Sales Forecasting from Your Books: Practical Methods That Work for SMEs

Sales forecasting has a reputation as complex, statistical, and expensive. For an SME, that reputation is mostly wrong. The cheapest and most reliable forecasting input you have is your own sales history, and a handful of straightforward methods get you 80% of the way without buying a tool or hiring a statistician. Here is the working approach.

Why even bother forecasting

A forecast — even an imperfect one — drives decisions that need to be made anyway:

• Inventory planning — how much to make / buy / stock
• Cash forecasting — what the next 90 days look like
• Capacity planning — do we need more people, more space, more machines?
• Pricing decisions — are we below capacity (cut prices, fill capacity) or above (raise prices, manage demand)?
• Borrowing planning — when will we need short-term credit; when can we repay

Not forecasting just means making these decisions on instinct. A simple forecast beats instinct in almost every case.

Method 1: Moving average

The simplest method: take the average of the last N periods (months, quarters, or weeks) and use it as the forecast for the next period.

• Simple Moving Average (SMA) — equal weight to each period. E.g., average of the last 6 months' sales is the forecast for next month.
• Weighted Moving Average (WMA) — recent periods weighted more heavily than older ones. E.g., last month × 3, previous × 2, the one before × 1, divide by 6.

Use SMA when sales are relatively stable. Use WMA when recent trends matter more (typically the case).

A practical period choice: 3-month WMA for short-term forecasts, 6 or 12-month SMA for longer horizons.

Method 2: Year-over-year with growth rate

When there's clear seasonality (most businesses), comparing to the same period last year often beats moving averages.

• Take same month last year
• Multiply by your year-over-year growth rate (e.g., 15%)
• That's your forecast

If your sales in October last year were ₹40 lakh and you're growing at 20% year over year, October this year forecast: ₹40 lakh × 1.20 = ₹48 lakh.

This handles seasonality naturally — a December peak last year produces a December peak forecast this year, scaled by growth.

The growth rate should be computed from recent history (last 6 months actuals vs same 6 months prior year), not picked arbitrarily.

Method 3: Decomposition — trend + seasonality + noise

For businesses with both growth and seasonality, the cleanest approach decomposes sales into three components:

• Trend — the underlying direction (growing, flat, declining)
• Seasonality — the repeating pattern across the year
• Noise — random month-to-month variation

The forecast = Trend (extrapolated) × Seasonality factor for the period.

Practical version:

• Compute a 12-month moving average — this is the trend
• For each month, compute (actual ÷ trend) — this is the seasonality factor
• Average the seasonality factor for each month over the years you have data — gives you stable seasonality factors
• Forecast next year's trend (linear extrapolation), apply each month's seasonality factor

The math is spreadsheet-doable. Most accounting tools that produce sales trend graphs are doing this implicitly.

Method 4: Bottom-up — by customer and product

For B2B businesses with concentrated customers, top-down statistical methods often miss the truth. Instead:

• Forecast each major customer individually (most B2B businesses know their top 10 customers' likely buying patterns)
• Aggregate the long tail using historical averages
• Add new-customer pipeline

This typically beats statistical methods for B2B because it captures known information (a key customer's expansion plan, a competitor's loss of a key contract you're picking up) that no time-series method can see.

The downside: requires more manual work and customer-by-customer thought.

Forecast vs target — keep them separate

A common confusion: forecasts and targets are different things.

• Forecast — what we expect to actually happen
• Target — what we'd like to achieve

The forecast drives operational decisions (how much to make, when to order). The target drives sales effort and motivation. They should differ, sometimes significantly. Conflating them — using the target as the forecast — leads to overstocking and over-investment.

A discipline: ask sales for their forecast (what they realistically expect) separately from their target (what they're committing to push for). Use forecast for operations; use target for accountability.

Tracking forecast accuracy

A forecast is only as good as its track record. Track:

• Forecast accuracy = 100% − |Actual − Forecast| ÷ Actual
• By period (last 6 months' accuracy)
• By product / category / customer (where are we systematically wrong?)
• By forecaster (if multiple people make forecasts)

A consistently optimistic forecast (always above actual) is just as problematic as a consistently pessimistic one. The bias matters more than the absolute accuracy — bias is fixable; random error is harder.

A practical accuracy benchmark: ±15% at the monthly company level is excellent for SMEs. Beyond that you're getting into diminishing returns from method sophistication.

What forecasts cannot do

A forecast extrapolates. It cannot see:

• Sudden category-level shifts (a viral trend, a regulatory change)
• Competitor moves (a major launch, a price war)
• Macro shocks (the pandemic was the obvious example)
• Internal disruptions (a key salesperson leaves, a stock-out, a production issue)

A forecast is a baseline; the management overlay handles the things forecasts can't see. Both are needed.

How Booksmor helps

Booksmor produces sales forecasts at total, category, customer and channel levels — using moving averages, year-over-year growth, and trend-plus-seasonality decomposition. Forecast vs actual tracking, accuracy reporting per period, and integration with inventory planning (so the forecast drives reorder points) come built in. Start a 30-day free trial and forecast from real data, not gut feel.