Back to Blog

ERP Project Manager Interview Questions 2026

ERPHiringIT Hiring

Last updated: November 20, 2026

By Tom Kenaley, Senior Partner and President, KORE1

The best ERP project manager interview questions test governance, not scheduling: how a candidate baselines requirements, classifies change, checks integration capacity, rehearses cutover, and keeps a system validated after go-live. Familiar behavioral prompts never reach any of it.

1,350,000 calls a year against a license for 130,000.

That was the gap between two ways of wiring the same NetSuite integration for a mid-market pharma manufacturer, and it took no insight to find, only division. A design that touches the vendor’s API once for every unit shipped eats about ten times the allowance. Not close. A design that batches by document needs somewhere between 10,000 and 20,000 calls. The proposal we have benchmarks from did that arithmetic on page six. Plenty of programs do it in month eight, right after the first load test starts getting throttled. Expensive month to learn it.

I ask every ERP project manager finalist some version of it now. I’m not testing the math. I’m testing whether they would think to ask for the allowance at all.

Our ERP implementation project manager staffing desk sits in my group, and we get paid when a client hires through us, so discount my opinions accordingly. You can run everything below without us.

Some of the ground has a home already. The general behavioral loop is in our project manager interview questions, and the cutover and rollback round for live production systems is in the IT project manager version. Whether to hire at all, what the seat pays, and the marked-up vendor quote exercise are in our guide to hiring an ERP project manager. The person doing the configuration gets their own list, the ERP consultant interview questions. This page covers only the program-governance questions that are particular to ERP.

Packet of blank pages with four orange flags and a red pen on a gray desk beside a steel ruler and a coffee mug

Table of Contents

What This Interview Is Really Testing

An ERP project manager interview tests whether a candidate can govern a software program from the buyer’s side of the table: baselining requirements, classifying and pricing every change, holding a vendor to its statement of work, and deciding when a go-live is actually ready. It does not test whether they can draw a schedule.

Panorama Consulting’s 2026 ERP Report, released March 4, found that more than a quarter of organizations exceeded their project budgets, with additional technology needs named as the leading cause. A Panorama spokesperson described how that happens. Organizations discover serious misfits late, so they reach for more technology, wider scope, and custom builds. Panorama sells ERP consulting and has a point of view, but the mechanism matches what clients tell us when a program is already in trouble.

Read that as a job description. The person in the seat is the one who finds the misfit in month two. If you’re writing the posting now, our ERP project manager job description template has the blocks.

Requirements and Baseline Questions

Start here. A program that baselines bad requirements will spend its whole budget proving it, and the candidate’s answers in this section tell you more than the ones about schedule.

Here is a line from a real requirements document: “The system shall be user friendly.” What happens to it?

The strong candidate won’t baseline it, and says why. Nothing can be tested against “friendly.” Friendly to whom? They’ll ask what outcome the business actually needs, then rewrite it as something checkable, such as a clerk with an hour of training being able to enter and route a vendor bill without help. They’ll also want a unique identifier on the requirement so a test script can trace back to it later, and a note on how much it matters if it fails, because testing depth should follow criticality rather than uniform rigor.

Weak candidates swap the adjective for a better adjective. Same requirement, nicer word.

The partner offers to draft the requirements for you. Do you let them?

Mostly no, and the reason matters. A partner can run the workshops and hand over a template, and a good one will. Authorship and approval should stay with you, because acceptance criteria that originate with the builder let the builder write the exam it will later be graded on. Listen for a candidate who has watched that happen. One regulated buyer we know insisted its own quality group approve every requirement before a line was configured, and the integrator, to its credit, wrote its response around that position instead of fighting it.

A requirement approved last week turns out to be wrong. Walk me through reopening it.

It’s a change now, not a conversation. Paper first. The answer should include who asked, what it does to cost, schedule, and anything already tested, and a written approval from a named person on each side before anyone touches it. “We’d just fix it, we’re a collaborative team” is the most common answer. It’s also the least useful. Collaboration is how the original misunderstanding got baselined.

Change Control, Where the Money Actually Goes

The hiring guide has the change log exercise. These two are verbal. They sort candidates fast.

Finance returned the mapping spreadsheet eleven days late, and the data load slipped two weeks. Who pays?

Depends what the delay caused. If the integration team sat idle, that’s waiting, and waiting earns schedule relief only. If the team had to rebuild transformations because the mapping that finally arrived was wrong, that’s rework, and rework is billable. Different invoices. The two get confused constantly, which is why vendors can send an invoice for a delay that cost them nothing. A candidate who separates them without being prompted has probably sat across from a change order before.

Ask who decides which one it was. The answer should be written down, with both sides signing.

Three small changes land in the same week. Each is under a day of work. Do you assess them one at a time?

