How to choose an APS software: 7 criteria that actually matter

Honest preface

Over the last 15 years I've seen all the big names: international packages, Italian specialists, all-in-one suites, niche solutions. They all have at least one thing they do well. None is the absolute best: there's the right one for your company.

When a client asks me "what's the best APS?", my honest answer is always: "depends on these seven questions".

1. Data model: what kind of production does it actually represent?

Not all APS think of a factory the same way. Some are great at:

• process industry (continuous, batch, recipes) — chemical, pharma, food;
• discrete with deep routings — mechanical, packaging, automotive;
• highly variable make-to-order — fabrication, custom shops;
• flow shop / mixed model — electronic assembly, white goods.

If your model is project-based with deep BOMs and the APS was born for pharma batches, you'll be fighting the tool for years. Starting question: show me three clients with my exact production model.

2. Algorithm: optimizer or constraint solver?

There's a big difference between:

• an APS based on a mathematical optimizer (linear programming, MILP) seeking the optimal plan against an objective;
• an APS based on a constraint solver generating a feasible plan that respects the rules;
• an APS based on heuristics / priority rules (FIFO, EDD, slack-based, etc.).

None is right in the abstract. For many Italian manufacturing SMEs, a good constraint solver with readable rules beats an MILP solver the planner doesn't understand. Plan explainability is a feature.

3. Integration: how does it live next to your ERP?

An APS that doesn't talk to your ERP doesn't live. Key questions:

• Native APIs or standard connectors for your ERP (SAP, Microsoft Dynamics, Sage, Galileo, Mago, IFS, etc.)?
• Realistic sync frequency — real-time or nightly batch?
• Who is the source of truth for master data, BOMs, routings, orders?
• How does it handle shop-floor feedback from MES or terminals?

An APS without real progress data plans blindly. Verify real integrations, not slides.

4. Ability to model your real constraints

Not the constraints in the brochure. Yours. Examples:

• sequence constraints (light to dark color, not the other way);
• finite tooling shared between machines;
• operator skills (shift X isn't qualified on machine Y);
• energy constraints (can't run A and B together due to contractual limit);
• chemical compatibility between consecutive products;
• physical buffer constraints.

Bring your 5 most annoying constraints to the demo. If the APS models 3 of 5, you have a problem. If it models 5 of 5 visibly with difficulty, you have a different problem.

5. UX: will the planner actually use it?

This is the most underrated criterion. A beautiful APS the planner hates becomes shelfware. Things to watch for in demos:

• replan speed: I change one urgent priority — in how many seconds do I see the new plan?
• Gantt readability: can I see the issue at a glance?
• quick actions: drag-and-drop on the order, controlled override, what-if scenarios?
• comments and traceability: can I note why I made an override?

If the planner needs 4 clicks for an override, after a month of pressure they'll be back in Excel.

6. Realistic implementation time

Question to ask the vendor candidly: "what's the average time from kickoff to steady-state go-live, for a company of my size and complexity?".

If they say "3 months", be suspicious. For an SME with master data to clean, ERP integration, two or three plants, the realistic number is 6-12 months to go-live, 12-18 to steady state. Vendors selling the 3-month dream are the ones who at month 5 will ask you for paid scope changes.

7. Total cost of ownership, not just license

Costs that will hurt over three years:

• license/saas: visible, you know it from day one;
• implementation: 2-4x annual license in year one, depending on complexity;
• integration: often underestimated, especially with custom or legacy ERPs;
• master data maintenance on your side: people, time, data governance;
• continuous training for planner turnover and new products;
• custom developments every APS needs after 18 months when real cases surface.

A "cheap" license APS with painful integration and a small community ends up more expensive than one three times pricier with a solid ecosystem.

The rule of thumb I use

When I help a client in selection, I first write a technical and functional requirements document where each requirement has a weight and a priority (must-have, should-have, nice-to-have). Vendors answer in writing, then we run demos on our case, not theirs.

The vendor who says "let's do a standard demo and talk later" is already telling you how the project will be run.

The final question

It's not "which is the best APS". It's:

"Which APS, at the same price, the planner picks after seriously trying both on a real case from my business?"

That's the answer that counts.