ERP Selection in 7 Steps: The Complete Guide

An ERP selection succeeds when you run it as a structured process rather than a gut decision: you first analyze your needs, pour them into a requirements specification, narrow the market down via longlist and shortlist, test the favorites in demos and reference calls, uncover the gaps with a fit-gap analysis, and finally decide based on the real total cost. This guide walks you through the seven steps for finding an ERP system that still fits five years from now.

The most common mistake is to look at a single product too early. Anyone who sees a demo and falls in love unconsciously shapes their requirements around what they just saw. Reverse the order: first you know what you need, then you look at who delivers it. That is exactly why ERP selection exists as a discipline in its own right.

Step 1: Analyze needs and map current processes

Before you evaluate any system, you need to know what your company actually does. Map your core processes end to end: from inquiry to paid invoice, from goods receipt to return. For each process, record who is involved, which systems are in play today, and where things get stuck.

Enrich the processes with hard numbers. How many orders, items, locations, entities, users? How much does volume fluctuate seasonally? These volume figures will later decide questions of scalability and licensing cost. Turn them into measurable goals — such as "halve order lead time" or "real-time inventory management" — so you can prove the project's success afterwards.

Involve stakeholders early

An ERP selection rarely fails on the technology and often on the people. Involve the departments that will work with the system every day right from the start. They know the edge cases that appear in no process diagram, and their acceptance decides the later project success. Deliberately factor in change management at this stage already.

Step 2: Write the requirements specification

The requirements specification is the heart of your selection. It describes, from the client's point of view, what the system must be able to do — solution-neutral, without pre-empting any particular software. Structure the requirements by functional area (sales, purchasing, warehouse, finance, production) and prioritize each one.

A three-tier prioritization has proven effective:

The requirements specification is distinct from the functional specification: the selected vendor produces that one later and describes how they will concretely implement your requirements. A clean requirements spec is the precondition for making the offers comparable at all — and protects you from scope creep during the project.

Put DACH obligations into the requirements spec

Regulatory requirements in the DACH region belong in the requirements specification from the outset, not in later rework. For Germany, the key ones are:

• GoBD: Accounting-relevant data must be archived in an unalterable, traceable and audit-proof manner, with process documentation.
• E-invoicing (B2B): The obligation to receive structured invoices per EN 16931 (e.g. XRechnung, ZUGFeRD) has applied since 01/01/2025. The obligation to issue them is staggered: generally from 01/01/2027 for companies with more than 800,000 euros in prior-year revenue, and from 01/01/2028 for everyone else.

Austria and Switzerland have their own rules (RKSV and GeBüV respectively) — when in doubt, check the concrete implementation with your tax advisor. What matters is that your candidate supports these formats and deadlines as standard.

Step 3: Build a longlist

Only now does the market come into play. Research systems that fundamentally fit your industry, company size and desired operating model. A longlist typically comprises 8 to 15 candidates. A good starting point is a neutral ERP directory where you can filter systems by criteria.

Watch out here already for fundamental decisions that rule out many candidates:

• Operating model: Cloud ERP as SaaS, or rather on-premise in your own data center?
• Orientation: Industry solution, broad suite or best-of-breed approach?
• Target size: Is the system built for your segment — or for significantly smaller or larger companies?
• Ecosystem: Are there partners, interfaces and a solid roadmap?

The range spans from lean SME systems like Billbee or JTL through mid-market solutions like xentral, weclapp or Odoo to enterprise platforms like SAP S/4HANA Cloud or Oracle NetSuite. Important: at this point it is only about "fundamentally fits", not yet about a ranking.

Step 4: Narrow down to a shortlist

Evaluate the longlist rigorously against your must-have criteria from the requirements spec. Every unmet knockout criterion eliminates a candidate — no discussion. This melts the list down to 2 to 4 serious vendors that you examine in depth.

For this shortlist you request structured proposals. Give the vendors your requirements specification and have them state, per requirement, whether they meet it as standard, via customizing or not at all. This makes the answers comparable and provides the basis for the later fit-gap analysis. If you want to set up the comparison of several systems systematically, a structured ERP comparison helps as a framework.

Step 5: Review demos and references

Standard sales demos show what the system does best — not what you need. Turn it around: give each vendor two or three of your real processes as a script and have them run exactly those live. That way you see how many clicks a typical order entry costs and where the system stumbles on your edge cases.

Question reference customers deliberately

Do not rely on glossy references. Ask for contacts to existing customers of similar size and industry and put concrete questions to them: how did the rollout really go? Were budget and timeline held? How does support respond? An honest reference call reveals risks that appear in no proposal — such as looming vendor lock-in or expensive release updates.

Step 6: Run a fit-gap analysis

The fit-gap analysis is the moment of truth. You set every requirement against your candidate and mark whether it is covered as standard (fit) or a gap remains (gap). For each gap you assess three things: the effort of the adaptation, the technical and economic risk, and the question of whether you could not simply do without the requirement.

The decisive insight: every customization you build around the standard costs money not just once, but again with every release update. A system with 80 percent standard fit and few critical gaps is almost always the better choice than one that only fits after extensive customizing. So check honestly whether a gap really is a must — or just a cherished process that you could adapt after all.

Step 7: Compare TCO and sign the contract

The license price is only the tip of the iceberg. Compare the candidates on the real total cost of ownership (TCO) over a realistic period of usually five years. Only then do a cheap system with expensive adaptation and a pricier system with high standard fit become fairly comparable.

Account for at least these items:

• License/subscription costs per user and module over the contract term
• Implementation costs: consulting, implementation, data migration and training
• Adaptations from the fit-gap analysis — including update maintenance
• Interfaces and integration to shop, marketplaces, DATEV/BMD, shipping
• Ongoing operation: support, hosting, internal administration
• Change and exit costs at the end of the term

Before you sign, negotiate the details: a clearly defined scope, binding service levels, responsibilities during the rollout, and fair exit and data-export clauses. If you lack the experience for a neutral evaluation, independent ERP consulting can review the proposals and contracts on equal terms.

Conclusion: ERP selection as a structured process

A good ERP selection is not a gamble but the result of seven cleanly executed steps: analyze needs, write the requirements spec, build longlist and shortlist, review demos and references, run the fit-gap analysis, and decide based on the real total cost. Whoever keeps to this order does not buy the system with the prettiest demo, but the one with the best fit to their own processes — and makes a decision that still holds up in five years.