Start researching ERP and the market appears to consist of about seven companies. The same names dominate the search results, the same comparison posts recycle the same shortlists, and within an afternoon most buyers have concluded that the decision is a choice between a handful of large suites.
The actual US market looks nothing like that. A recent editorial survey of 302 ERP products actively sold in the United States — researched individually against vendor documentation and public records — produced several findings that ought to change how a first-time buyer approaches the process. Four of them matter more than the rest.
Finding One: The Generalists Are the Exception
Of those 302 products, only 35 sit in the broad horizontal categories most buyers start with — six enterprise systems, eighteen mid-market, eleven small business. The other 267 are segment or industry systems.
Thirty-five are built for discrete manufacturing and job shops. Twenty-nine for distribution. Twenty-three for construction. Fifteen for the public sector. Eleven for apparel. The tail runs all the way out to seed-to-sale cannabis tracking and grain accounting.
The practical consequence is significant. Buyers routinely take a general-purpose suite and pay to customize it into something that handles their industry's specific processes — job costing, lot traceability, progress billing, style-color-size matrices, compliance reporting — when a system already exists that does those things natively. Customization is the single most reliable way to inflate an implementation budget and complicate every future upgrade.
Before shortlisting any general erp software, it's worth establishing whether a vertical product already covers your process out of the box. Frequently one does, and the search never surfaces it because specialist vendors don't outspend enterprise suites on marketing.
Finding Two: Almost Nobody Publishes a Price
This is the finding that most changes how buyers should behave. Ten of the 302 products publish enough pricing detail to budget from. The rest quote through partners, on request, after a discovery call.
That opacity has real effects. It means any five-year cost figure you find in a blog post is an estimate built on assumptions that may not resemble your situation. It means comparing vendors on price is impossible until you're deep into a sales process with each one. And it means budget conversations with a board frequently happen on numbers nobody can substantiate.
The workable response is to build the cost model yourself, from your own inputs, and use vendor quotes to fill it in rather than to define it. A five-year total cost of ownership for an erp system needs to account for considerably more than license or subscription fees: implementation services, data migration, integration work, training, any customization, ongoing support, annual uplifts, sandbox and test environments, and — the line item almost every organization underestimates — internal staff time diverted from their actual jobs for the duration of the project.
Going into vendor conversations with your own model already built changes the dynamic entirely. You're validating assumptions rather than being handed a number.
Finding Three: The Deployment Decision Is Still Yours
Conventional wisdom holds that the cloud question is settled. The data says otherwise: only 131 of the 302 products are cloud-only. Another 163 sell both cloud and on-premises deployments, and eight remain Windows or IBM i products with no true SaaS option — several with substantial, actively supported installed bases.
There are legitimate reasons organizations still choose on-premises or hybrid: data residency and regulatory constraints, existing infrastructure investment, latency-sensitive shop floor operations, sites with unreliable connectivity, or simply a cost model that works better over a ten-year horizon. There are equally legitimate reasons to go cloud — reduced IT overhead, predictable operating expense, faster access to new releases.
The point isn't which is correct. It's that a vendor telling you the decision has already been made for the industry is describing their own product line, not your requirements.
Finding Four: Where the Vendor Sits Matters
256 of the 302 vendors are headquartered in the United States; 46 are foreign, predominantly European enterprise suites plus a handful of open-source projects.
That distinction shows up in practical ways that rarely appear on a feature comparison: support hours aligned to your working day or not, the depth of the local partner network, how thoroughly US sales tax, payroll and regulatory reporting are handled natively versus through a localization layer, and how quickly a critical issue escalates to someone who can actually fix it.
None of these disqualify foreign vendors — some are excellent — but they belong in the evaluation rather than surfacing during implementation.
What This Means for Running a Selection
Put the four findings together and a clear sequence emerges.
Write requirements before you take a demo. This is the highest-leverage step in the entire process, and it's the one most commonly skipped. A vendor demonstration is a sales instrument, professionally constructed to showcase strengths and route around gaps. Walking in without a written requirements document means the demo defines what you think you need. Walking in with one means you're scoring against your own criteria. Involve the people who will actually use the system — finance, operations, the warehouse, the shop floor — because their process knowledge is where the real requirements live.
Shortlist by fit, not by familiarity. Given that 267 of 302 products are specialists, the odds favor at least one strong vertical candidate for most operations.
Model cost independently. With ten of 302 publishing prices, self-service budgeting is the only way to arrive at a defensible number.
Treat references seriously. Ask for customers of similar size, in a similar industry, who implemented recently — and ask them specifically what went wrong, not whether they're satisfied. The useful answer is always in the second question.
Where Projects Actually Go Wrong
Two areas account for a disproportionate share of overruns, and neither is about software selection.
Data migration is consistently underestimated. Legacy systems accumulate years of inconsistent, duplicated and partially structured records. Cleaning, mapping and validating that data is slow, unglamorous work, and it cannot be compressed. Organizations replacing older platforms — Dynamics GP, NAV and SL, Infor XA and LX, Oracle E-Business Suite, Sage 100 and 500 are among the most common — should budget for this seriously and start it early.
Change management is the other. Technical readiness and organizational readiness are different things, and a system that works perfectly is worth nothing if the people expected to use it are still running the old process on a spreadsheet. The first ninety days after go-live determine whether adoption holds.
A Note on Sources
One last practical point about researching enterprise resource planning software: most comparison content online is paid placement. Rankings, "top ten" lists and lead-generation directories are frequently ordered by commercial relationship rather than merit, and the disclosure is usually somewhere you won't read it.
The signals worth looking for are structural. Does the resource state its editorial policy? Does it rank, or does it list neutrally — alphabetically, say, which is the one ordering nobody can purchase? Does it publish figures it can't confirm, or leave them out? Does it exclude products it couldn't verify as actively sold, or pad the directory to look comprehensive?
Those questions are worth asking of any source, including this one.
The Bottom Line
The US ERP market is wider, more specialized and considerably more opaque than a first afternoon of searching suggests. Most buyers are choosing from a shortlist assembled by marketing spend rather than by fit — which is how organizations end up customizing a general suite into something a specialist product already did.
Write the requirements first. Look for the vertical system before the horizontal one. Build your own cost model. And treat any published five-year figure as a starting hypothesis rather than an answer.