How to Forecast Food & Beverage Sales and Stop Losing Revenue

Say you’ve got a new wholesale client starting shortly. You’ve built 200 extra cases a week into the forecast. Yet, production is running a standard week.

The rest writes itself:

The account’s first order arrives → Production is short → The relationship that took three months to close starts on the wrong foot.

Working with SMB food and beverage manufacturers, we at FlexiBake often see sales and production working with different forecasts. As a result, the team ends up with a short delivery, a margin that doesn’t add up, or a pallet coming back on the truck.

So what’s inside? Reasons why most food sales forecasts fall short, and a five-step framework for building one that actually connects to production.

TL;DR

• Most food manufacturers have everything they need to forecast sales accurately. The forecast is still wrong because none of it is connected.
• A sales forecast that doesn’t reach production is just a number in a spreadsheet.
• Four things that break sales forecasts: returns not in the total number, pricing and volume disconnected, customer-level patterns buried in the total, and the forecast not connecting to anything downstream.
• The fix is a five-step cycle: capture the demand signal, validate against inventory, trigger purchasing, translate it into a production plan, and feed variance back in.
• Each step makes the next forecast more accurate.

You’re Sitting on More Forecasting Data Than You Think

The data needed for an accurate forecast already exists in your business. Most manufacturers just haven’t connected it yet. And it’s not a food manufacturing problem only. A survey of 317 demand forecasting professionals across manufacturing, retail, and consumer goods found the same thing: data sitting in different places, teams working from different numbers, and no one owning the process that ties it together.

Here’s where that data lives in your business, what signals it’s sending, and what it costs when they’re ignored.

Standing Orders and Sales History: Months of Data, Mostly Underused

At least two or three years of delivery history sits in most manufacturers’ order systems right now. The café that takes 40 sourdough loaves every Tuesday. The grocery account that runs the same case of chipotle sauce every Friday. Every November spike, every slow August. It’s all there, documented, timestamped, waiting.

The problem is, most of it never feeds a forecast. Someone opens last year’s file, bumps it up 10%, and leaves it at that. The rest gets filled in by memory, gut feel, and a few red cells that need checking with someone who’s out that week.

Last year’s numbers. Plus 10%. Ask Dave. Send to production. Repeat.

When all sales history lives in one system, Monday morning looks different. Take La Provence Bakery, a wholesale bakery in California. Charles Dardaine, their Plant Manager, says that once daily and standing orders were inside FlexiBake, the ordering process that used to require manual work “ran itself.”

Sales history only tells half the story, though. Here’s another neglected factor that could improve your forecast.

Returns and Slow-Movers: The Signals You’re Not Reading

Your sales history shows 1,200 units of lemon herb dressing sold last month. But 180 came back. Building next month’s forecast from 1,200 means starting 17% too high.

The same problem shows up differently. For example, an SKU sitting at the same inventory level for three months despite a promotional push is telling you to produce less of it, and probably has been for weeks.

Most manufacturers do track returns. Somewhere. A credit note here, a driver bringing product back on the route there. But almost none of it feeds the next forecast. It comes off the P&L, and next month opens with the same number.

Sales alongside what came back by customer, by product, as a percentage (Revenue vs Returns report, FlexiBake). The number that should be adjusted in every forecast.

Getting the forecast wrong costs money. How much and how fast depend on what you make.

Fresh or Long Shelf Life: The Cost of Getting It Wrong

The cost of a bad forecast shows up in the bin or on the pallet, and it shows up fast.

For fresh manufacturers (bread, pastries, meals, and deli), an overestimate goes in the bin the next morning. Stales are a physical, immediate record of how wrong the forecast was.

For long shelf life manufacturers (frozen foods, snacks, confectionery, packaged goods), the consequence is slower, but the exposure is larger. An overestimate results in a pallet of Ginger Kombucha sitting at the same inventory level throughout an entire seasonal window. Yes, it won’t spoil fast. But it becomes redundant if it doesn’t move before the next promotional cycle. You’ve paid to make it, you’re paying to store it, and it’s not moving.

The data to prevent both outcomes is already in the business. So why does a forecast still go wrong?

Why Most Food Sales Forecasts Fall Short

A forecast that looks convincing can still be wrong in four specific ways. Most are, at least.

1) Returns aren’t in the number. You’re forecasting from a number that’s higher than what you actually sold, and next month’s forecast starts from that same inflated number.

2) Pricing and volume aren’t connected. You forecast 500 cases. Then a holiday promotion changes the unit economics on the three top SKUs. The volume number stays the same, but the margin for those 500 cases is different.

3) Customer-level patterns are buried in the total. The total looks fine. But you’re up 20% with Customer A and losing Customer B. By the time Customer B’s decline shows up, you’ve already overproduced.

4) The forecast doesn’t connect to anything. Production hasn’t seen it. Purchasing hasn’t seen it. The price change that went through last Tuesday hasn’t been updated. You have a number, but nobody is responsible for fulfilling it.

The root cause is the same for all four: everything lives in a different place, and nobody’s connecting it. The fix is a closed loop: demand signal, inventory check, purchasing, production plan, variance feedback, back to the start.

That loop is what the next section is about.

From Demand Signal to Production Plan: The FlexiBake 5-Step Forecasting Framework

The fix for poor sales forecasts is a connected cycle: five steps where each one answers a question the previous section raised. Together, they take a demand signal from a number on a screen to a production schedule, a purchase order, and a feedback loop that makes the next cycle more accurate than the last.

Any manufacturer can apply this logic. I’ll show how it’s done through the example of FlexiBake.

Step #1: Capture Your Sales Signal

→ Your sales signal is the demand data showing what your customers will order. This is where you capture it before the production team needs a number.

