Last updated: August 28, 2026
By Gregg Flecke, Senior Talent Acquisition Partner, KORE1
EDI analyst interview questions in 2026 should test acknowledgment reasoning, ASN accuracy against the physical shipment, partner-spec deviation handling, and one real production failure, not definitions of X12 transaction set numbers. Most published question lists stop at the definitions. Nobody has ever failed a definition. The better questions all share one property. They put something broken in front of the candidate and watch.
I sat in on a final round in Memphis last spring. Third-party logistics company, four people on the panel, and a candidate they had already half decided to hire. He was good. Every answer landed where it was supposed to. He could define a functional acknowledgment, he knew what an 856 carried, and he had the right platform names on his resume with a believable number of years behind each one.
Ninety minutes of it. Nobody handed him anything.
Right at the end, the operations lead, who had barely spoken, slid a printed page across the table. One interchange. A header, a couple of functional groups, a few hundred segments, wrapped badly by whatever printer it came off. “What do you see?”
He turned it around twice. Then he said he usually worked inside the mapping tool. Nobody followed up.
That was the interview. Everything before it had been vocabulary, and vocabulary is the one part of this skill set a person can acquire in a weekend.
You should know what I get paid for before you weigh any of this. KORE1 places EDI and integration talent. A hire that washes out in four months is a bad outcome on our side of the table as well as yours, which is why I would rather hand you the loop than sell you a search. Take the loop below and run it yourself. I mean that. If you would rather not, that is what our EDI analyst staffing desk does, inside the broader IT staffing services practice, and these reqs sit close enough to the ERP that a fair number route through ERP consultant staffing too. This page assumes the req exists and the band is set. If either is still open, start with how to hire an EDI analyst instead and come back.
The Loop Changes With the Seat
One assumption before the questions, because it changes all of them. The seat you are filling has a shape, and the shape decides which questions carry signal and which are decoration. Retail supply chain, healthcare claims, and ERP-side mapping share a standard and very little else. Our guide to hiring an EDI analyst works through that decision properly, and the template for writing the req is the faster way to pin it down if the req is still vague.
Assume it is settled. Here is the one question I would open with for each shape, and who shows up to complain when that part of the estate breaks.
| Seat | Who calls when it breaks | The question that sorts real from recited |
|---|---|---|
| Retail and supply chain B2B | Accounts receivable, once the deductions post | “The 856 passed validation and we still got charged back. Where do you look first?” |
| Healthcare X12 | Revenue cycle, when cash slows down | “Walk me through reconciling an 835 against what the billing system thought it sent.” |
| ERP integration and mapping | Whoever owns the ERP, usually on a Monday | “A partner sends a field we have no home for. What happens next, and who decides?” |
| Trading partner onboarding | Sales, when a new account cannot transact | “How long does onboarding take, and what actually makes it take longer?” |
| Operations and monitoring | Nobody, which is the problem | “How would you find out that a partner stopped sending, before they call?” |
Then watch what happens when you ask about a seat they have never worked. The honest reaction is a flat admission followed by a plan. What you get from the other kind is an answer shaped roughly like the question, delivered with total confidence, and the confidence is exactly the problem. Flag that candidate.
The Acknowledgment Ladder Filters Faster Than Anything Else
If every other question in this post got thrown out, I would keep one. “A partner says they never received the invoice. Our system shows it sent. Walk me through your next twenty minutes.”
It looks like a support question. It is actually four competencies wearing a trench coat. Ask it early.
The strong answer walks down a ladder. Did the interchange leave us at all, or did it stack up in an outbound queue behind something else. Was there a TA1 back, which validates the ISA and IEA envelope and nothing beyond it. Did a 997 come back, and if so, what did the AK9 say, because “accepted with errors” is not the same as accepted and a partial acceptance is the most-ignored status in this line of work. Then, and only then, did an 824 come back reporting that the content failed once the partner’s application actually tried to consume it. A document can be structurally flawless and still be business-garbage. The X12 standards themselves draw that line clearly, and candidates who have lived it draw it without being prompted.
