Ultimately, though, any reasonable tools or models are only predictive in the sense that they map forecast to bookings based on knowing the type of customer that you have already sold to with a saleforce that is seasoned enough to know how to translate their body language into some form of number-morphing formula.  There is no provision for a new type of customer, a larger (or smaller) deal size than is normal for your company nor a change in the experience level of your sales team.

In my experience, any such tools that have been employed have been repeatedly beaten by a process I have used – it’s called: ask the Sales VP.  As the end of each quarter grew near, I asked the head of sales to tell me what’s going to come in.  His/her response was based on his knowledge of how each individual sales person (or territory or region or division) worked; the type of customers that were involved in the biggest deals in the pipeline; the level of support those prospects had; how high up in the organization that contacts had been made, and so forth.  Even then, many of deals that made the final list were because: “My gut tells me that one will come in.”  He/she took into account a myriad of factors, some of them only valid for that particular time and place.  It’s difficult for models to adapt like this.  Perhaps there is some form of calculus that I have not discovered though.  In the end, I always had comfort that the smaller the difference between the bottoms-up forecast that came from the salespeople and the Sales VP’s “gut feel,” the more likely it was we would hit a particular number.  The larger the difference indicated problems that need to be immediately ferreted out.  

For this method to be successful, of course, one has to have a sales manager that has the experience to make the call.  I have always believed that the level and experience of this person is one of the key factors in the success or failure of an organization.

As mentioned elsewhere in these comments, this final Sales VP cut could be considered just another tool in you quiver.  The wisdom and experience of the head of sales was the secret sauce.  It’s certainly not the only data or process you can work with, but I’ve found it the most accurate.

However, if individual sales are relatively small, then a mathematical forecasting process MUST take place, because there is no way to be that tuned-in with the individual customers.  And in a larger company, there is no way to get a decent forecast from the larger organization by hoping that individual managers are good at it – but on the other hand larger companies can gather good statistics over time and have enough prospects in the pipeline that the statistics can work.

There are three things that need to happen to make “expected value” forecasts work:

1. The “stages” need to be objectively defined, rather than subjectively assessed by the salesperson.
2. The “percentages” need to be based on HISTORICAL ACTUAL results rather than subjectively estimated probabilities.
3. N (the number of individual prospects in the pipeline, and the number of historical datapoints) needs to be large relative to the amount of any one sale.

Applying percentages and equations to sales forecasting is trying to force fit a science on something that is at best an art and one with incosistant training and capability of the artists.

The size of the deal is important.  The stage of the deal is important.  But multiplying the two makes no sense at all.  However, there are some mathmatic things and simple “equations” that can be applied to add some usefulness to the din.

Every organization I have been in from Hewlett-Packard to several startups have tried to apply a method for acquiring forecasts.  The one that I like the best is where we’ve taken the anticipated order size (ie 50K) and multiply it by 2 percentages that the sales dude is left to determine:  percent chance that ANY order will be placed by this customer and the percent chance that a placed order will be ours.

The first (chance of placement) takes into account the sales person’s knowledge of the customer, buying habits, stage of the order, and overall gut feel.  This number replaces the straight percentage based on stage that you’ve discussed, which I agree that by itself is virtually worthless limited.  This number is an attempt to quantify all the sales person’s knowledge, not just the over simplistic stage which reflects NONE of this good stuff.

The 2nd percentage (chance that if placed, you get it instead of competition) reflects the competitive nature of the deal.  If you’re head to head on two deals and give each 50%, it says and reflects that you are likely to get one, but probably not both.

By multiplying these two percentages by the deal size, you end up with a weighted order size that directly reflects the gut feels of the sales person.  Again, on a sales person by person basis, this forcast is virtually guaranteed to be wrong.  (100K order, 90% sure it will place, 70% chance I’ll get it gives you a forecast of 63K which can’t happen since ultimately it is either 100K or zip.

But, if I am working 10 such deals, statistically it becomes more reliable, and if I’m one of 100 reps, the 1000 deals being worked starts to provide a number that you might be able to plan against.

Ultimately, any forecdast that is mathematically derived is only useablle within the context of the rules you’ve used.  Multiplying by the stage of the order IS inherantly flawed since it assumes all deals will result in orders to your company, its just a matter of time.  If you have ten 100K deals at the 10% mark, this model says you have high confidence that you are going to book 100K.  

The final factor to throw in is the estimated month of closing.  With 20, 50 or 100 sales people, you end up with a set of numbers that is low maintenance for the sales person to maintain, and one that IF INTERPRETED REASONABLY gives you some basis for cash flow and production planning.

At a couple of my companies, depending on what information we need we look at this forecasts on a quarterly basis, and we play games like taking 50% of the current month, adding 50% of the previous month and 10% of the following month to reflect the high tendency for orders to slip and smaller but still possible tendency for orders to fall early.

Tools.  Nothing but tools.  But, they can be useful.