One at a time, and then together. Each gets its own classification and its own reasoning on the record, because a change that touches nothing important still needs a written reason for why it touches nothing important. Then somebody assesses the combined effect, since three changes that are harmless alone can disturb something that was already tested. Bundling small changes to skip the assessment is a habit. An old one. Ask whether they’ve ever caught it, and who on their team they caught.

  • Which of your last program’s change orders did you refuse, and what did the requester say?
  • Who held the pen on classification? Not who typed it. Who decided.
  • What was the agreed scope at go-live, in one sentence, and where was it written?
Two gray chairs facing each other across a small round table with a notebook, a stack of pages with orange clips, and two mugs in a glass meeting room

Integration and Capacity

This is the section almost nobody has in their loop, and the one where a PM who came up through the business side gets found out. Nobody expects code. They have to know which questions to make the builders answer.

The partner proposes calling the vendor’s API once per unit shipped. What do you ask for before you say yes?

Three things. The licensed call allowance, the real monthly volume in documents and in units, and the number of calls each transaction needs. Then the division. In writing. In the case I opened with, a unit-level design would have used about 1.35 million calls a year against 130,000, and a transaction-level design lands under a quarter of the allowance and leaves room for growth. The better candidates add one more request, which is to measure the call volume in testing and not just assert it on a slide.

The weak answer is “I’d trust the architects.” You might. They’re also the people who drew it.

A vendor says its callback “just posts with an API key.” What goes on the plan?

Here’s a trap from NetSuite that I like because it’s so small. A RESTlet won’t accept a bare API key from an outside caller. Oracle’s help page on RESTlet authentication points external clients to OAuth 2.0 or token-based authentication, and it carries a warning that support for token-based authentication on RESTlets ends in NetSuite 2028.2. If the vendor’s callback can only send a static header, somebody has to build an authenticated relay in front of the RESTlet. Small piece of work. Real one, though, and nobody priced it.

What you want from the candidate is not the technical answer. It’s the move. They put it on the plan as a named confirmation item for the first phase, with an owner and a date, and they decide up front what it costs if the answer is no. Weak candidates say the developers will sort it out, and the developers find out in the fourth month of testing.

Other platforms have the same shape with different nouns. Whether it’s SAP S/4HANA, Dynamics 365, or Acumatica, every integration has an authentication method and a throttle, and a program manager should be able to say where both are written down.

Data, Cutover, and the First Close

General cutover judgment, who can call the rollback and what number triggers it, is covered in the IT project manager list. These are the ERP-specific ones.

A finalist for a Dallas distributor told us her third full data rehearsal ran past the cutover window, and she moved go-live a week instead of squeezing the weekend. The CFO hated it. She was right. She told the panel that story as a point of pride, which is the right instinct.

  • How many full rehearsals of the data load did your last program run, and what was different about the final one?
  • Who signed that the opening balances were right? Not the PM. Find out who, and whether the candidate knew their name.
  • What was on the plan for the first month-end close after go-live, and who was in the room for it?
  • Which data did you cut from migration, and who agreed to the cut?

That last one separates people fast. Data migration is the thing vendors often exclude or scope thinly, and a candidate who can name the exact table that got left behind has lived through the argument about it.

If the Plant Is Regulated

Life sciences, devices, food, and aerospace shrink the pool, and these three questions are most of the reason. Skip the section if nothing you make has to be validated.

The integration is simple to validate. The platform underneath it was never qualified. Now what?

Raise it in planning, not at performance qualification. A qualification of serialization flows that rest on unqualified inventory, lot traceability, and status control only shows the integration behaved as specified on a platform nobody has qualified. The recognized fix for a system already in production is a retrospective qualification, which is a documented gap assessment against intended use followed by targeted qualification of the functions that matter, using operating history as evidence where it holds up. It’s smaller than buyers fear and it is priced separately from the integration. Candidates who know the ISPE GAMP 5 second edition usually say this before you ask.

A partner says its low-code integration platform cuts the validation effort. What do you ask?

Where the scripted steps are. Realistic builds on those platforms still contain scripted transformations, and those remain bespoke code inside a configured product. GAMP 5 treats the software categories as a continuum assessed per component, so the saving is smaller than the brochure suggests. Then the second round of questions. You’ve added a supplier to qualify and a second environment to your validated boundary, so what did that cost?

Where in the plan do electronic signatures get priced?

Short answer. The design phase. Audit trail and approval routing come with most platforms. What 21 CFR Part 11 asks for on a signed record is different, and the signature manifestation rule requires the printed name, the date and time, and the meaning of the signature to appear with the record. Getting that, plus the link between signature and record, usually takes deliberate configuration or an added component. When nobody priced it, it surfaces in operational qualification, expensively. A good candidate asks which records carry signatures before asking anything else.

After Go-Live, and What Changed in 2026

The seat doesn’t end at cutover, and two questions get at the part that does not.