Weak answers resend the file. Immediately. Then resend again. No envelope check, no acknowledgment check, no question about whether the partner’s mailbox is polled or pushed. They are fast and they sound certain and they are the reason somebody eventually creates duplicate purchase orders in the ERP.
One follow-up carries a surprising amount of weight. “What do you do with a 997 that comes back accepted with errors?” I have watched senior-titled candidates blink at that. More than once. It is a normal Tuesday for anyone who has really owned a production environment.
Push on the ASN Until Something Specific Comes Out
Every other transaction set costs you time when it breaks. The advance ship notice costs you money, on a schedule set by whoever is receiving the truck. Walmart, Target, and Amazon each publish their own supplier standards with their own scorecards and their own math, and no two of the three calculate it the same way. Walmart’s supplier requirements are the public front door for that. The dollars are somebody else’s section to argue. Not yours.
Yours is one question, asked narrowly. Not “tell me about ASNs.” This. “The 856 passes validation and the retailer charges us back anyway. Where do you look?”
Listen for the hierarchical structure first. Shipment, order, pack, item, and whether that HL tree actually mirrors how the goods were physically built. Then the serial shipping container code on the label, because an ASN promising an 18-digit SSCC that the receiving scan cannot find is a compliance failure no matter how clean the syntax was. Then timing, since an ASN that lands after the truck is worth nothing. Then quantities at the line level. Boring, and load-bearing. Then the question I most want to hear, which is whether anything downstream in the warehouse changes the shipment after the ASN is generated.
That last one is where the expensive failures actually start, and it is more common than people expect. The map was never wrong. The process moved out from under it. Nobody told the map.
There is a second question here that is organizational rather than technical, and it separates analysts from people who have only supported analysts. “Who owns the ASN when a third-party logistics provider packs the cartons?” There is no clean answer. What you want is a candidate who treats it as a boundary to negotiate and document rather than a thing that just sort of happens.
Onboarding Questions, and the Companion Guide Nobody Reads
Ask how long it takes them to onboard a new trading partner. Then ask what makes it take longer. The second half is the whole question.
Somebody who has done twenty of these will tell you the timeline is rarely about mapping. It is about waiting. Waiting on the partner’s certification team. Waiting on label approval. Waiting on a test cycle that only runs on Tuesdays. Always Tuesdays. And underneath all of it, the companion guide, which is the partner-specific document explaining exactly how they deviate from the base standard, and which is almost always both mandatory and slightly out of date.
The version I like best goes at it sideways. “Tell me about a partner spec that turned out to be wrong.”
Anyone with real hours has this story. The guide says 5010, the test files are 4010. A qualifier is documented as optional and rejected as required. A segment appears in production that appears nowhere in the documentation. What I am listening for is not the anecdote, it is what they did next, because the correct move is to get the partner on the phone with evidence in hand rather than to quietly code around it and never tell anyone. Coding around it works. It also compounds. It builds a map that exactly one person on earth understands, and eleven years later that person retires.
Worth asking about transport too, and worth keeping it short. AS2 with signed MDNs, SFTP with keys, a VAN with interconnects, an increasing number of partner APIs. A candidate should be able to tell you which they have run and, more usefully, what goes wrong with each. Certificate expiry is the classic. Every single year. It takes down AS2 at two in the morning, it is entirely predictable, and it still happens constantly because nobody owned the calendar.
Two Questions Only a Payer-Side Analyst Can Answer
If the seat touches claims, two questions will separate the candidates faster than anything else in this post, and neither requires you to know healthcare EDI yourself. You only have to recognize a real answer when it arrives.
A healthcare seat lives in the 837 claim, the 835 remittance, the 834 enrollment file, the 270 and 271 eligibility pair, and the 277CA. Under HIPAA, covered entities have to use an adopted standard from ASC X12N or NCPDP for these transactions, and the currently mandated X12 version is 005010, per CMS administrative simplification guidance. X12 published updated 008060 versions of the HIPAA implementation guides in late 2025, but federal rulemaking has not adopted them, so anyone telling you in an interview that the industry has moved on is a step ahead of the regulation.
