Every morning at 4 AM, the head of production at an F&B franchise in Greater Jakarta decided how much food to cook for fourteen outlets. His inputs were yesterday's leftover reports, phone calls from outlet supervisors, and eleven years of instinct. Some days he was brilliant. Other days three outlets sold out of the bestseller by 2 PM while two others threw away trays of it at closing.
This central kitchen system case study is about replacing that guessing game with data the business already had. Not with AI, not with a warehouse robot, but with a straightforward system that moved outlet sales numbers to the kitchen automatically and turned them into a production plan.
I am sharing it because the pattern applies to almost any franchise or multi-outlet F&B business in Indonesia: the information you need already exists at the point of sale, it just never travels to the place where decisions are made.
The starting point: phone calls and instinct
The franchise ran a classic central kitchen model. One production facility prepared semi-finished goods overnight: sauces, proteins, dough, portioned components. A delivery run distributed everything to outlets between 6 and 9 AM. Outlets did final preparation and sold through the day.
The daily production decision worked like this:
- Around 8 PM, outlet supervisors sent leftover counts by WhatsApp, when they remembered
- Around 9 PM, the production head compiled whatever arrived into a paper worksheet
- At 4 AM, he set quantities, adjusting by feel for weekends, paydays, rain, and nearby events
The costs were visible everywhere once you looked:
- Waste ran at roughly 11 percent of production cost. For a kitchen spending around Rp 900 million a month on ingredients, that was close to Rp 100 million cooked and discarded.
- Stockouts hit the top three items several times a week. An outlet that sells out at 2 PM loses its best hours and trains customers to stop coming.
- Everything depended on one man. When the production head took leave, waste and stockouts both spiked within days.
What we actually built
The tempting move was a full ERP. The owner had already collected two quotations, both above Rp 800 million, both promising to digitize everything from procurement to payroll. We did something much smaller.
Every outlet already had a POS system recording each transaction. The data was sitting there, used only for end-of-day cash reconciliation. The build had three parts:
- A nightly sync pulled item-level sales from every outlet's POS into one database. No new hardware, no retraining of cashiers.
- A demand model with modest ambitions. For each outlet and item, it looked at the same weekday over recent weeks, adjusted for trend, and flagged known calendar events like paydays and public holidays. Deliberately simple, so the team could understand and challenge it.
- A production sheet, generated at 8 PM. Recommended quantities per item per outlet, aggregated into totals for the kitchen, with the reasoning visible: last four Tuesdays this outlet sold 62, 58, 65, 61 of this item, recommendation 64.
The critical design decision: the system recommended, the production head decided. He could override any number, and the system logged both figures. This did two things. It kept his eleven years of knowledge in the loop, and it built an honest record of when instinct beat the data and when it did not.
Total build was under Rp 200 million including the pilot period, roughly a quarter of the cheapest ERP quote.
What happened in the first six months
The first month was rough in an instructive way. The production head overrode about 70 percent of recommendations, and the override log showed his numbers and the system's numbers were usually within 10 percent of each other. He was doing manually, from memory, what the system did from data. That earned trust faster than any presentation could have.
By month three, overrides had dropped to around 20 percent, concentrated exactly where they should be: days with local events, weather disruptions, and new menu items with no sales history. The system handled the routine, the human handled the exceptions.
The measurable results after six months:
| Metric | Before | After 6 months |
|---|---|---|
| Waste (share of production cost) | ~11% | ~6% |
| Stockouts of top 3 items | Several per week | Rare, mostly event days |
| Time to compile production plan | ~2 hours nightly | ~20 minutes review |
| Dependence on one person | Total | Any supervisor can run the sheet |
The waste reduction alone was worth around Rp 45 million a month, which means the system paid for itself in under five months. The owner's favorite outcome was different, though: for the first time, he could see per-outlet, per-item performance in one place, which reshaped decisions about menu pruning and where to open the next outlet. That visibility is the same argument I make in build an owner dashboard you will actually look at.
The lessons that transfer to your business
Your POS is an underused asset. Most multi-outlet businesses use point-of-sale data for cash control and nothing else. The demand signal your operations team needs is already being recorded every day.
Recommend, do not dictate. Systems that overrule experienced operators get sabotaged or ignored. Systems that make recommendations transparent, and log overrides without blame, get adopted and then improve.
Buy the smallest thing that removes the biggest pain. The Rp 800 million ERP would have delivered this feature eventually, in module seven, after a year of implementation. The focused build delivered it in ten weeks. You can always expand later, and by then you will know exactly what to expand toward. If you are evaluating vendors for something like this, my software vendor red flags checklist covers the warning signs.
Consistency is a brand asset. The waste savings justified the project on a spreadsheet. But the quieter win was that customers stopped encountering sold-out favorites, and franchisees stopped fielding complaints about it. For a franchise, predictability at every outlet is the product.
The takeaway
If you run a central kitchen feeding multiple outlets, and production quantities are still decided from memory and WhatsApp messages, you are paying a daily tax in waste and stockouts that a modest system can eliminate. Start by asking one question: does our outlet sales data reach the person planning production, automatically, every night? If the answer is no, that single pipe is likely the highest-return technology investment available to you right now, and it does not require an ERP to build.