← All posts

Choosing Production Scheduling Software for Small Factories: 7 Questions That Separate Real Tools from Demos

Over the past six months I've sat in on many software evaluations with small and mid-sized factory owners and PMC managers. The pattern is almost always the same: three vendors in one month, each with a polished demo, each claiming to be "lightweight, easy to use, live in days." Then the salespeople leave, the owner looks at three quotes, and still has no idea which one to pick.

The problem isn't the products. It's the selection method. Feature comparison sheets are nearly useless at this price point, because every vendor's feature list looks the same. What actually separates them are the questions that get skipped during the demo and always surface three months after go-live.

Below are the seven questions I recommend asking before signing anything. They don't point at any specific vendor — only at decision criteria. You can take them to any supplier, including us. They are especially relevant for high-mix, low-volume discrete manufacturers with 3+ BOM levels — the factories that need PMC scheduling most, and are most often burned by "general-purpose production planning software."

Data ownership Q1 3-year TCO Q2 BOM depth Q3 Finite capacity Q4 Kit-to-stock loop Q5 Implementation Q6 Data portability Q7

Question 1: Where does the data actually live?

How to judge: ask where the data is stored, which functions work when the network is down, what format the data can be exported in if you terminate the contract, and whether export costs extra. Vague answers are a red flag.

This is the most overlooked and most consequential question. Mainstream SaaS solutions store data on the vendor's cloud servers. On-premise APS or private-deployment solutions store it on your own disk. Neither is universally better, but for small factories there are concrete differences worth asking about:

If you've been searching for "on-premise APS" or a "private deployment scheduling system," you're really asking the same thing: can scheduling decisions stay inside the factory?

Question 2: What is the real 3-year total cost of ownership (TCO)?

How to judge: ask for a 3-year TCO table listing every possible charge — additional users, additional documents, integrations, data export. Do not accept "we'll figure it out later."

Many factories compare only the year-one quote and discover in year three that total spend is far above expectation. SaaS bills typically include: base subscription, per-user charges, implementation, training, integration fees, and overage charges for storage or document volume. A buyout scheduling system typically includes: software license, implementation, and an optional annual maintenance fee.

Put two vendors side by side in a 3-year table and the answer often reverses what the year-one quotes suggest. For factories with annual revenue in the tens of millions of RMB and a few dozen to a few hundred employees, buyout pricing usually has a clear cost advantage over a 3–5 year horizon; SaaS wins on low initial outlay and zero server maintenance.

Question 3: How deep is your BOM? Does kit checking read stock, or recurse through levels?

How to judge: take a product with 3+ BOM levels and ask the vendor to run a kit check live. Look for three things — recursive expansion through all levels, sub-assembly stock consumed first, and a shortage list that drills down to the specific material. If any one is missing, it is not built for discrete manufacturing.

This is the single question that separates a "lightweight tool" from a PMC-grade scheduling tool. Many lightweight solutions only support single-level BOMs: the product hangs directly off its materials. But in discrete manufacturing, a finished product often has 3–5 levels: finished good → sub-assembly → sub-sub-assembly → purchased part.

A usable material kit check for scheduling needs to:

A 5-level BOM kit analysis takes 20 minutes by hand in Excel. A proper tool should compute it in under a second. During the demo, hand them your most complex product and ask them to compute it on the spot — not their prepared sample.

Question 4: Is scheduling a date guess, or a finite-capacity calculation?

How to judge: ask the vendor to compute "how many hours for 100 units" using your product parameters and your line configuration, then compare it to your veteran planner's estimate. A deviation above 20% means the scheduling logic is not trustworthy.

"Scheduling" is used loosely. Some tools' idea of scheduling is letting you type a start date and an end date, with no hour calculation at all. Real finite-capacity scheduling has to compute how many hours a work order occupies a given line, based on the product's characteristics and the line's configuration.

The algorithms differ entirely by production mode: fixed takt (e.g. SMT, seconds per unit), flexible takt (e.g. manual assembly, minutes per unit ÷ stations), single-station (e.g. injection molding, cycle minutes ÷ machines), batch, continuous flow, and project-based. If a tool only supports one mode, it falls apart the moment you change industries. This is also why so many APS scheduling software products demo well and schedule poorly after go-live.

For factories that routinely deal with rush orders disrupting the plan, one extra question is worth asking: after an insert, can the system automatically re-schedule? And when it does, does it avoid moving work orders already in progress? If the answer to either is weak, PMC will be back in Excel by month two.

Question 5: Does material picking close the loop with work orders, stock, and signatures?

How to judge: ask the vendor to walk through the full flow from "create a work order" to "material pick deducts stock." The more manual re-entry steps in the middle, the higher the adoption friction.

Many factories migrate off Excel-based scheduling only to find scheduling and material picking are still two separate tracks: a work order exists in the system, but the pick list is still paper, and the warehouse manually keys in stock afterward. The result: inventory never reconciles, and loss rates can't be calculated.

