How to Create a Sales Forecast You Can Plan Routes Against

To create a sales forecast, pull your order history into weekly buckets, average the comparable weeks to get a baseline, adjust for seasonality and anything you know about the period ahead, then convert the result into the units your operation actually runs on. For a business that delivers its own orders, that last step is the one that matters: a forecast that stops at a revenue figure cannot tell you how many drivers to roster.

The whole exercise takes about two hours the first time and twenty minutes a week after that. You need order history, a spreadsheet and a decision about how far ahead you are forecasting. You do not need software, a data analyst, or a finished view of what your method should be. The method emerges from doing this once and looking at where it was wrong.

This is the build. Why the number is worth having at all is covered in sales forecasting for a delivery operation, and the trade-offs between the different approaches are in the comparison of forecasting methods that work when you deliver your own orders.

The Bottom Line

• Build the forecast in units first and convert to revenue afterwards. Units are what routes, drivers and stock consume.
• Use weekly buckets for the operating forecast. Monthly buckets average away the exact peak you are trying to staff for.
• Forecast peaks from the same week in prior years, never from a recent average. An average built from ordinary weeks always under-calls a holiday.
• Write forecast and actual side by side every week. The teams that track weekly reach 87% accuracy against 52% for those tracking irregularly (Gangly).

Step 1: Pull your order history into weekly buckets

Start with the raw record of what you have already sold, exported from wherever your orders live: your order management system, your delivery platform, your accounting software, or the invoice book.

You want two years if you have them, one year at minimum, with four fields per order: date, customer, product or category, and quantity. Add delivery zone or postcode if you have it, because you will want it in step 5.

Then aggregate into weeks. Use the same week numbering every time and be consistent about which day a week starts, because inconsistent week boundaries are the most common source of nonsense in a first forecast. Weeks beat months here for one reason: you roster drivers by the week, and a monthly total cannot tell you that the volume is concentrated in seven days of it.

If you have less than a year of history, you can still forecast a short horizon from recent trend. You just cannot see seasonality yet, so you will be supplying that from judgement until a full year accumulates.

Step 2: Choose your forecast horizon

Run two horizons rather than one, because two different sets of decisions have two different lead times.

The operating forecast: four to six weeks, in weekly buckets. This is the one that drives driver rosters, route counts and stock purchasing. Six weeks is usually far enough to cover the longest purchasing lead time and near enough that the estimate means something.

The planning forecast: twelve months, in monthly buckets. This one drives hiring, vehicle decisions, and whether you take on a large seasonal account. It does not need weekly precision and will not survive contact with reality beyond about a quarter, which is fine because its job is to size decisions rather than schedule them.

Anything beyond twelve months is planning, not forecasting. Write it if your bank wants it, but do not expect to manage against it.

Step 3: Build the baseline from comparable periods

Now produce the first number: what would happen if nothing changed.

For a business with a full year of history, compare like for like against the same week last year rather than against a recent average. Week 7 is Valentine’s week for a florist and week 47 is Thanksgiving for a caterer whatever happened in the four weeks preceding them.

Then apply your growth rate. Take total orders in the last complete quarter against the same quarter a year earlier, express the difference as a percentage, and apply it to the comparable week. If you did 1,000 cases in week 12 last year and you are running 12% up year over year, your baseline for week 12 is 1,120 cases.

For a business with less history, use a four-week moving average of recent weeks as the baseline instead, weighting the most recent weeks more heavily if you are growing. It will be flat where the real demand is not, which is what step 4 exists to correct.

Step 4: Adjust for seasonality and what you already know

The baseline is arithmetic. This step is where your actual knowledge of the business enters, and it is usually worth more than the arithmetic.

Work through four adjustments in order:

• Known calendar events. Holidays, local events, school terms, the week your largest customer closes for the summer. Forecast these from the same period in prior years, not from the baseline.
• Customer changes. Accounts won, accounts lost, accounts that have visibly slowed. New accounts should be added by hand, because no historical method can see them.
• Your own actions. A promotion running, a new product launching, a price change, a route you are no longer serving.
• Capacity limits. If the forecast exceeds what your kitchen or your vans can physically do, record both numbers. The constrained figure is what you will ship; the unconstrained figure is the only evidence you will ever have of demand you turned away.

