Sales forecast: how to calculate one you can trust

A monthly sales forecast is calculated like this: add expected payments on open deals to the money already received, adjusting each for the probability of payment by the end of the month. You need the amounts of the remaining payments, their due dates and a history of similar deals. The business owner gets an estimate of future incoming payments and can see which buyers it depends on.

It gets harder when this total has to support a decision: increase purchases, use production capacity or hire another person. The head of sales is confident about the customers. But confidence that an order will go ahead says nothing yet about whether there will be enough money by the date payment to the supplier is due. This is the gap to close before discussing how accurate the percentages are.

What should you forecast: orders, revenue or incoming payments?

Forecast the measure that the owner's decision depends on. An order, revenue and payment relate to different events; you cannot add them into one total.

From here on, we calculate incoming payments from sales. Amounts already received come from payment records; future amounts come from deals and payment schedules. The rules for VAT, refunds and cancellations must be the same in the forecast and the actual figures. When forecasting money in the account, count incoming payments and refunds at the amounts that will pass through the account.

For example, a contract worth 1 million ₽ provides for an advance of 300 thousand in November and the remaining 700 thousand in December. November's calculation of future incoming payments includes 300 thousand. If the advance has already been received, it is part of the actual figures and is not counted again as expected money. The contract value stays the same here; the payment calendar changes.

Keep the sales plan alongside it, on a separate line. The sales plan is how much the company intends to receive. The forecast is how much it expects under the current deals and conditions. Adding the amount needed to reach the sales plan to the forecast means hiding a task that still needs to be done.

What data do you need for a sales forecast?

The calculation needs a list of future payments linked to deals, and a history of how similar buyers progressed towards payment. The owner only needs to approve what goes into the calculation; the head of sales should compile it together with the person who keeps track of the money.

The working spreadsheet or export from the CRM system needs these fields:

• Deal and buyer — so the same payment does not end up in two rows.
• Remaining payment amount — after money already received and agreed changes to the order.
• Purchase stage — what the buyer has already confirmed.
• Date the deal entered the stage — how long the deal has been in its current state.
• Expected payment date and its basis — an agreement, a schedule or a required step before payment.
• Comparison group — product, new or regular customers, similar purchase terms.
• Person responsible and update date — who is responsible for the information and how recent it is.

With instalments, one deal takes up several rows, one for each payment. For repeat purchases, the future order is counted separately: a past payment does not prove that the next one will arrive on the same day.

The “proposal sent” stage describes something a salesperson has done. For the calculation, it is more useful to know whether the buyer has received the proposal, whether the contents of the order suit them, and who makes the decision. The article “Sales are happening. But which figure can you trust?” examines the gap between a CRM entry and the basis for a forecast in detail.

How do you choose a sales forecasting method?

For the coming month, a sales team working with individual deals bases its main calculation on the sales funnel. Sales history provides a guide, while a calculation based on future enquiries is needed for a period that includes enquiries yet to arrive.

Calculating the average for past months is simple: add up incoming payments and divide by the number of months. For example, 4, 5 and 6 million ₽ give an average of 5 million ₽. But if the most recent month included a one-off order, you cannot carry it into the next month without a basis for doing so. Compare seasonal months with those from the same season, accounting separately for changes in prices and the mix of buyers. The history tells you what happened under the previous conditions.

For new enquiries, the calculation is different: multiply the number of expected enquiries by the share of buyers who will pay within the chosen period, and by the average payment. For example, 120 new enquiries × 20% × 100 thousand ₽ = 2.4 million ₽. Here, 20% is the share who pay within the required time, and 100 thousand is the payment for that period, not the full order value.

Only add this amount for future enquiries that are not yet in the current sales funnel. Otherwise, some buyers will be counted twice. If there is no history to support the volume of enquiries and the share that result in payment, show several scenarios with different starting figures. It is too early to choose the most appealing one as the only forecast.

How do you calculate a sales forecast from the sales funnel?

Multiply each payment you have yet to receive by the probability of receiving it by the chosen date, add up the results, then add the money actually received since the start of the period. Subtract expected refunds separately. Take probabilities from comparable historical data, rather than assigning them to stages for show.

For example, you reconstruct the position as at the 10th of previous months. There were 20 similar deals at the relevant stage; 12 of them brought in the scheduled payment by the end of their respective months. The share of payments made on time is 12 ÷ 20 = 60%. The remaining eight stay in the original group used for the calculation, even if the buyers paid later or left without buying.

Compare deals with similar lengths of time at the stage and similar payment terms. A tender requiring several approvals and a repeat order from a regular buyer need different groups. Twenty deals in the calculation above illustrate the arithmetic; that is not enough historical data in all cases. When sales are infrequent, each deal has a large effect on the result, so you also need scenarios for the outcomes of large orders.

