Last updated: August 21, 2026
By Robert Ardell, Co-Founder and Strategic Advisor, KORE1
Hire an ERP consultant when the work in front of you changes how the system is designed, and hire another administrator when it only changes how the system is configured. That one distinction settles the argument faster than any symptom list, including the seven below. Difficulty is not the test. Design is.
Sit with two numbers for a second. 1,350,000 against 130,000.
Both are annual NetSuite API calls. The first is what a serialization integration would have burned had it been built to touch the API once per physical unit. The second is what the company was actually licensed for. Over by a factor of ten, before anyone wrote a line of code, and the whole thing was visible in a volume table during discovery. Batch the same traffic at the document level instead of the unit level and the integration lands near 10,000 to 20,000 calls a year with headroom to spare. Same requirements. Same platform. One arithmetic decision between a system that works and a system that starts throttling in month two.
A very good ERP administrator would not have caught that. Not for lack of skill. Nobody would have put the question in front of them, and it does not look like an administrator question until after it has cost you.
Fair warning about who is talking. KORE1 runs an ERP recruiting practice and we place administrators and consultants both, so I have no fee riding on which way you go. What I have is twenty years of watching companies open the wrong req and spend four months proving it. There is a section near the bottom that tells a fair number of you to hire nobody at all.
The Word Consultant Is Measuring Two Different Things
An ERP consultant is brought in to change how the business runs inside the system: process design, the data model, the integration pattern, the controls, the boundary between what the ERP owns and what it does not. An ERP administrator keeps the system running as designed. Users, roles, fields, forms, saved searches, approvals, the report finance needs by Thursday.
Now look at what the market pays each of them, because the numbers hide something.
Glassdoor’s 2026 data puts the average ERP administrator at $105,393, with the middle half landing between $84,103 and $133,161. ZipRecruiter puts the average ERP consultant at $113,186, which works out to $54.42 an hour, middle band $80,500 to $145,000. Roughly eight thousand dollars apart. If those two figures were the whole story you would not need this article, or me.
Except neither figure describes the thing you are probably about to buy. Both track employees who carry the title on a badge. An engaged specialist prices somewhere else entirely. A NetSuite serialization proposal I read this year priced delivery at a blended $215 an hour across every role on the team, with $275 for the independent validation lead, the only rate differential in the entire document. Four times the ZipRecruiter hourly figure, for work that shared a job title with it.
Same word. Two entirely different transactions. So when somebody in your leadership meeting says let’s just get a consultant, find out which one they mean before the number goes in the budget.
The Line Is Not Difficulty. It Is Whether the Design Changes.
Change control turns out to be the cleanest way to draw this, and I am borrowing the frame from a real quality document rather than inventing one. Foretopia, whose CIO Colin Boothe writes for us on ERP data migration, sorts every change on an engagement into three classes before anybody touches anything.
| Class | What It Covers | Who Should Own It |
|---|---|---|
| A | No effect on agreed scope, cost, schedule, or validated state. Includes correcting a defect against a specification that already exists. | Administrator. Record it, approve it internally, move on. |
| B | Affects scope, cost, or schedule, but not the validated state of the system. | Consultant territory. Somebody has to price it and own what it breaks downstream. |
| C | Affects the validated state, whether or not scope, cost, or schedule move at all. | Consultant, plus a reviewer who is not the person implementing it. |
Look at what that table leaves out. Nothing about how hard the change is. Nothing about hours. A twenty-minute change can be Class C and a three-week change can be Class A, which is precisely backwards from how most companies decide who to call, and it is why so many of these decisions get made on gut feel about effort.
Two more rules from that same document belong in any ERP shop, regulated or not. Changes are never bundled together to avoid assessment. And the person who reviews a change is never the person who built it. Nearly every ERP mess I have been asked to help staff out of broke at least one of those, usually both, usually with everyone acting in good faith.
Seven Signs the Work Has Crossed the Line
Run these against your actual open request list. Not against how the system felt during last quarter’s close.
1. The close keeps getting longer and nobody can name the step that added the time
Everyone feels it. Nobody can point at it. That combination is the signal, more than the number of days itself, because a close that is slow for a known reason is a queue problem and a close that is slow for an unknown reason is a design problem wearing a queue costume.
One controller we worked with in Orange County went from a six-day close to an eleven-day close over about fourteen months. No single change caused it. Four separate approval routings had been added by four separate people solving four unrelated complaints, and each one was reasonable in isolation. Nobody had ever looked at them together, which is not something an administrator is usually asked to do and is roughly the entire job on the other side.
2. A spreadsheet became the system of record, and somebody will defend it in a meeting
Shadow spreadsheets are normal. The tell is the defense.
When you ask why revenue allocation lives in a workbook on a shared drive and the answer is a shrug, you have a training gap. When the answer is a fifteen-minute explanation of why the ERP cannot handle it, and that explanation is technically correct, you have a design gap. Those two conversations sound similar in the room and cost wildly different amounts to fix. Listen for the confidence.
