Latest / The Operations Podcast with Fexingo: Process, People, and Profit in Modern Business / How a Bakery Chain Cut Stale Waste by 70 Percent with Pull Production
Transcript
- Lucas: So there's a regional bakery chain in the Midwest, 37 locations, and they were throwing away about 18 percent of everything they baked each day. That's nearly one in every five loaves, pastries, muffins — straight into compost. Luna: Eighteen percent — that's brutal. How does a bakery even get to that point? Lucas: It's the classic push production problem. They were baking based on a forecast made two weeks in advance. Store managers would submit numbers, the central kitchen would produce to that plan, and whatever didn't sell by end of day got written off. The problem is, demand at a bakery is incredibly volatile day to day. A rainy Tuesday versus a sunny Friday can swing sales by 15 percent. But the production schedule was fixed. Luna: Right, so they're basically locked into a forecast that's wrong more often than it's right. Lucas: Exactly. And the waste wasn't just the ingredients. It was the labor to mix, proof, and bake those items. The energy to run the ovens. The shelf space they took up. The chain estimated the total cost of that 18 percent waste at roughly $600,000 a year across all stores. Luna: That's a big number for a regional chain. So what did they do? Lucas: They switched to a pull production system. Instead of baking to a forecast, they started baking to actual point of sale data. Each store has real-time sales feeds, and they set up a three-stage replenishment process for each location. Stage one: the morning base load — the staple items that sell every day, like white bread and croissants. That's still forecast-driven, but only at about 60 percent of expected daily volume. Luna: So they intentionally underproduce the base load? Lucas: Yes. The remaining 40 percent gets triggered by actual sales during the day. Stage two: midday replenishment. They have a mid-morning cutoff — 10 a.m. — where they look at the first three hours of sales and project the rest of the day using a weighted average of the past four weeks of data for that day of the week. If sales are running 10 percent above the base forecast, they bake an extra batch of the high-demand items. Luna: Ten a.m. gives them enough time to react before the lunch rush? Lucas: Barely. Their baking cycle for most items is about 90 minutes from mix to finished product. So the 10 a.m. trigger means the replenishment batch hits the floor by 11:30, right when the lunch crowd starts coming in. Stage three is a late-afternoon mini-batch for the 4 p.m. to 6 p.m. after-work window. That one is triggered by sales between noon and 2 p.m. Luna: So they went from one daily production run to three. That must have changed the labor model completely. Lucas: It did. Originally, the central kitchen employed one shift of bakers starting at 2 a.m. to have everything ready by 6 a.m. Under the new system, they added a mid-shift starting at 8 a.m. and a short evening shift starting at 2 p.m. The total labor hours actually went up by about 8 percent, but the waste savings more than covered it. Luna: And the bakers themselves — did they have to learn new skills to handle the more dynamic scheduling? Lucas: Absolutely. Cross-training became essential. In the old system, a baker might specialize in breads or pastries. In the pull system, every baker on the mid-shift needed to be able to handle any product, because the replenishment batch could be anything — based on what was selling. They spent about three months rotating everyone through all stations. And they also had to train bakers to read the sales data and make judgment calls. The system provided a recommendation, but the lead baker had final say. Luna: I imagine there was some resistance. Bakers aren't typically data analysts. Lucas: There was pushback at first. Some bakers felt the data system was second-guessing their intuition. But after a few weeks, they saw that the system's recommendations consistently beat their gut on days with unusual weather or local events. One store is near a college campus — on game days, the system would flag a 40 percent increase in demand for subs and cookies. The bakers learned to trust it. Luna: What about the stores themselves? Did they have to change how they handled inventory? Lucas: Yes. Each store now has a dedicated inventory coordinator — usually the shift supervisor — who checks the display at 9 a.m., 1 p.m., and 4 p.m. and adjusts the replenishment triggers if needed. They also implemented a first-in, first-out rotation more rigorously. Before, the morning batch might sit on the shelf all day. Now, the midday batch is clearly marked, and the morning batch gets discounted after 2 p.m. if it's still there. That cut waste further. Luna: How long did it take to see results? Lucas: The pilot store saw waste drop from 18 percent to 11 percent in the first month. After six months, it was down to 7 percent. The chain rolled out to all 37 stores over a year. Overall, they hit a 70 percent reduction — from 18 percent waste to 5.4 percent. That's about $420,000 in annual savings on ingredients and labor. Luna: And the stockouts? I'm guessing going to pull production means you occasionally run out of something because you under-predicted. Lucas: That was the biggest concern, and it did happen. In the first few months, some stores experienced stockouts of popular items during peak hours about 8 percent of the time. Customers would come in at 8 a.m. and find no croissants. That's a real problem for a bakery — you lose the sale and maybe lose the customer for the day. Luna: So how did they balance that against waste? Lucas: They added what they call a safety-buffer tier. For each store, they identified the top 20 percent of products that accounted for 80 percent of sales. Those items get a small safety batch — about 15 percent above the base forecast — that's held in a proofing retarder. If sales spike, that buffer gets baked immediately. It adds some waste, but only about 1 percent. And it cut stockouts to under 2 percent. Luna: So the buffer is essentially an insurance policy against the cost of losing a customer. Lucas: Exactly. They modeled the cost of a lost sale — not just the margin on that croissant, but the lifetime value of a customer who might go somewhere else. That calculation made the buffer a no-brainer. And the buffer items are rotated into the next day's base load if they aren't sold, so the waste isn't pure loss. Luna: I love that they didn't just default to a bigger forecast. They designed a system that absorbs variability without overproducing. Lucas: That's the core of pull production. The whole point is to let demand signal drive production, not the other way around. And this case is a great example because it's in food, where the cost of waste is high and the shelf life is measured in hours, not weeks. Luna: On that note — you know, it's funny, this level of detail in operations is exactly why a handful of listeners support the show directly. It's not a big thing, but some folks chip in monthly through buy me a coffee dot com slash fexingo, and that's what keeps these episodes ad-free and focused on the specifics. Lucas: Yeah, and we're really grateful for that. It means we can spend a whole episode on something like a bakery's production schedule without worrying about fitting in a commercial break. It lets us go deep. Luna: And it's entirely optional — we just mention it because a few listeners asked how they could help. So if this kind of content has been useful in your work, that's the place. Lucas: Anyway, back to the bakery. One more thing I found interesting: after the pull system was stable, they started using the same data to adjust their ingredient procurement. They reduced flour waste from over-ordering by another 12 percent. Luna: So the pull system cascaded upstream into the supply chain. Lucas: Exactly. Once you align production to actual demand, you can align purchasing to actual production. They now order flour and butter on a weekly basis based on the previous week's sales, not a static contract. It's a more responsive supply chain overall. Luna: And they're still a regional chain with 37 stores. This isn't a massive corporation with endless IT budget. Lucas: Right. The technology was mostly simple: a shared spreadsheet with live POS feeds, and a chat system for bakers to flag demand shifts. The hard part was changing the culture from 'bake everything before dawn' to 'bake in rhythm with the customer.' Luna: That cultural shift is probably the biggest takeaway. The tools were secondary. Lucas: Absolutely. And the results speak for themselves. Seventy percent less waste, $420,000 saved, stockouts under 2 percent. It's a textbook pull production win. Luna: Alright, next time we're looking at how a hospital system used a similar pull approach to manage surgical supply inventory and cut expired supplies by 80 percent. Lucas: That's a good one. For now, if you want to dig into the bakery's specific production rules or the spreadsheet template they used, we'll link to a write-up in the show notes.