121 Silicon Valley, Inc. Advisory & Search 

CFO Sales Advisory November 9, 2009

Using Tropical Fish Charts to Manage Visibility

At the highest level, there are three main metrics in a sales forecasting system:  (1) total pipeline, (2) expected sales in the period (usually calculated as the sum of all pipeline deals, weighted by close probability), and (3) business closed so far.  For reasons that I think this audience understands, all three are critical to track, each for different reasons.  Even more valuable is comparing where you are in the current quarter to where you were at the same point in the last two or three quarters. 

While the information in a sales forecasting system captures critical information, virtually everything in the system is just a guess:  the size of each deal, when it might close, how likely it is to close, etc.  For that reason, the main value of a forecasting system is not to predict future results with scientific precision, but to help impose a uniform discipline on the sales process and to give management a sense of overall trends.  This makes a sales forecasting system an ideal candidate for presenting its key information in charts, rather than tables.  Here’s an example, recapping the weekly progress of the above three metrics for each quarter, through week 35 of the year: 

Here’s how to read this chart, which I call the “tropical fish chart” because of the visual impression I get from it: 

1. The broken lines are the total pipeline, the solid lines arethe weighted average/expected outcome, and the dotted lines are business closed quarter to-date. 
2. The Q1 forecasts are in blue, Q2 in red, and Q3 in green.  This company’sstandard practice is to begin tracking forecast progress for each quarter at the beginning of the prior quarter. 
3. There are solid vertical gridlines demarcating each quarter-end (weeks 13 and 26, so far).  On the last day of the quarter (OK, in the last minute of the quarter!), the three curves for that quarter should converge at the quarter’s actual result, shown as a small square in that quarter’s color.  [Note that I’ve included a green square for Q3, which is the target for the quarter.] 
4. I’ve included a thinner, broken vertical line at the current week (week 35, or 9 weeks into Q3) and also at the same point in the two prior quarters (i.e., weeks 9 and 22). 
5. In most organizations, the total pipeline curve peaks at an amount much larger than the eventual result for that quarter, and the closed business curve increases monotonically starting in the first week of that quarter.  The weighted average curve’s behavior is harder to predict, since it’s driven by both the total size of the pipeline and the rate of progress toward closing deals throughout the quarter. 

I like this chart because it provides an at-a-glance perspective on how sales are progressing through the quarter, and whether things are on track for achieving target results.  Here are a few observations you could make from the sample chart shown above: 

• The total pipeline for Q3 is light compared to where we were at the same point in the two previous quarters, especially considering that the Q3 target is significantly higher than the Q1 and Q2 results. 
• The weighted average number seems to be roughly on track toward achieving the target result, in spite of the relatively smaller total pipeline – perhaps this is because the fewer deals this quarter are making better progress toward eventual closing. 
• Closed business is much stronger at this point in the quarter than at the same point in Q1 and Q2.  This supports the argument that we are making better progress on the fewer total deals in the pipeline.  [And may generate a thank-you note from the manufacturing department!] 

As I’ve said here and in previous posts, sales forecasting systems are not a good place to look for “truth” in the sense that we usually think of that word with respect to numbers.  The main purpose of these systems is, after all, to predict the future, and that’s always messy.  But summarizing the information from the sales forecasting system in a coherent, consistent way is a powerful management tool for understanding how well the sales organization is functioning. 

CFO Sales Advisory October 18, 2009

Deal Close Probabilities:  SWAG vs. Zero Latitude

One of any sales forecasting system’s critical tasks is estimating just how much business the sales organization will actually close.  That is, you’re trying to boil that (hopefully) giant pile of suspects and prospects down to a single, critical number.  By far the most widely-used method for doing this is to assign a close probability to each deal, and the total forecast is then the sum of each deal’s size, times its close probability.  But how should you assign a probability to each deal?  You have two basic choices: 

1. Sales management uses its best judgment to assign a probability to each deal. 
2. The forecasting system assigns a probability to each deal, based on which of a series of predefined stages that deal is at in the sales cycle.  (e.g., face-to-face meeting = 15%, delivered proposal = 35%, in contract negotiations = 90%, etc.)

Many argue that Method 1 (management judgment) is the more “precise,” since we all know sales is a complex, subjective process.  Some “bluebirds” that arrive over the transom at the last minute do close, while some in the final stages of a months-long campaign don’t.  That observation is especially true since the object of all this computation is to estimate how much business will close in a given period. 