How much will your customers order and when? The answer looks different depending on how far ahead you’re planning — this week’s delivery schedule, or next quarter’s revenue target.

Fresh and route-based manufacturers are mostly planning days ahead: daily and weekly order patterns from standing-order and route customers. Long shelf life manufacturers are planning months ahead: SKU-level trends, seasonal peaks, and promotional effects across a quarter or more. Most businesses need both at different moments.

When You’re Planning Days Ahead

A route-based bakery running 15 accounts needs to know what to load on the truck on Thursday. Most of those accounts follow a predictable pattern — the same café takes the same loaves most weeks, with small variations. In FlexiBake, Forecast Sales Orders pulls that history, applies whatever adjustment makes sense (a new account coming on, a seasonal uplift, a promotion running next week), and generates the actual orders directly into the system.

For fresh manufacturers specifically, there’s one more variable worth building in before the orders go out. Each one has a stale tolerance, the percentage of returned product they can absorb without it hurting the week’s margin. Most just absorb whatever comes back and adjust next time by gut feel. In FlexiBake, you set that target return percentage upfront, and the system back-calculates how much to produce to land there.

For an operation like Mancini’s Bakery in Pittsburgh — 100 years old, fifth generation of bakers — that kind of structured order generation changes how the whole week starts. Nick Mancini says that forecasting has let them ‘bake to par and be prepared for future orders.

When You’re Planning Months Ahead

A sales manager heading into Q4 needs to tell the production team what demand looks like for the next three months before anyone starts scheduling runs. Creating a Sales Forecast in FlexiBake’s Analysis Centre is built for that. Select the forecast period, choose a customer, and FlexiBake pulls last year’s actual sales month by month as the starting point. The salesperson adjusts from there: a new product line, a growing account, a promotion that didn’t exist last year.

The output is a revenue prediction by customer and product that feeds directly into Schedule Production, so the production team is working from the same forward-looking picture as the sales team.

Step #2: Validate Against Your Inventory

→ This is where you find out whether you can actually fulfill what’s on the books before the production starts.

The worst moment to discover a stock gap is when someone has already started mixing.

Before the schedule is locked, run an inventory check against what’s on hand. Do you have enough eggs for tomorrow’s croissant run? Enough packaging for a 10,000-unit run three weeks out?

FlexiBake’s inventory report runs in two passes: once for ingredients, once for packaging. Set a date range, hit refresh, and gaps are visible before they reach the production floor.

Step #3: Trigger Your Purchasing

→ This is where inventory gaps become purchase orders.

Found a gap in Step 2? For instance, butter is short by 40kg. Or the packaging supplier delivered half the order. Now someone needs to order the missing materials before the production window closes. Often, that means a phone call, an email to the supplier, and a manually created PO all happening under time pressure.

In FlexiBake, purchase orders are generated directly from scheduled production by raw material, quantity, and required date. You can review, adjust, and email to the supplier from the same screen.

Step #4: Translate the Forecast Into a Production Plan

→ This is where the sales forecast becomes a production number.

A sales manager has forecast a 20% volume increase on two frozen SKUs for Q4 because a new wholesale account starts in six weeks. Production is scheduled from open orders. The order lands in week six. Production never saw it coming because the forecast never left the sales team’s spreadsheet.

For fresh manufacturers, the window is shorter, but the stakes are just as immediate. A new café account has been confirmed for next week. Production is running standard quantities. The first delivery comes up short.

To schedule production well, the planner needs to see not just what customers have ordered but what the sales team expects. In FlexiBake, the sales forecast created in the Analysis Centre can be pulled into Schedule Production as a dedicated column alongside open order quantities. Here, anticipated demand and confirmed demand are all on the same screen.

For fresh manufacturers, that forecast column covers the gap between a new account being confirmed and the first order arriving. For long shelf life manufacturers, it’s the primary input for runs that need to happen weeks before orders are placed.

Case in point: Café de Medici runs a daily order operation where production scheduling accuracy affects everything that follows. After moving to FlexiBake, Rebecca Ferryman, their Operations and Accounting Coordinator, describes their production schedule as “so much more accurate and efficient”, which, for a business where the daily order is the heartbeat of operations, is exactly the outcome Step 4 is designed to deliver.

Step #5: Feed Variance Back Into the Next Forecast

→ This is where the cycle closes and where every future forecast gets more accurate.

Most forecasting processes end when the product ships. The question of whether the forecast was right (and by how much) rarely gets answered formally. It becomes the next gut-feel adjustment, and the same overestimate or underestimate repeats the following month.

The variance step breaks that pattern. Two questions need answering after every cycle:

How much did we actually sell?
How much came back?

For fresh manufacturers, returns are the critical signal. A spike in stales on a specific product or route tells you the forecast was too aggressive, or a promotional price drove volume that didn’t translate into real sell-through. FlexiBake’s Revenue vs Returns reports (available by customer or by product) show sales revenue alongside returns and the percentage for each. That percentage is the number that adjusts next month’s forecast.

For long shelf life manufacturers, the signal is in the product-level returns data. Is the Ginger Kombucha coming back at a higher rate than last quarter? Are units moving more slowly than the forecast assumed? The Revenue vs Returns Detail report shows sales and return percentage by SKU. A product with softening sales and rising returns is telling you the forecast needs to come down before you’re sitting on a pallet nobody wants.

Finally, when you open the next forecast cycle, last year’s actuals are already populated alongside the new forecast columns. The gap between what was predicted and what actually sold is visible the moment you start planning.

A Sales Forecast That Connects to Production Is a Plan

That’s the cycle. Five steps, one loop, and each rotation makes the next forecast more accurate than the last.

A forecast that triggers a production schedule, updates when prices change, and improves every cycle because variance feeds back in… sounds like a plan, doesn’t it?