Twice a year the platform changes underneath you. Ask what they’d do the week a release preview lands. Good answers run the existing regression set against a preview sandbox, assess what the release touched, and document it to the standard the periodic review expects. Weak answers wait for production to tell them. In a regulated account that’s how a validated state lapses without anyone deciding to let it. Nobody chooses that.

The new one is AI. Partners are drafting customization code with generation tools, and that’s fine, but ask this: would you let the tool that drafted the customization also draft its test scripts? The right answer is no, a person writes the tests, because independence of the test from the code is the control. The best written policy I’ve read on this requires a named, qualified author of record for every deliverable, generation against an approved design instead of a loose prompt, and review before merge by someone who isn’t the author. A candidate doesn’t need to quote that. They do need to know that the question is worth asking in selection, while there’s still room to negotiate.

How to Score the Answers

Write the strong and weak signals down before the first interview. This is the sheet we give clients.

Question AreaStrong SignalWeak Signal
RequirementsRefuses to baseline what can’t be testedPolishes the wording and moves on
Change costSeparates waiting from rework unprompted“We’d negotiate it”
Integration capacityAsks for the allowance and the divisionTrusts the architects
UnknownsTurns them into owned, dated plan items“Dev will sort it out”
CutoverMoved a date after a failed rehearsalNever missed a date, can’t say why
ValidationRaises the unqualified platform earlyDiscovers it at the protocol stage
AI-assisted workA person writes the tests, someone else reviews“The tool handles it”

Two scorers, separate sheets, one to four per row, compared only after both are finished. A three-point gap on the same row means the interviewers heard different answers, and sorting that out is where the useful ten minutes of the debrief go.

Answers That Cost a Candidate the Offer

Four, and they’re consistent. The first is any story where the partner’s project manager handled whatever went wrong. That person works for the partner, which is the whole reason you’re hiring a second one. Wrong side of the table. Then a go-live count offered as the proof of competence, with no account of what changed after the first approved scope. Counts aren’t outcomes. Third is the candidate who has never read a contract, and says so as though it’s a virtue, because “legal handled that” is not an answer about a program whose scope lives in a statement of work. The last is quieter. Ask what they’d do in week one, and if the plan arrives complete, with a governance model and a RACI and a communications calendar, they built it before learning anything about your company.

If You’re the Candidate

Leave the polish at home. Bring one number you checked yourself and one thing you got wrong. The candidates who win these loops describe a rehearsal that failed, a change order they refused, a question they put to a partner that the partner didn’t enjoy. Hiring managers have heard the polished version of every ERP story already. If a panel only asks about methodology, it’s fair to ask what’s actually in trouble. That tells you whether you want the seat. Know the going rate before the call, too, and the ERP project manager salary guide shows where this seat lands in 2026.

Before You Schedule the Loop

How many rounds should an ERP project manager loop run?

Two substantive rounds and a short reference conversation, finished inside ten working days, because experienced ERP project managers are usually on contract somewhere and take another offer while you coordinate calendars.

KORE1’s average time-to-hire across IT roles is 17 days, and most searches that blow past it die in scheduling, not in sourcing.

Should our implementation partner sit on the panel?

No, not as a scorer, because the partner has a margin to protect on every call this person will make, and a candidate who pushes back on them in the interview is showing you the job.

Introduce them after the offer instead.

Can we test any of this without a live ERP system?

Yes, hand them a redacted packet with a past change order, a one-page integration design, and ten real requirements, give them thirty minutes, and ask what they would send back.

Thirty minutes is plenty.

A great candidate has run three go-lives but never in a regulated plant. Disqualify?

Not automatically, since regulated experience narrows the pool hard and is usually worth waiting for on validated work, while an unregulated mid-market program can lean on change-control habits that transfer.

The validation questions above simply won’t apply there. Ask what they’d want to learn in the first month. Honest answers to that tell you most.

Do platform-specific questions change by system?

A little, because the governance questions stay identical across NetSuite, SAP S/4HANA, Dynamics 365, and Acumatica, and only the integration authentication and release-cycle questions change with the platform your program actually runs.

Same bones, different nouns.

Ask the Question Your Own Program Is Afraid Of

The arithmetic question works because it’s specific to a real design. Yours should be too. Pull the last status report your sponsor read without comment, find the number nobody has divided yet, and put it in front of the candidate. They’ll sort themselves in about four minutes.

We’ve placed technical and program talent across 30+ U.S. metros since 2005, our recruiters average 15+ years in the business, and 92% of the people we place are still in the seat a year later. A lot of ERP program seats begin on contract staffing terms and convert later. If you’d like a second read on your question set, or a slate of project managers who’ve already handled questions like these, talk to a recruiter on our ERP desk. If you’re still early and weighing why programs stall in the first place, our look at the staffing root causes behind ERP project failures is a good place to start.