Write the reasoning next to each adjustment, not just the revised number. When the forecast misses, the reasoning is what tells you which assumption to fix.

Step 5: Convert the forecast into routes and driver hours

This is the step that most sales forecast guides leave out, and it is the reason to forecast at all if you deliver your own orders.

Take the weekly unit forecast and divide it the way the operation divides: by day, then by zone. A week of 1,120 cases is not a schedule. Eleven hundred cases weighted two-thirds toward Monday and Tuesday and clustered in three postcodes is a schedule.

Then apply your own operating ratios, which you derive from history rather than from any benchmark:

• Orders or cases per route, from your recent completed routes
• Average drops per hour in each zone, since a dense urban zone and a rural round are not comparable
• Stock required per unit sold, including the wastage rate you actually run at

The output is a number of routes, a number of driver hours and a purchase quantity. Those three numbers are the forecast’s real deliverable. A forecast that stops before them has not changed any decision.

Worked example. A wholesale bakery forecasts 1,120 cases for week 12. History says Monday and Tuesday take 60% of the week’s volume and a route carries about 90 cases. That makes roughly 672 cases across two days, or eight routes, against four on the remaining three days, for a total near twelve routes and a requirement for at least two extra driver shifts early in the week. Flour is ordered against 1,120 cases plus the 4% wastage rate the last six months actually produced.

Step 6: Check the forecast against what actually shipped

A forecast that is never compared to the outcome is a writing exercise. Put a column for actual next to every forecast column and fill it in at the end of each week.

Two numbers and one sentence is enough: forecast, actual, and why they differ. Do this weekly and you will have, within a quarter, the single most valuable artefact in the whole exercise: a record of how your own forecasts fail.

Read the pattern rather than any individual week:

• Consistently high: structural optimism, or a target that has crept into the forecast. Subtract it.
• Consistently low: an unrecognised growth trend, or a method that flattens your peaks.
• Erratic both ways: the method does not fit the business and no amount of tuning will fix it.

Keep expectations proportionate. The median sales team lands at 70% to 79% accuracy and only 7% clear 90% (Gangly); a benchmark of landing within 15% is common (A Sales Growth Company). The standard to hold yourself to is not a percentage but an explanation: every miss over about 15% should have a nameable cause.

Step 7: Set the weekly review habit

The first forecast takes two hours. Keeping it alive takes twenty minutes a week, and that twenty minutes is where nearly all the accuracy comes from.

Pick a fixed slot ahead of the week you are rostering. A Thursday for a Monday start is typical. In it you do three things: fill in last week’s actual, roll the horizon forward one week, and adjust the next four to six weeks for anything you have learned. Then send the result to whoever builds routes and whoever buys stock. If they have not seen it, it has not affected anything.

That handover is also the honest test of whether you need a tool. When the file becomes the bottleneck rather than the forecasting, the categories available are set out in the review of sales forecasting software for businesses with delivery routes.

A simple sales forecast template

If you are building the sheet from scratch, five columns per row and one row per week is the whole thing:

Add columns to the right for routes, driver hours and key stock items, each derived from the forecast units using your own ratios. Keep one tab per year so the comparable-week lookups stay easy.

Frequently asked questions

What is the formula for a sales forecast?

The simplest version is expected units sold multiplied by price per unit. For a delivery business, build the unit side first, taking baseline units from comparable periods and adjusting for seasonality and known changes, then apply price afterwards, since units are what drive routes and purchasing.

How far in advance should you forecast sales?

Run a rolling four to six weeks in weekly buckets for operational decisions such as rostering and purchasing, plus a twelve-month view in monthly buckets for hiring and vehicle decisions. Beyond twelve months you are planning rather than forecasting.

How much sales history do you need to build a forecast?

Twelve months lets you see a full seasonal cycle and is the comfortable starting point; two years is better because it lets you separate seasonality from growth. With three to six months you can forecast a short horizon from recent trend and supply the seasonality from judgement.

What should you do when the forecast is wrong?

Record the size of the miss and the reason in writing, then decide whether the cause is recurring. A one-off such as weather needs no change. A recurring cause such as consistently under-calling holiday weeks means changing how you build the baseline for those weeks.

Should a sales forecast match your sales target?

No. A forecast is your honest estimate of what will happen and a target is what you want to happen. If the target is written into the forecast, you will roster drivers and buy stock against a number nobody expects to reach.