Now the questions themselves. Both are cheap to ask. Neither takes long.
“Where does a 999 tell you something a 997 would not?” The 999 was built to report implementation guide syntax edits, which means it can point at the specific loop and segment inside the guide that failed rather than just flagging the functional group. A candidate who has worked payer traffic answers this comfortably. A retail analyst usually has not seen one. That is fine.
“An 835 came back and the posted amounts do not reconcile to what we billed. What is your process?” Strong answers separate the adjustment reason codes from the actual data problem, check whether claims were split or bundled on the payer side, and go looking for the claims that never made it past the 277CA in the first place. That last part is the tell. The claims that hurt are not the denied ones. They are the ones that silently never arrived, and if the seat you are filling sits in revenue cycle, our healthcare IT and revenue management staffing practice is closer to the right door than a general IT req.
The API Question, and What a Good Answer Sounds Like
Every EDI interview in 2026 eventually gets to some version of “is EDI dying.” It is not, and a candidate who says it is has told you something about their judgment. Something useful.
What is actually happening is hybrid. APIs handle the front door for new partners and real-time lookups, event-driven patterns handle notification, and X12 keeps carrying the high-volume regulated and retail traffic where the trading community already agreed on a format thirty years ago and has no reason to renegotiate. Cloud-native platforms and iPaaS layers like Boomi, Cleo, Jitterbit, MuleSoft, and IBM Sterling have made onboarding faster. They have not made the standard go away. Nothing has.
So ask the question that matters. “A partner offers you an API instead of AS2. What do you ask before you say yes?”
Good answers get suspicious in a productive way. Does the API replace the EDI or wrap it, and do we still owe them an acknowledgment of some kind. What are the retry and idempotency semantics, because a duplicate purchase order created by a retry is a worse outcome than a failed file. Who owns reconciliation now that there is no functional acknowledgment to reconcile against. Is the API versioned, and what happens when they deprecate v1. Where does the audit trail live. Ask all five. A candidate who just says yes because APIs are modern is going to hand you a partner connection nobody can prove delivered anything.
The inverse question is at least as useful. “What would you refuse to move off EDI, and why?” The answer I like involves regulated transactions, high-volume partners with mature scorecards, and anything where the acknowledgment chain is the audit trail. If you are already carrying a pile of one-off connections and wondering how to untangle them, our post on decoupling load-bearing point-to-point integrations covers the architecture side of the same conversation.
Three Broken Mornings
A question gets you the answer somebody rehearsed. Hand them something already broken and you get the actual person. Use the broken things. Pick one of these for a senior loop and give it a full thirty minutes.
Morning one. The duplicates. Overnight, a batch of purchase orders came in twice. The ERP created both sets. Procurement has already released some of them. Two hours ago you resent a file a partner said they never got. Go.
What you want is somebody who connects those last two sentences without being led there. Interchange control numbers exist precisely so a receiver can detect a duplicate, and plenty of environments either do not check them or check them in a window too short to catch a re-drop from the prior day. The strong candidate contains the damage first, gets a list of the duplicated POs to procurement before procurement finds out from a vendor, and only then goes looking for whether the control number logic or the resend procedure is the actual defect. The weak candidate starts debugging. Understandable. Wrong order.
Morning two. The silence. A partner that sends forty 850s a day sent zero on Thursday and zero on Friday. Nobody noticed until Monday, when a buyer asked why an account had gone quiet.
This one is not about the fix. It is about monitoring, and monitoring is the single most underbuilt thing in most EDI environments I see. Alerting on failures is easy and most shops have it. Alerting on absence is harder, because absence looks identical to a quiet week, and it requires somebody to have written down what normal looks like per partner. Ask what they would have instrumented. If they describe a volume baseline with a threshold and a named owner, they have been burned before. Hire that.
Morning three. The Friday change. A mapping change went into production at four on Friday afternoon. It is Monday. Something is wrong with pricing on one partner’s invoices and finance is asking.