3. “The system can’t do that” is the answer, and nobody can name the constraint
Here is where the API story pays off. There was a real constraint in that project, and it was a licensed call allowance of 130,000 a year. Nameable. Checkable. Arithmetic, not opinion.
Most of the time nobody can produce a constraint like that, because the real answer is that the system was configured a certain way in 2019 by somebody who has since left and nobody has revisited the decision. That is not a limitation. It is a leftover. An administrator will tell you honestly that the system does not do it today, which is true, and a consultant’s job is to find out whether it cannot or merely does not.
There is a related version of this worth naming. Sometimes the constraint is real and specific and nobody in the building knows it exists. A NetSuite RESTlet, for instance, will not accept a plain API key. An external system calling in has to authenticate with OAuth 1.0 token-based authentication or OAuth 2.0 signed requests, which means a vendor whose integration only speaks API keys needs an authenticated relay sitting in front of it. Small addition. Real one. Discovered in week two of a build, it costs a few days. Discovered in user acceptance testing, it costs a release.
4. You are adding an entity, a currency, a warehouse, or a company you just bought
Structural additions are the cleanest consultant trigger on this list, and they are also the ones companies most reliably underestimate, because the ERP will technically let you do all of them without help.
- A second legal entity changes your chart of accounts, your intercompany eliminations, and probably your close calendar.
- A new currency changes revaluation, reporting, and every historical comparison anyone runs.
- A new warehouse changes fulfillment logic, inventory costing, and whichever integration talks to the building.
- An acquisition changes all of the above at once, on a deadline set by someone who does not work in your department.
Any one of those is a Class B change minimum. Handing the whole set to your administrator and calling it a project is how companies end up with a second entity live in the system and a finance team that quietly keeps consolidating in Excel for another two years.
5. Somebody asked for evidence and you produced a story instead
An auditor, a customer’s quality team, a lender, a regulator. The request is always some version of show me how this works and show me that it has always worked this way.
Good change control produces records that can be handed over without reconstruction. That phrase is doing heavy lifting. If answering the question means three people spending a week rebuilding a timeline from memory, Slack messages, and a saved search somebody wrote for a different purpose, the gap is not documentation. Documentation is the symptom. The gap is that changes were made without assessment, and that is a design and governance problem, which is consultant work every time.
6. One person knows why everything is the way it is
Ask three people on your team why a custom field exists. If all three say ask Dana, you do not have an administrator. You have a dependency.
Nothing about this is a criticism of Dana. She has been holding a system together for six years, probably well, probably while also doing a different full-time job that appears on her actual job description. But an instance whose logic lives in one person’s head cannot be assessed, cannot be handed over, and cannot be safely changed by anyone else, and the day she takes a role somewhere else you will pay a consultant three times what you would have paid one this year to reverse engineer her decisions.
7. You are about to buy an integration and nobody has done the arithmetic
This one has a hard number attached, which makes it easy to check before you sign anything.
Volumes, call allowances, batch windows, retry behavior, what happens when the other system is down for ninety minutes. If the vendor’s proposal does not address the licensed limits of your own platform, and if nobody on your side has multiplied transactions by calls per transaction, then the design has not happened yet regardless of what the statement of work says. That is not a build you staff with an administrator. It is a design decision with money attached, and the arithmetic takes an afternoon at most. Do it first.
Three or more of these firing at once, and the next hire is a consultant. One of them on its own is usually a ticket.
What Each Path Actually Costs Across a Year
Rough numbers, and I would rather label them rough than dress them up. These blend the aggregator data above with what we see land in real KORE1 offers and contracts.
| Path | Rough Annual Cost | What You Get, and What You Do Not |
|---|---|---|
| Second full-time administrator | $105,000 to $135,000 plus burden | Ticket throughput doubles. Design questions still go unanswered, now with two people not answering them. |
| Fractional or part-time administrator | $40,000 to $70,000 | Right answer for eight to fifteen hours of real weekly demand. Wrong answer if the backlog is structural rather than volume. |
| Scoped consulting engagement | $35,000 to $60,000 for discovery and design, more to build | A named design, a decision record, and a fixed price on the next phase. No hands on your ticket queue. |
| Consultant first, administrator after | Both, staggered across the year | The pattern that works most often. Design gets settled, then you hire the person who runs what was designed. |
One thing worth pulling out of that table, because it causes real arguments in procurement. A discovery-and-design phase priced somewhere in the $40,000 range is not a consultant being expensive. On the serialization proposal I mentioned, discovery, risk assessment, and the validation plan came in at 182 hours and $44,110 as a fixed fee, and the build behind it was priced as a range that firmed up only after the requirements document was approved. That structure is the point. You are buying a real decision gate, not a bigger commitment. Our breakdown of 2026 ERP consultant hourly rates goes deeper on what drives the spread between platforms.
A related number from the same document is worth carrying around. Validation ran roughly 30% of base hours, which is normal for that class of software and is the reason a materially cheaper proposal is usually cheaper by exclusion rather than by efficiency. Read the exclusions section first. Every time.
