This is a manufacturing inventory system case study about a family business, but the real story is not about software. It is about a father who ran a factory from memory for 25 years, and a son who had to prove that counting things properly was not an insult to that legacy.
The company, a second-generation plastics manufacturer in the Tangerang area with about 60 employees, makes packaging components for consumer goods brands. When I first sat with them, their inventory system was, quite literally, the founder's head plus a warehouse supervisor's handwritten ledger. And here is the uncomfortable part of every manufacturing inventory system case study I have been involved in: that memory-based system had worked. For decades. Until it did not.
The breaking point
Three things converged in 2021. Orders grew past what two people could track mentally. The founder began stepping back for health reasons. And a major customer introduced penalty clauses for late delivery.
The symptoms were textbook:
- Stockouts on fast-moving raw material. Production halted for two days waiting for resin that the ledger said was in stock. It was not. Estimated cost of that one stoppage: around Rp 180 million in delayed shipments and overtime.
- Dead stock nobody could see. A physical count, the first full one in years, found roughly Rp 900 million of raw material and finished goods that had not moved in over a year. Some of it degraded beyond use.
- Purchasing by instinct. The founder ordered material based on feel. His feel was genuinely good, but it did not transfer. When his son placed orders during his father's absence, he either over-ordered out of fear or under-ordered and caused stoppages.
The son, in his early thirties and back from a corporate job, saw the problem clearly. His father saw something different: a son who wanted to replace judgment built over 25 years with a computer.
What they did not do, and why it matters
The son's first instinct was to buy an ERP. He had sat through vendor demos quoting Rp 400 to 700 million for implementation. This is where the project could have died, and where most similar projects do die: automating a process that nobody actually understands yet.
The advice that changed the trajectory was blunt: you cannot automate what you cannot count. Their inventory records were so far from physical reality that any software, no matter how expensive, would have been computerizing fiction. Garbage in, expensive garbage out.
So phase one had no software at all.
Phase one: count accurately for 90 days
For three months, the goal was a boring one, make the records match the shelves:
- One full physical count, done over a long weekend, with the founder personally walking the warehouse. This mattered politically. The dead stock discovery was his discovery, not an accusation.
- Location discipline. Every material got a labeled bin location. No more "it is somewhere in the back."
- A paper-then-spreadsheet movement log. Every item in or out got recorded same-day into a shared Google Sheet by the warehouse supervisor. Simple columns: item, quantity, in or out, reference, who.
- Weekly cycle counts. Every Friday, count 20 items and compare against the sheet. Record accuracy started at 61 percent. By week ten it was above 95 percent.
Cost of phase one: label printing, some shelving, and discipline. Under Rp 15 million.
Phase two: modest software, real rules
Only after the counts held steady did they introduce software, and it was deliberately unglamorous: a mid-tier local inventory application at around Rp 2.5 million per month, connected to barcode scanners at the warehouse entrance. Not a full ERP. The evaluation logic was similar to what I describe in No-Code Tools: Build Business Apps Without a Developer, match the tool to the actual requirement, not the ambition.
The software mattered less than the two rules they attached to it:
- Minimum stock levels with named owners. Each of the 40 most critical materials got a reorder point, calculated from real usage data they now had. When stock hit the point, purchasing was triggered by rule, not by feel.
- The founder's judgment, encoded. The son sat with his father and turned instinct into numbers: which suppliers slip on delivery during rainy season, which customer's forecasts run 20 percent optimistic. That knowledge went into the reorder points and safety stock. His experience was not replaced. It was written down so it could outlive him.
That reframing, from "the computer replaces you" to "the computer remembers what you know," is what finally won the founder over.
The results after a year
Twelve months from the first physical count:
| Metric | Before | After |
|---|---|---|
| Inventory record accuracy | ~61% | 96-98% |
| Dead stock value | ~Rp 900M | ~Rp 420M |
| Production stoppages from stockouts | 5-6 per year | 1 |
| Time to answer "do we have X?" | Hours, sometimes wrong | Under a minute |
Dead stock cut by roughly half, partly sold at discount, partly written off honestly, mostly prevented from recurring. The freed working capital funded a new molding machine without a loan.
What this case actually teaches
Three lessons I would underline for any owner reading this:
Count before you compute. The 90 days of manual discipline were the project. The software was a formality afterward. Companies that skip this step buy ERPs that everyone works around within six months, and end up with the kind of invisible mess I describe in Technical Debt Explained: Why Your App Gets Slower to Fix, except in their operations instead of their code.
Succession is a systems problem. The founder's knowledge was a real asset that existed in exactly one place. The project's biggest return was not the Rp 480 million in recovered dead stock. It was making the business runnable by someone other than one man.
The champion needs the elder's blessing, not his defeat. The son won by making his father the co-author of the new system, not its victim. Any second-generation reader planning a similar fight: structure it so the founder discovers the problem himself.
The practical takeaway
If your stock control lives in someone's memory, start with a physical count, location labels, and a movement log, and hold that discipline for 90 days before spending anything serious on software. This manufacturing inventory system case study ended well because the sequence was right: accuracy first, rules second, software last.
If you are the second generation trying to make this argument at your own family's dinner table, and you want an experienced technical partner in your corner rather than a software vendor with a quota, that is exactly the kind of engagement I take on. Start at the partnership page.