Sales have long been an Assumptions driven mechanism, thereby having too much reliance on the gut feel of a Sales Rep, Sales Manager, often leading inaccurate sales forecasting, fudged pipeline and non-transitional Sales Funnel, leading inconsistent Revenue projections and tough quarter on quarter results. Today, with the availability of Data, and leveraging Data Science techniques, there can be an efficient Sales Engine Augmentor created, that can help Sales Managers to forecast with more precision, leading to faster sales cycles and unearthing minute customer centric characteristics. We tried leveraging Data Science on following accounts enabling the Sales and Inside Sales organization within an enterprise:
“Is Sales based on Assumptions better or we drive it through Data Points, determining effective Sales Funnel and Agile Conversions ?”
Sales Cycle and Management
Aimed at the Sales Managers, the analysis looked at the profile of sales cases within their responsibility. Aim was to highlight sales cases to be examined in more detail rather than the Sales Manager having to evaluate for themselves what to focus on, thus making the sales coaching sessions shorter and more effective
Sales Velocity
Evaluate each sales person against past performance and then use that to project future performance given the content of their sales funnel. It also highlights areas of change within their existing funnel, for example number of leads (too many as well as too few), types and ages of sales cases. Used to suggest coaching and mentoring pairs
Sales Lead Viability
Sales leads are scored on their content. This is used as an indication or urgency and using text processing route the lead to the most appropriate group, for example if web contact mentions jobs in a particular country that would go to HR there
Marketing analysis forecasting
Model which methods and campaigns had the most impact overall and which customers responded to which type of campaign the best. Challenge here was to tie together multiple sources of data to produce a viable data set to work with, I.e. web traffic, email, leads and sales case
1. Lead Quality
Content of each sales was evaluated to see for new contacts how likely they would become new customers. For existing contacts, how likely there could be new opportunities
2. Matching sales cases
Acquisitions resulted in the enterprise having multiple CRM systems in operation. In an attempt to minimise effort assignment to find duplicate sales cases between each system. Despite customer data being essentially the same: name, address, registration number, tax number and DUNS number, the quality of this wasn’t good enough to use. Created a number of matching algorithms to score if the customer and the sales case was the same. Eventually created lists for local sales management highlighting which sales cases were likely to be the same. As another benefit, it also highlighted some as to then unknown customer behaviour.
3. Define a Model Sales Case
For each of the main product types a definition was created which described the ideal sales case. Initially this was a combination of manual and automatically generated input, you may want a 20K sales case but 15K is more usual. Sales cases then compared against the model to see how realistically they could be served and to highlight ones for different treatment
4. Sales Process: Actual flow vs happy flow
Purpose was to understand which stages a sales case goes through in its lifecycle. Process mining techniques were used to see if there were optimal steps which led to a successful outcome. Together with comparisons against perceived optimal flows, produced insights which led to a better actioning of sales cases at given points in their life cycle
5. Volumes
Model sales cases over time to forecast likely sales
6. Sales Path
For a given sales case what is the next best action to be carried out
7. Model customers
Definition of companies who would be most likely to use our services. Using external data to identify companies who matched the best with the ideal customer profile
8. Customer Behaviour
-
- – Model customer behaviour in areas of
- – Offer and order frequencies
- – Offer to order probability
- – Offer to order time scales
- – Value of offer vs order received
- – Preferred sales person
- – Built an app that could then, based on the sales funnel, predict sales both in terms of value and time for one product type
9. Inside sales Data Experience (DX)
Design the data needs for Internal Sales role. Drew together a number of themes of customer behaviour, next best actions and viability so prioritise which opportunities should be done in which order to gain the maximum impact, i.e. time to order. This also brought in data from maintenance service as support to offers made and was supported by a quotation app which sped up costing repairs All placed in one app rather than using numerous applications and reports
