Choosing an APS for Business Central: A Checklist
If you’ve read the companion page on how Microsoft Dynamics 365 Business Central schedules by default, Business Central Is Infinite-Loading, you already know why shops start looking for a finite-scheduling add-on in the first place. What follows isn’t a feature matrix or a ranking — it’s the set of questions worth asking any advanced planning and scheduling (APS) product before you commit to one, ours included. Vendor demos are built to look good; these questions are built to find out what’s actually underneath.
None of the questions below are about picking a “winner.” They’re the questions worth writing down before a single demo, so a strong sales pitch doesn’t quietly answer them for you.
1. Is the engine exact, or is it heuristics?
This is the single most consequential technical question, and it’s also the easiest one for a demo to gloss over. An exact solver — typically built on constraint programming or mixed-integer programming — can, for a given problem, either produce a schedule it can prove is optimal, or tell you honestly how far its answer might be from optimal. A heuristic engine — rule-based sequencing, genetic algorithms, simulated annealing — produces a schedule quickly and it’s often a good one, but it generally can’t tell you how close to the best possible answer it actually is; “good” and “provably good” are different claims.
Neither approach is dishonest on its own — heuristics are a legitimate, common engineering choice, especially at very large scale where an exact solve can take too long. What’s worth pressing on is the vendor’s own language: “advanced heuristics” and “AI-optimized” are marketing phrases that say nothing about whether the underlying method can bound its own answer. Ask directly which category the engine falls into, and ask for a specific example of a schedule the tool produced alongside a statement of how far from optimal (or how provably close to it) that schedule was known to be.
2. Can you see it, and override it, and trust what happens next?
A schedule nobody can look at isn’t operationally useful, no matter how good the math behind it is. Ask whether the product shows the plan on something a planner can actually read at a glance — a Gantt-style board, ideally — and, critically, what happens when a human overrides it. Some products treat the generated plan as final unless you re-run the whole optimization; others let you drag an operation to a different time or resource and tell you, immediately, whether that move still holds up against every other constraint in the plan, or whether it just created a new conflict somewhere else. The second behavior is a meaningfully harder engineering problem than the first, and it’s the difference between a tool a planner actually uses daily and one that gets run once a week and then ignored.
It’s worth testing this one directly rather than taking a vendor’s word for it: during a demo, ask to drag a real operation to a time slot you already know is double-booked, and watch what happens. A tool with live feedback tells you immediately, in the same view, that the move creates a conflict — no separate report to go check, no “run the optimizer again to find out.” A tool without it lets the drag succeed silently, and the conflict only surfaces later, if it surfaces at all.
3. Is it built on your BC data, or does it live beside it?
Ask where the scheduling data actually lives. A product embedded in Business Central’s own data model reads your real production orders, routing lines, and work centers directly, and writes proposed dates back onto those same records — there’s one source of truth. A product that runs as a separate system requires your routings, calendars, and orders to be exported, mapped, or synced into a second data model, and every sync is a place where the two systems can drift out of agreement, silently, until someone notices a scheduled date in the external tool doesn’t match what Business Central itself shows. Neither architecture is automatically disqualifying, but the ongoing cost of keeping two systems in sync is a real, recurring cost that’s easy to underweight during a demo and expensive to discover after go-live.
4. Does it actually read your shop, or does it need you to describe it twice?
A finite scheduler is only as good as its model of your capacity — and if it can’t read your shop calendars, work center efficiency percentages, and machine counts directly from Business Central, someone has to enter that information a second time, in a second place, and keep it current forever. Ask specifically whether the tool reads shift calendars, efficiency, and parallel-machine counts automatically from your existing Business Central setup, or whether it requires a separate capacity model maintained by hand. The second answer isn’t fatal, but it’s a standing maintenance cost that compounds every time a shift pattern or a machine count changes in real life and someone has to remember to update it in two places.
5. What does it not do yet — and will the vendor say so plainly?
Every finite-scheduling product, including ours, has a roadmap: capabilities it handles today and capabilities it’s still building toward — automatic alternate-machine selection, labor as a shared constraint alongside machines, material-availability awareness, splitting oversized batches across machines or days, what-if scenario comparison. None of that is disqualifying on its own; what’s worth testing is whether the vendor tells you clearly which bucket a given capability falls into, or whether every question gets an enthusiastic yes regardless of what’s actually shipped and working today versus what’s still on a roadmap slide. A vendor willing to draw that line clearly, in writing, is a vendor whose other claims are more likely to hold up too.
6. What does it actually cost to turn on?
Cost isn’t just the license fee. Some APS products are architected around a lengthy, consultant-led configuration project before you see any value — weeks or months of modeling your shop before the tool produces its first usable schedule. Others are designed to produce a workable plan with close to zero configuration on day one, refined afterward as you tune it to your specific priorities. Ask what the realistic timeline is from purchase to a planner actually using the tool for a real schedule, and ask what portion of that timeline requires the vendor’s own consulting time versus something your own team can do.
7. Ask for a captured run, not a slide
The most useful thing you can ask any vendor for is a real, captured schedule run against data that resembles your own shop — not a rehearsed demo scenario built to show the product’s best angle, and not a slide claiming a capability exists. A captured run shows you the actual before-and-after: what the shop’s calendars and orders looked like, what the engine proposed, and what changed. If a vendor is confident in what their tool does, this is an easy, low-cost thing to produce. If a specific capability can only be described rather than shown, that’s worth noting as a data point in itself, not a disqualifier — some things genuinely are hard to demo cleanly — but it’s still worth asking why.
Running the evaluation on your own data
The single most useful thing you can do to turn these seven questions into a real answer, rather than a set of talking points, is to stop evaluating on the vendor’s demo data. Pick two or three real production orders from your own shop — ideally including one that’s genuinely awkward, with a changeover-heavy sequence or a resource you know is tight — and ask each vendor to schedule them against your actual routings and calendars. A tool that looks impressive against a clean, curated demo scenario and struggles against your own messy, real data has told you something important that no slide deck would have.
It’s also worth asking what happens on day one after go-live, not just what the finished, tuned product looks like in a mature deployment. A shop evaluating an APS is rarely evaluating the tool in isolation — they’re evaluating how much of their own team’s time, and how many weeks of calendar time, stand between signing a contract and a planner trusting the tool’s output enough to actually work from it instead of falling back to a whiteboard or a spreadsheet.
Where SmartFlow stands against its own checklist
In the interest of holding ourselves to the same questions: SmartFlow APS schedules with an exact constraint solver, not a heuristic — its schedules are computed against provable capacity limits, not approximated. It’s embedded directly in Business Central’s own production order and routing data, with no second system to keep in sync, and it reads your shop’s real calendars, work centers, and efficiency settings automatically rather than requiring a duplicate capacity model. The plan renders on an interactive drag-and-drop board built for BC users to review and adjust before anything commits back to the order. Like every product in this space, it has a roadmap of capabilities still being built — and per this site’s own honesty standard, what’s shipped and what’s still ahead is stated plainly rather than blurred, on every page it appears.
For the concept-level foundation behind these questions — what finite scheduling actually guarantees, and what it costs to get it — see What Is an APS (Advanced Planning & Scheduling) System? and Finite vs Infinite Capacity Scheduling, Explained Properly.