There is a long list of useful applications of predictive analytics in sales operations. In principal, at any stage where you can ask yourself whether to proceed or not or how to proceed you could theoretically employ a predictive model.
A very common application is the so called lead scoring. This is all about prioritizing your sales resources. In most sales situations there is a long list of leads in the database. Sales reps plan follow-ups on those leads, make phone calls and visit their prospects and so on. All this is time consuming. So, whenever you spent time on a lead that has a low probability of becoming a customer of yours this will incur opportunity cost - because if you spent your time on a “better” lead you will eventually increase your revenues.
Now, a lead scoring model will evaluate each lead with regards to the probability of conversion. Of course, you can build more complex models like expected lifetime value that a lead will create but the logic that is behind all this is always similar.
Let’s revisit the situation of the sales rep. Usually all the follow-ups and new incoming leads from the marketing department consume the day. This way the sales rep will most likely miss some of the best leads that will fall through the cracks. Making you of lead scoring the predictive model comes up with the so call lead score. So, instead just working through the follow-up list it makes sense to focus on the leads that have a higher probability of conversion.
The gain you expect is hard to predict without knowledge of the current situation. It mainly depends on the quality of the current prioritizing approach. The accuracy of the predictive model is another driver but - unless the data is in a very poor state or unless there is hardly any data - most of the times the predictive model will just do its job.
Depending on your sales process you can have the lead score updated whenever new information is available, e.g. the lead asked for a quote (or has not reacted to a follow-up on a quote).
There are many more applications you could think of. Common tasks are upselling and cross-selling and of re-engaging. In these cases the predictive models will similarly propose which customers are most probable to react positively to your efforts.