1

hopefully this isn't too dumb of a question. I've been thinking about it for a while now and don't have a clear answer. I saw this question, where the answer mentions the Poisson Distribution, but it doesn't seem like the Poisson applies here because of the time aspect of the Poisson (please feel free to correct if wrong).

Suppose I have a set of sales data like this:

| Opportunity | Dollar Value |

| Sales Opportunity 1 | \$34,000 |

| Sales Opportunity 2 | \$78,000 |

| ... | ... |

| Sales Opportunity x | \$98,000 |

Each "opportunity" represents a currently-pending sales deal. The actual value of the deal is subject to change as salespeople add/remove line items from the deal.

Given the standard deviation of this dataset, how could I calculate the probability of a certain deal closing out above a certain dollar level, e.g. $100,000?

For instance, sales opportunity 1 is at \$34K, and opportunity 2 is at \$98K, what is the probability of opportunity 1 closing above \$100K, and what is the probability of opportunity 2 closing above \$100K?

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
quantif
  • 59
  • 6
  • You can explicitely calculate the cdf $F$ of this distribution, and then evaluate $1-F(\$100,000)$. – Alex R. Jun 28 '17 at 23:52
  • I updated the description to make myself more clear, I can see how that answer would make sense given the way I worded the question. it seems like it is a little bit more complicated than that. – quantif Jun 28 '17 at 23:57
  • Unless each sales opportunity comes with a set of features behind it (i.e. the customer's age, income level, etc...), it's not clear how you expect to do better than just assessing the prior distribution over all dollar values. Whereas if you include such features, you could try a simple logistic regression on $>\$100,000$. – Alex R. Jun 29 '17 at 00:00
  • That is interesting, I hadn't thought about bringing in more features, thanks. I was thinking that were would be a way to calculate the probability of a certain value being greater than a certain level, given a measure of variance. – quantif Jun 29 '17 at 00:04
  • I have some features I can bring in, I'll probably try logistic regression. – quantif Jun 29 '17 at 00:05

0 Answers0