behalf, in the form of never finding out you exist.
The common thread is that every one of these attests that you once acquired something, and the question anyone actually needs answered is whether you have it now. That question can only be answered by watching someone work. All four are imperfect approximations standing in for it, and all four are getting worse at roughly the same rate. What replaces them is the oldest method available: asking harder questions and listening carefully to how far the answers go.
— Chris