peterosborn.com

A senior sales exec I know prides himself on his team being within 5% of forecast. He's spent much of his career selling big ticket technology sales into complex organisations, which are notoriously hard to forecast. So if he's that good, would he be able to forecast sales of cold drinks at Glastonbury this year? I doubt it. The sales dynamics are different, and the forecasting method would need to be different too. Matching method to dynamic is crucial.

Forecasting a single large sale into a big organisation is a lengthy and political business, fraught with complexities and a blizzard of separate influences. Calling one of these sales is all about knowing what you don't know, understanding corporate behaviour and the attitudes of the people in this particular customer. None of this is relevant to a forecast of drinks sales at Glastonbury, just as the weather will have little effect on when a contract will get signed for a risk management system at a major bank in Denmark.

A frequent mistake I see is the use of factoring - the practice of applying a probability to the value of sales. It's vital to understand where this will work and where it won't. There are three attributes to understand about an opportunity in order to make a forecast: Value, timing, and probability, and they are separate characteristics. You may have a 20% chance of a sale at £1,000, but that doesn't mean it's worth £200 on your forecast - either you'll make it at £1,000 or you'll get zip. Factoring like this only works if you have a large enough population of sales. If a stage is rated at 20% (1/5) then you must have significantly more than five sales at that stage for factoring to be anything better than dangerous.

I have yet to see percentages arrived at scientifically. They are most often set subjectively. Sometimes they're just the default percentages set in the CRM system when it was implemented. There's an astonishing range of ranking methods: "Hot Prospects", "Green and Amber Sales", you name it. All these are subjective, and should be a big warning about the forecasting culture.

The first step to choosing the right method is to understand the cycle. How much time lies between now and the customer buying decision in question, and what process lies between the customer decision to buy and when the sale is certain. Buyers of soft drinks at Glastonbury do not make their buying decisions in May, but we can take a view based on previous years, the long range weather forecast for June, and the likely number of attendees. Provided that we have the drinks and the staff to sell them, there's little else we need to know to produce a forecast of sales.

A similar methodology can be used to decide how to forecast more complex sales. Another key influence is whether you sell or take orders. If a customer has pretty much decided to buy before he makes contact with you, then your only foreknowledge of customers comes from trends and a guess at the results of marketing activity, because there is little in the early stage of the pipeline. So you should be able to judge incoming customers, and then apply judgement about timing and probability of close by knowing your hit rate in the past.

More problematic are the small number of lumpy sales into complex organisations, of the kind my friend is used to making. In this world, there is little substitute for building up your forecast from a detailed understanding of each opportunity, what you know, what you don't know, whose buy-in you have, whether budget has been approved, and so on. Here, though, the quickest way to get the forecast wrong is to make one; one is no use, and it'll almost always be guaranteed to be wrong.

It's very hard to come up with one forecast in the world of complex selling, and the forecast can only be understood with a good knowledge of the relative optimism of the people providing input. Much more useful information comes from producing a forecast range: Low and Slow, Upside and Early. This allows and requires a better judgement about the possible outcomes. It also gives a valuable input to planning for contingencies.

Some science can also be applied to a judgement of probabilities. Once the customer has told you that they've decided to buy, you ought to be able to make a good call on whether the sale will come in or not. After this point, you can be more certain, and opportunities can justifiably be forecast. You ought to know the steps to close, and they should be reasonably visible and generic. You might insist on an opportunity closure plan for each item in this forecast.

Before this point, before the customer has decided to buy from you, there are a large number of potential influences, for and against. If the majority of possible sales lie in this group, then your forecasting methodology will need to be designed to reflect all of the significant influences.

Good forecasting is built from a scientific understanding of the nature of the sales cycle and the influences on it. Subjectivity and crystal balls have no part to play.

(c) 2010, 2011 Peter G. Osborn