When building a sales forecast, whether it’s a simple Excel table or a more sophisticated model, it’s always a good idea to think about what data are you including in it, and what information you decided to omit, for model simplicity. Salesperson’s tenure usually gets omitted, especially in the simpler models, because it’s not obvious how to include it.

Based on my experience from on the machine learning models that power TwelveZeros’ Sales Forecasting product I see that for many customers salesperson tenure is a big factor impacting forecast quality.

In our models, we break down tenure into two different factors: one is the number of years of experience in Sales in general, of the analyzed salesperson. The other one is the number of closed opportunities in their current role.

Number of closed opportunities

The number of closed opportunities is more practical of those two. First, it’s directly applicable to the results you want to predict – it’s based on selling the same product, in the same organization, given similar conditions. In my experience, just the two numbers – how many opportunities did the salesperson work and how many did they close, tell a lot about chances for future success.

Based on that you can calculate salesperson’s historical win rate. If the rep’s historical data show that they close 20% of their opportunities, and for this quarter they’re committing to close 40% of their pipeline, you’ll know that that number is unrealistic and should be reduced.

It’s also important to remember about the confidence interval when looking at rep’s win rate. A salesperson that won 50% of their opportunities (1 out of 2) is not necessarily “better” than a person that won 30% (30 out of 100 deals). You can get pretty sophisticated with the statistics here, but I think even without that, just having those numbers in front of you will help you judge how trustworthy is a commitment from each salesperson individually.

Number of years of experience

We’re using the number of years of experience of the sales rep as one of the inputs to ML models and that’s interesting too – especially at the beginning, in the first months after being hired. But after a few months, when the ML model gathers enough data about the actual track record of the salesperson, this information becomes less significant.

I think there’s no easy way of using years of experience in a simple Excel-based model, and it doesn’t bring much value (unless you’re hiring a lot), so maybe it’s not worth the trouble.

Machine Learning based sales forecast

In general, Machine Learning forecast models can take into account multiple factors, or even factor combinations (seniority might matter in selling one product line, but might not matter at all when selling a much simpler product). If you’d like to learn what factors impact your sales forecast, give TwelveZeros a try by filling the form below.

Co-founder, Head of Data and Machine Learning at TwelveZeros. Out of the 10 years of his professional experience building software, he spent last 6 solving problems of sales organizations using a scientific approach, data, and AI tools.