Four Times the Answer Is Another Admin, or Nobody
We place ERP administrators too, and I would rather say this out loud than have you find out in month three.
If your backlog is genuinely a volume problem, meaning the requests are ordinary and there are simply too many of them, hire the administrator. Two hundred tickets of field changes and report tweaks do not add up to a design failure. It is arithmetic of a different kind, and NetSuite administrator staffing solves it more cheaply than any engagement will.
If the real demand is eight to fifteen hours a week, do not create a full-time seat you will have to justify next budget cycle. Fractional administrator coverage exists for exactly that shape and it converts to full-time cleanly when the volume shows up.
If the pain is concentrated in one upcoming project with a date on it, you want contract staffing rather than a permanent hire or a full engagement. Finite work, finite commitment. Carrying a permanent seat afterward because the project felt important is how companies end up with an expensive person and an empty backlog.
And if nobody in the room can articulate what is actually wrong, the honest first move is a diagnostic rather than a hire. Our ERP readiness assessment is free and takes about fifteen minutes, and it has talked more than one company out of a search. That is fine. A search opened against a vague problem takes three months and produces a candidate nobody can evaluate, which is a slower and more expensive version of the same confusion.
Worth reading alongside this if the specific question is who does what: our piece on the difference between a NetSuite administrator and a developer splits the technical side of the same decision.
The Objections We Hear on the Intake Call
How do I tell whether my admin is underperforming or just out of scope?
Count how many open requests would change something about how the business runs, versus how the system is set up. If most of the backlog is the first kind, your administrator is being handed design work and is losing at it the way anyone would. The tell that it is genuinely a performance issue looks different: routine, well-specified requests sitting untouched for weeks with no explanation. Different problem, different conversation, and it is worth being sure which one you are having before a review.
Can a consultant actually fix this in a two-week engagement?
Two weeks buys a diagnosis, not a rebuild. That is not a dodge. It is the honest shape of the work. A short engagement should end with a written statement of what is wrong, what it will take to fix, and what happens if you do nothing, priced. Anyone selling you a two-week fix for a design problem is selling you configuration and calling it consulting, which you will discover in about five months.
Our ERP is fine. Finance is the problem. Do we still need this?
Finance being the problem is usually the finding, not the counterargument. Systems and process are the same object viewed from two chairs, and an ERP that faithfully automates a broken approval chain has done its job perfectly while making everything worse. What decides it is whether anyone owns the process design. If the answer is that finance owns it and IT owns the system and the two meet quarterly, you have found the gap.
What does a scoped engagement produce that I can actually hold?
Three artifacts, and you should refuse to pay without them. A written design or requirements document specific enough to test against, a decision record explaining why each choice was made rather than just what was chosen, and a priced plan for the next phase with its assumptions and exclusions stated. Slides are not deliverables. If the engagement ends in a deck and a verbal summary, you bought a meeting.
We are not regulated. Does the change-control discipline still apply?
Yes, minus the paperwork. Regulated shops formalize change classification because an auditor will ask, but the underlying practice earns its keep in an unregulated business for a simpler reason: it forces someone to state what a change touches before it ships. Skip the validation documents. Keep the classification, the record of why, and the rule that the reviewer is never the builder.
Who should own the consultant relationship, IT or finance?
Whoever owns the number at the end of the month. In practice that is usually the controller or the CFO for ERP work, with IT owning access, environments, and integration security. The failure mode we see is a shared ownership arrangement where neither side can approve a change order without the other, which turns a two-day decision into a three-week one. Name one approver and put that name in the engagement letter.
Where to Start This Week
Pull your open request list. Sort it into two columns, changes to how the system is configured and changes to how the business runs. If column two is longer than column one, you already have your answer and no amount of additional administrator capacity is going to move it. Not one hour of it.
Then check one number before you do anything else. NetSuite ships two upgrades a year, and SAP, Dynamics 365, Epicor, and Sage Intacct all run their own cadences. Find out who at your company assessed the last one against the things you customized. If the answer is nobody, that is the eighth sign, and I left it off the list because it is less a symptom than a standing condition.
For context on the wider market, the Bureau of Labor Statistics projects management analyst employment up 9% from 2024 to 2034, with roughly 98,100 openings a year against a median wage of $101,190 as of May 2024. Demand for this work is not softening. And the McKinsey and University of Oxford study of more than 5,400 IT projects, published in 2012 and still the most-cited work on this, found large projects running 45% over budget while delivering 56% less value than predicted. Old research. Depressingly current.
If you want a second read on which role your list points to, talk to one of our ERP recruiters and we will scope it before you commit to a search. KORE1 has placed technical and finance talent since 2005 across more than 30 U.S. metros. Our average time to a first qualified submit is 17 days, and 92% of the people we place are still in the seat at a year. The intake call is where most of these reqs get fixed, and a fair number of them get closed instead.
