How to Forecast Ecommerce Sales

An ecommerce sales forecast is built from traffic, conversion and order value by channel, adjusted for returns, seasonality and planned marketing. Because every step is measured in your store's analytics, ecommerce is one of the easiest businesses to forecast with drivers and one of the easiest to get wrong by ignoring how drivers interact: more advertising spend does not buy proportionally more traffic. This guide builds a channel-level model with a worked example, adds returns and net revenue, explains marginal versus average return on ad spend, covers peaks and promotions and shows how to connect the forecast to inventory and cash. It is part of budgeting and forecasting and applies the driver approach in how to forecast revenue.

The basic formula

Revenue = Sessions × Conversion rate × Average order value

Forecast each part separately, by channel, because channels differ in cost, conversion and order value.

Step 1: Break traffic into channels

Step 2: A worked monthly forecast

A hypothetical store plans the next month with these channel assumptions (from the last three months of data):

Check orders: 40,000 × 0.024 = 960; 30,000 × 0.018 = 540; 15,000 × 0.04 = 600; 10,000 × 0.03 = 300. Total 2,400. Blended conversion = 2,400 / 95,000 = 2.53%. Revenue = 2,400 × $62 = $148,800. (Using a single average order value for all channels is a simplification; use separate values when they differ.)

Step 3: From gross to net revenue

Gross sales are not what the business keeps. Adjust for returns and discounts.

Return rates vary by category and should come from your own history, by product group. Include the cost of return shipping and the resale value of returned stock when you look at margin. Shipping charged to customers, taxes collected and payment fees affect the numbers that reach the income statement; follow your accountants' guidance on presentation.

Step 4: Marketing spend and diminishing returns

Paid traffic is the lever teams adjust most, and the relationship is not linear. Suppose paid spend of $12,000 produced the 30,000 paid sessions above.

• Average return on ad spend (ROAS) = paid revenue / spend = $33,480 / $12,000 = 2.79.

Now management considers a 25 percent budget increase ($3,000 more), and expects traffic to rise only 15 percent because the best audiences are already reached.

• New paid sessions = 30,000 × 1.15 = 34,500
• New paid orders = 34,500 × 1.8% = 621, an increase of 81
• Extra revenue = 81 × $62 = $5,022
• Incremental ROAS = $5,022 / $3,000 = 1.67

The average ROAS of 2.79 hides a marginal return of 1.67. Whether the extra spend is worthwhile depends on gross margin: at a 45 percent gross margin, each extra dollar of revenue earns $0.45 of margin before other costs, so $5,022 of revenue produces $2,260 of gross margin for $3,000 of spend, a loss on the first order unless repeat purchases follow. Repeat purchase value changes the answer, which is why customer lifetime value belongs in the decision. See unit economics.

Step 5: New and returning customers

New customers and returning customers behave differently: returning customers convert at higher rates and cost less to reach. If your platform allows, forecast them separately.

Step 6: Seasonality and events

Peaks (holidays, sales events) drive a large share of annual revenue for many stores, and their timing, discount depth and promotion calendar matter. Build a calendar of planned events with expected lifts based on previous equivalent events, and avoid applying the same seasonal index to a year with a different promotion plan. See adjusting forecasts for seasonality. For a sales event, forecast: sessions lift, conversion lift, average order value change (discounts often reduce it) and returns, which may be higher after promotions.

Step 7: Connect to inventory and cash

Ecommerce stores often run out of stock at peak, or end the season with surplus.

Stockouts depress both revenue and paid-traffic efficiency, since ads send visitors to out-of-stock pages.

A worked promotion forecast

Take an ordinary day: 3,000 sessions, 2.5% conversion and a $62 order value give 75 orders and $4,650. A four-day sales event is planned, with these assumptions based on a previous event:

• Sessions 2.2 times normal: 6,600 a day
• Conversion 3.5%
• Average order value 15% lower because of discounts: $62 × 0.85 = $52.70

Event day orders = 6,600 × 0.035 = 231. Revenue per event day = 231 × $52.70 = $12,173.70. Four days = $48,694.80, against a normal four days of 4 × $4,650 = $18,600. The gross uplift is $30,094.80.

Not all of that is new. Promotions pull some demand forward from the weeks after. The extra orders are (231 - 75) × 4 = 624. If 15 percent of those would have been bought anyway in the following two weeks, 94 orders × $62 = $5,828 disappears later. The net uplift is about $24,267 (30,095 - 5,828). Margin is lower still because discounts apply to every order, not only the extra ones, and returns after promotions are often higher. A promotion plan should be judged on net contribution, not on the headline revenue of the event days. See unit economics.

From a store forecast to a company model

When the business adds categories, markets or wholesale accounts, a single channel table is no longer enough. Build each stream as its own driver block and sum them in a bottom-up revenue model, with a common assumptions sheet, so that marketing, merchandising and finance work from the same numbers.

Step 8: Measure accuracy and update weekly

Ecommerce data arrive daily. Compare the forecast with actuals weekly for sessions, conversion, order value and revenue. A miss in revenue should be traceable to a miss in one driver. Track accuracy and bias using the methods in how to measure financial forecast accuracy, and re-forecast the rest of the month or quarter when drivers move.

A simple checklist

• Are sessions, conversion and order value forecast by channel?
• Are returns and discounts deducted to reach net sales?
• Is the marginal return on extra ad spend estimated, not just the average?
• Are event calendars and promotions explicit?
• Are stock levels consistent with the forecast range?
• Is there a low and a high case?
• Is accuracy tracked weekly?

Common mistakes

• Applying a single conversion rate to all traffic.
• Assuming ad spend scales linearly.
• Forecasting gross revenue when net is needed.
• Using last year's peak as this year's without checking the promotion plan.
• Ignoring stockouts in the historical data.
• Forgetting that discounts lower the average order value.
• Not linking the forecast to cash.

FAQs

How do you forecast ecommerce sales?

Multiply sessions by conversion rate by average order value for each channel, subtract returns, and adjust for seasonality and planned promotions.

What is ROAS?

Return on ad spend: revenue attributed to advertising divided by the spend.

What is the difference between average and incremental ROAS?

Average ROAS divides total ad revenue by total spend; incremental ROAS measures the extra revenue from the last dollars spent, which is usually lower.

How do I forecast peak season?

Use equivalent past events, planned promotions and expected lifts in traffic and conversion, and check stock and cash.

How often should I update an ecommerce forecast?

Weekly, comparing drivers with actuals.