Change control, plainly. Nothing exotic. Was there a test partner or a test loop. Is there a way to reprocess. Can they roll the map back without taking down the other eleven partners on the same translator, which is where a lot of environments discover their maps are less isolated than they assumed. And the human part, which is whether they tell finance what happened before finance figures it out. I weight that more than I probably should.
The Quiet Disqualifiers
Everybody catches the obvious problems. What gets through a panel is the candidate who sounds fluent for thirty minutes and thins out the second you ask a follow-up to a follow-up, and that person costs you a year.
Watch the candidate who names tools and never names work. “I know Sterling” is not an accomplishment. It is a noun. Ask what they built in it, for which partner, and what broke. If the answer stays at product level after two follow-ups, they have probably been adjacent to this work rather than in it.
Somebody who is relaxed about carton-level ASNs has not done many. That is not a knock. Just a fact. It is a scoping signal, and it means you should not put them in a retail compliance seat and expect a short ramp.
Then there is the candidate who cannot describe a single trading partner relationship. This job has humans on the other end of it. Every real EDI analyst I have placed can tell you about the partner contact who actually answers, the one who never does, and the certification team that takes three weeks. Give the conversation forty minutes and count the proper nouns. No retailer, no payer, no partner contact, no volume figure? You have been listening to a job description read back to you.
Ask them to read a raw file out loud. Print an ISA and a handful of segments, hand it over, and ask what they see. It sounds almost rude. Ask anyway. It is the fastest honest signal in the entire loop, and the people who have done this work visibly relax when you hand it to them.
One more. If every problem in their telling was the VAN’s fault, or the partner’s fault, or the ERP team’s fault, note it. EDI sits between four systems and three departments, and nobody in that seat gets to be a bystander. Nobody.
One Note on the Band Before Anyone Walks In
A band set from a single tracker will sink a good loop before it starts, and this title is worse for that than most. The published averages for “EDI analyst” range from the sixties to the low one-teens depending on which site you open, because each one is averaging a different mix of seats. Our breakdown of EDI analyst pay bands takes that apart properly, including the part where the word in the title moves pay more than the years behind it.
Two things worth knowing before you interview anybody. There is no federal number to anchor to, because the Bureau of Labor Statistics does not code EDI analysts at all. Its closest tracked occupation, computer systems analysts, showed a $105,850 median wage in May 2025, with 8% projected growth through 2035 and roughly 32,900 openings a year. That is a floor, not a rate. Nothing more.
And the candidate who can genuinely map and develop, rather than monitor and re-drop, will price closer to an integration developer than to an analyst. If that is who you are actually interviewing, our integration architect staffing practice is the right desk and the band needs to move. For a live range on your title and metro, the salary benchmark assistant beats any national figure on this page.
How to Build the Loop
Three rounds for a standard analyst seat. Sometimes two. Four if the role owns architecture or a migration.
Round one runs thirty minutes and covers two things. Which of the three jobs this actually is, and whether the band clears. Which partners have they owned, which transaction sets, at what volume. A good share of the applicant pool disqualifies itself right there, which is most of the point of the round.
Round two is the acknowledgment ladder, the ASN, and onboarding. This is the round that tells you whether the person can hold a production environment. Do not let it drift into trivia questions. Cut those on sight. Anything a person could answer after one night with a glossary belongs in the bin, replaced by something broken.
Round three is one of the three broken mornings, in conversation, with whoever runs the warehouse or the revenue cycle sitting in. Grade the path, not the destination. Somebody who talks their way to a wrong conclusion out loud, showing every turn, is worth more to you than somebody who lands the correct term and cannot say why it is correct.
Add a fourth round for senior seats and put the controller or the operations lead in it. EDI failures get discovered by other departments, always, so you want to hear that conversation once before it is a real one. No take-home. It measures patience rather than skill, and the strong candidates decline unpaid work anyway.
Six Things Hiring Managers Ask Us About Running the Loop
Is handing a candidate a real file a security problem?
Not if you prepare one. Take a real interchange, replace the trading partner IDs, the addresses, the item numbers, and the dollar amounts, and keep the structure exactly as it was.