Let us calculate a month using round numbers. Suppose 2 million ₽ has already been received, no refunds are expected, and your data gives the following estimates for the remaining payments:

Future incoming payments: 1.2 + 0.6 + 0.9 = 2.7 million ₽. Forecast for the whole month: 2 + 2.7 = 4.7 million ₽. The total of payments still outstanding — 6 million ₽ — does not, on its own, give you grounds to expect all 6 million.

In the table, the probability refers to receiving the stated payment in full. If the buyer has agreed to pay in instalments, first split the amount by payment date. Simply reducing the percentage for the whole deal will hide exactly which part of the money is due to arrive.

Why is the probability of a deal not the probability of payment?

The probability of a deal answers the question of whether the purchase will go ahead. A monthly forecast needs something else: will a specific payment arrive by the end of the month? Even a customer who has agreed to buy has to go through approval, paperwork and payment.

Calculate the remaining steps using confirmed timings. For example, if the steps happen in sequence, approval takes three working days, preparation for dispatch takes five, and payment once the order is ready takes another two. That makes ten working days. If there are seven left until the end of the month, the previous expected payment date no longer fits this sequence. A high probability of purchase does not make up the three missing days.

A shorter timeframe must be based on a change in conditions: the steps will run at the same time, the buyer will pay an advance, or the goods are already ready. A salesperson's intention to speed things up is not such a basis. How to find these delays before the monthly goal is missed is explained in the article “Sales plan: how to spot a shortfall before the middle of the month”.

For a quarter, calculate each month separately. Moving a payment from November to December preserves the quarterly total, but November's commitments remain in November. A single figure for the quarter hides this problem.

How do you account for large deals and shared risk?

A large payment needs to be shown to the business owner in separate scenarios: it arrives on time, or it is delayed. A deal's average expected contribution does not mean the buyer will transfer exactly that amount.

In the table, the order worth 1 million ₽ contributed 900 thousand to the forecast. But if the payment cannot be split, the buyer will transfer a million or will not transfer it this month. With the estimates for the other deals unchanged, the total without this payment will be 3.8 million ₽, or 4.8 million ₽ with it. These are two scenarios for one order; the lower amount is not guaranteed either, because the other deals are still open.

Here is what I need to see alongside the total: which expected payments will disappear together because of a single delay. Three different customers are waiting for the same batch of goods — a delay to that batch will push back all three payments. Averaging across customers will preserve the average estimate, but hide the size of the combined shortfall.

So build a cautious forecast scenario around specific events: a batch is delayed, a large buyer's decision is postponed, a repeat order does not arrive. An arbitrary “minus 20% just in case” does not explain under what conditions there will be less money, or what to do about it.

How do you check the accuracy of a sales forecast?

Save the forecast as at a specific date and compare it with actual incoming payments once the month has ended. To assess the method, compare forecasts made with the same amount of time left: for example, on the 10th of each month.

Take the size of the difference between the forecast and the actual amount, divide it by the actual amount and multiply by 100%. For example, you expected 4.7 million ₽ and received 4.1 million: the error is 600 thousand ₽, or about 14.6%. When the actual amount is zero, you cannot calculate a percentage — the difference in roubles remains. There is no universally acceptable margin of error: what matters to the owner is the amount that would force them to change a decision already made.

As well as the size of the error, look at its direction. Consistent overestimation means the basis of the calculation needs to be reviewed. Alternating overestimates and underestimates means you need to examine the variation; differences cancelling each other out over a quarter do not make monthly forecasts accurate.

Attribute the cause of each large difference to the amount, the timing or the purchase itself. The buyer paid less, paid later or decided not to buy — these are different errors with different fixes. Keep the old version of the calculation: if you overwrite it, there will be nothing to compare once the month is over.

How do you set up forecasting without the owner recalculating it by hand?

Make one person responsible for the final total, agree rules consistent with how money is recorded, and set a time for updates. The head of sales is responsible for expectations on deals; the person who records payments confirms what has been received. The owner gets the total and significant changes, rather than several competing spreadsheets.

For the first run, one month and the current sales funnel are enough. Then add future enquiries and a longer forecasting period. Before automating the calculation, the payment figures must match, advance payments must no longer be counted twice, and a saved version of the forecast must be in place. Otherwise, the new software will simply add up the old mistakes faster.

Update the working total on a set day of the week and when an event changes a decision about money. The report includes the actual amount, expected payments to the end of the month, the main things these depend on, and an explanation of the change. If the owner does this analysis every time, they are still doing the work of the overall head of sales. At RENTROP, turnkey sales management means handing over the day-to-day running of the sales department to a team, so the business owner does not have to piece together the situation from individual salespeople.

The request for one figure is understandable: it makes decisions easier. But 4.7 million ₽ with details of the payment on which almost a million depends is more useful than the same 4.7 million without an explanation. In the first case, the owner understands the cost of a delay before taking on commitments. In the second, they can only take the head of sales at their word.