Even so, my strong recommendation is to use Method 2 (the system assigns the probabilities), for the following reasons: 

• Motivating behavior.  Under Method 2, sales people know that the only way to make a deal seem nearer to closing is to complete a specific action (like having a meeting, delivering a proposal, getting something in writing from the customer, etc.). 

• Reliability.  It’s asking a lot to require every sales manager to review and update the close probability of every deal in his or her territory every time you run the forecast.  Is this really how you want managers to be spending their time? 

• Consistency.  Personality, style, and even cultural differences among sales managers may cause variations in how different managers will assess exactly the same deal. 

• Personal agendas.  One of the risks of Method 1 is that managers may alter probabilities to attract or deflect corporate attention, or simply to get the calculated territory forecast closer to the manager’s gut feel number.  (Note that this is not always conscious behavior!)

• Relevance.  Does all this really matter?  Remember that the object of a sales forecasting system is to provide a consistent view of sales organization progress from period to period, and to motivate effective sales management.  These are not quite the same as demanding that the weighted forecast halfway through the period exactly matches the eventual actual result. 

I suggest that Method 1 is the better approach only if (a) your company typically has a relatively small number of relatively large deals in your pipeline, (b) sales managers typically get personally involved in the close process, and (c) the ways deals get closed in your company are so diverse that a standardized view of deal stages just doesn’t make much sense anyway. 

In my last post, I observed that trying to improve systems and processes sometimes causes us to forget that “better is the enemy of good.”  Inputting close probabilities into a sales forecasting system is not only an example of that dynamic, but also shows that moderate precision that serves a real purpose is better than extreme precision that’s pointless. 

CFO Sales Advisory October 1, 2009

Forecasting Systems:  “Improvements” Can Make Them Useless

A useful and effective sales forecasting system is one of the Holy Grails of sales management.  The components of such a beast will be topics of this post from time to time.  But before we address any of those components in detail, I offer this fundamental truth about forecasting systems:  A mediocre system that’s used consistently is better than a perfect system that isn’t.

As profound as that statement may seem, it’s even deeper than that!  Obviously, no forecasting system is credible unless every rep and manager whose input is required (the “users”) inputs reliably and consistently.  What’s less obvious is this:  Changes to the system, especially efforts to “improve” the system, can themselves have a significant impact on how reliably and consistently the system gets used.  What if the changes you are considering will: 

• Make inputting more time-consuming?  Virtually all significant changes to the system processes require the users to spend more time inputting data.  It’s a slippery slope:  the basic system just asks for prospect name and deal size, the next iteration asks for sales cycle stage, then product line detail, then activity summaries, and so on and so forth. 

• Make inputting more complex?  What if, for example, the proposed changes will require users to specify which of seven, rather than three, stages in the sales cycle each prospect is at?  Or the changes mean inputting detailed product line information? 

• Require more judgment and subjectivity?  For example, one approach to assigning probability of close is to tie a single specific probability to each stage (e.g., 100% for “closed,” 90% for “in contract negotiation,” 75% for “user signed off,” etc.).  Another approach requires a sales manager to input his or her own probability assessment for each deal.  Which approach produces the more reliable weighted average forecast? 

• Generate unwanted visibility?  Additional detail in the system on each prospect is a two-edged sword.  On the plus side, it enables company-wide collaboration to help close a deal, and provides valuable information.  But from the sales force’s perspective, that extra support may feel more like meddling, and an opportunity to be micromanaged. 

• Benefit those other than the system’s users?  For example, detailed product line information on all prospects might be much more valuable to marketing or engineering, than to the users who are just trying to close deals. 

Some of the consequences of these “improvements” can include:

• Input errors because the users just can’t (or won’t) devote sufficient time to inputting correctly,

• Input errors because the input procedures are more complicated,

• Inconsistent data in the system resulting from individual judgment and cultural differences,

• Data simply not entered, because users are concerned that (a) others will meddle, or (b) that they won’t personally benefit, through increased sales, from the significant additional effort that’s being asked of them. 

There’s no question that a good sales forecasting system is a living, breathing animal that is constantly evolving.  But every time you consider enhancing the system with the objective of getting more accurate, more reliable forecasts, you need to ask yourself:  When does “better” become the enemy of “good”?