Structure is the whole point of the exercise. A candidate who can read an envelope will read a sanitized one just as well, and one who cannot will stall on either. Print it. Handing over a piece of paper removes the “I would look it up” escape hatch that a screen share leaves open, and it takes ten minutes to build a file you will reuse for every EDI search you run from here on.
Our strongest candidate has no EDI title anywhere on their resume. Should that worry us?
Frequently it is the opposite. A lot of the best people in this work arrived through a warehouse management system, an ERP team, or an accounts receivable desk that got tired of researching deductions.
Title history is close to worthless here and the certifications are worse. Both mislead. Ask what they own rather than what they were called. If somebody spent three years reconciling ASN discrepancies for a distributor and taught themselves the mapping tool because nobody else would, they have already done the job under a different name. What actually matters is whether they have seen a production environment misbehave, and that is a question, not a resume field.
How much should experience with our specific translator matter?
Less than most panels weight it. Somebody who has genuinely owned mapping in one platform picks up another in a few weeks, because the hard part was never the interface.
Ignore the logo. What matters is depth of the environment rather than the name on it. An analyst who has run a large IBM Sterling B2B Integrator install with hundreds of partners has managed a kind of complexity that a small SPS Commerce footprint does not teach, and that difference is real regardless of what you run. Ask about partner count, transaction volume, and whether they inherited the environment or built it. Those three answers tell you more than the product name on the resume.
Our last EDI hire looked great on paper and washed out in five months. What did the loop miss?
Nine times out of ten, the loop measured what a person can say rather than what they have carried. Those two things sit unusually far apart in EDI.
Go find the question list you used last time. Anything on it a candidate could have answered off a reference sheet was never going to sort anybody, so pull those and drop in the broken mornings above. Hand the finalist a raw file. Rebuilding the loop takes an afternoon. A second washed-out hire takes two quarters.
Who belongs on the panel?
Whoever gets the phone call when EDI breaks. In most companies that is the warehouse or revenue cycle lead and somebody from finance, and neither of them is usually invited. That is the bug.
IT running this loop alone is the most common structural mistake I see. The questions that matter are about what happens to a shipment or a claim, and the people who feel that are the ones who can tell whether an answer is real. Keep the panel to three. Four is already a committee, and by five you are watching people perform for each other instead of listening to the candidate.
How many rounds is too many for this role?
Four is the ceiling and three is usually right. The pool for this title is small enough that a fifth round mostly buys you the chance to lose someone to a company that moved faster.
Speed is a real competitive advantage in EDI specifically, more than in most IT searches, because strong candidates are rarely on the market by accident and rarely on it for long. If your process needs five conversations to reach a decision, the problem is the questions rather than the number of rounds. Compress. Then contract-to-hire is a legitimate way to buy certainty without stalling, and we run both models through our contract staffing and direct hire staffing desks.
The Person You Actually Want in This Seat
Tidy answers are not the signal. The person you want has already had the bad version of this job, the quarter where something structurally perfect was quietly wrong and nobody caught it for six weeks, and they now check the boring things first because of it. They fixed it, eventually. They still bring it up unprompted. None of that is on a resume, and none of it surfaces unless your questions have something broken inside them. Hire that person.
Fill the loop with things that are already broken. Everything downstream of that gets easier to run. Our twelve-month retention on placements sits at 92% across 30-plus U.S. metros, and the reason it holds is boring. We screen for the bad quarter, not the good interview. Running this search alone is fine. If you would rather not, talk to a KORE1 recruiter and bring three numbers with you. Transaction sets in scope, partner count, and what last year’s deductions actually cost. We source this talent through our logistics and supply chain IT staffing and manufacturing ERP staffing practices, and through the ERP recruiters who work these desks every week.
Hiring across the broader ERP stack instead of specifically into EDI? Our ERP consultant interview questions cover the same discipline one layer up. Where the seat leans toward requirements and process rather than mapping, our business analyst interview questions are the closer fit. And if you are still deciding what the req should say, the job description template and the 2026 pay bands for the role come before any of this.