A closed-loop pick process should include at minimum: work order → multi-level BOM pick list → print with QR code → warehouse scans → fills actual issued quantity → electronic signature → automatic stock deduction → signature archived for traceability. Miss any step and you fall back to manual entry.

Question 6: Is implementation a remote help desk, or someone who actually gets your Excel running?

How to judge: ask whether they offer a data-initialization service, whether they can handle your existing Excel formats, and whether they have import cases from your industry. Vendors offering only remote docs and videos carry far higher implementation risk.

Digitalization failures at small factories are rarely about missing features. They are almost always about data initialization. Asking a factory to enter 150 products, several hundred materials, and dozens of BOM rows from scratch is not realistic.

So ask: Can you import the Excel data I already have? Can we get my single most painful scenario working first? The answer to that question predicts go-live success better than any feature list. In one kitchen-appliance ODM case, we used a standalone conversion tool to turn two Excel files accumulated over years into a working scheduling system on the same day, without re-entering a single record from scratch.

Question 7: If I want to switch systems tomorrow, can I take all my data?

How to judge: ask about export format, whether all business data is included, whether it costs extra, and whether fields are restricted. A vendor willing to put "complete data export" in the contract is usually more trustworthy.

It sounds inauspicious, but this question is the litmus test of vendor confidence. Some SaaS solutions restrict data export: partial fields only, extra fees, or formats that are hard to migrate. This is data lock-in, and it puts the factory in a weak position at renewal.

A healthy standard is: the system should always let you leave, and you stay because it works. On-premise deployment plus open export formats (JSON / Excel) satisfies this by design.

Appendix 1: On-premise vs SaaS — typical differences

The table below is not meant to prove one is better. It surfaces the differences that are invisible during a demo and only appear after go-live. Walk through it row by row with each vendor.

Dimension On-premise + buyout SaaS subscription
Data locationFactory-owned disk / LAN hostVendor cloud servers
Works offlineCan schedule, check BOMs, report work on LANUsually unavailable
3-year TCO (50-user scale)One-time license + implementation, no mandatory annual feeSubscription × 3 years + potential per-user or volume fees
Multi-level BOM kit depthVendor-dependent; recursive logic usually customizableMost lightweight tools support only one or two levels
Data exportFull JSON / Excel export, no field restrictionsSome vendors restrict fields or charge per export
Maintenance costOne always-on PC as host; backups manual or semi-automatedVendor-managed; zero maintenance for the user

Appendix 2: A printable selection scorecard

Turn the seven questions into a scorecard and score each vendor before comparing. Rate each item 1–5, with 5 being best. Don't just look at the total — focus on Q3 and Q4. Those are the core capabilities of a PMC tool, and a low score there means the solution should be rejected even if it's cheap.

Criteria Vendor A Vendor B Vendor C
1. Data ownership (location, offline access, export freedom)
2. 3-year TCO (including all hidden fees)
3. Multi-level BOM and recursive kit checking
4. Scheduling based on finite-capacity hour calculation
5. Closed loop across pick list, stock, and signature
6. Data initialization and implementation service
7. Complete data portability
Total (max 35)

One honest closing note: no software fits every factory. The goal of selection isn't the most features — it's the best match for your BOM depth, production mode, and data sensitivity. Once you've asked all seven questions, the answer is usually obvious.

This framework is best suited to high-mix, low-volume discrete manufacturers with 3+ BOM levels. If you're a 15-person pure-assembly shop with a single-level BOM, a lightweight SaaS or even Excel may be the smarter choice. Selection isn't about "heavier is better" — it's about fit.

Frequently Asked Questions

Q: Can an on-premise scheduling system run without dedicated IT staff?
No dedicated IT is required. Day-to-day use is just opening a LAN address in a browser, and backups are a one-click export of the database file. In host mode, one PC stores the master data and other PCs sync as clients — after setup, you basically never touch a server.
Q: Does a one-time purchase mean upgrades cost extra later?
Typically, a buyout license covers the current major version, with minor fixes free. Whether major-version upgrades are charged depends on the vendor — ask before signing and put it in the contract. That detail matters more than the word "buyout" itself.
Q: Does multi-level BOM kit checking require high-end hardware?
A 5-level BOM kit analysis across 5,000 materials typically completes within 1 second on an ordinary office PC. We have published a full performance test with reproducible data — worth reading alongside this article.

Want to ask these 7 questions to PMClite in person?

On-premise · One-time purchase · Multi-level BOM recursive kit check · Live in 1–3 days · 30-day full-feature trial

Book a demo focused on your BOM depth →

This article discusses selection methodology and does not target any specific vendor. Deployment models, cost structures, and BOM levels described are common industry patterns; actual vendor offerings may differ. PMClite is an independent product and has no commercial relationship with any unnamed vendor mentioned in this article.