Written in collaboration with : Abhinav Singh
Identifying the exact variables in a sales conversation that ultimately convert prospects into customers is not an easy task. Successfully closing a sale depends on several variables— many beyond human control— like customer persona, the urgency of need, market conditions, sales representative, customer emotions and more. As many of these variables are hard to trace, it is not easy to identify exact reasons for sales success. However, doing so would help organisations improve performance, replicate sales success, stall preventable losses, and increase revenue.
At SalesKen, bringing clarity by tracing these variables is one of the key problems we solve for.
The idea has been to generate actionable feedback for sales representatives to achieve their goals by giving recommendations on their sales activities, utilising:
- The conversations that have been happening between the sales representative and the prospect
- The prospect’s profile, and
- The representative’s pitching style, to improve the chances of the prospect buying the product.
Subjectivity is a fundamental problem in developing a heuristic system that reviews sales activity to tell the representative what works or doesn't. Such a flowchart and quality scorecard are subjective. Different sales experts can have different opinions about what might work for a certain prospect. This has thus far limited the kind of feedback that we provide to the representatives about what needs to be done to make a deal work.
There is a fundamental problem in trying to review a sales activity and develop a heuristic system that can then tell the representative what works and what doesn't. Such a flowchart and quality scorecard are subjective. Different sales experts can have different opinions about what might work for a certain prospect. This has thus far limited the kind of feedback that we provide to the representatives about what needs to be done to make a deal work.
Additionally, a lot of manual configuration is required. Sales leaders, who are adept with the script and processes, are often busy and unavailable.
The system does not adjust to the different sales associates who may have different styles of selling, does not change with time and needs excessive effort to adapt to a different team, product or to a slightly different use case and doesn’t accommodate diverse prospects who may have different needs and require different aspects of the sales script.
The SalesKen Advisory Solution
In SalesKen’s quest to solve these problems, we have developed a system which can do this in a highly scalable and automated way. We are training decision forests on the signal tags that SalesKen already generates. These signals range from conversational, emotional, prospect-related and generic. You can read more about the signals we can catch on the conversations we analyse here. We are able to achieve as high as 90% accuracy in the models we have trained on these data points in predicting the results of leads based purely on these signals.
Using the predictions made by our models we are able to solve a very fundamental problem in sales pipelines. The pipeline stages have static conversion probability, which greatly limits the accuracy of revenue prediction. By employing the pitch score and lead win chance in combination, we are able to make the predictions on the revenue more accurate. Over time we have seen such systems getting quite certain about the cash flows. This allows SalesKen to accurately predict the outcome of sales conversations and give our customers greater realistic visibility into their pipeline.
Keeping in mind the need to help our customers increase their sales, we set out to extract actionable insights from these mathematical models that can accurately predict the outcomes. What if the model could predict the least effort that could be made to drastically improve the likelihood of conversion of the prospect? It turns out that by deciphering where the model’s decision tree pushes the probability of conversion down the most, we are able to identify actionable insights out of these lead prediction models.
Key Takeaways From SalesKen Advisory
Along these lines, we are making Salesken Advisory more widely available. The key takeaways from SalesKen Advisory are as follows:
- The ability to make exact recommendations to the sales representatives about how they should proceed on their conversations with a particular prospect. We represent this information in actionable information cards. Moreover, these information cards are accessible before the sales activity starts, thereby setting the agenda for the associate to pursue. It also mentions how many deals have used this strategy successfully.
- Because we are able to customise what works for a sales representative in their conversation with prospects, we are also able to build a pitch delivery profile that is specific to the representative . This, in turn, lets us customise the recommendations for the representative based on their unique pitching style. We make both positive and negative recommendations on what to do and what not to do.
- We understand the importance of timing when giving feedback and, in that spirit, run predictions on our models on every sales activity. This enables us to make near real-time recommendations to the representative based on what has worked out in similar situations.
- Often, we see sales representatives struggling to independently engage in conversations with the prospect. By having models do the heavy lifting of ideation, we can move on to focus on the specific prospect, thereby enabling every sales representative to be as much in control of the deal irrespective of their familiarity with the context.
These features work out of the box as soon as you start tagging the leads in SalesKen as won/lost. You can use our already published connectors for CRMs like Salesforce, Zoho, LeadSquared and, Hubspot to start seeing this data today. Do go through our User Guide on SalesKen Advisory for more information on how to get started.
We are delivering Salesken Advisory information across multiple channels including WA, email, SalesKen device and web application to both sales leaders and sales representatives. We are segregating the advisory data at various levels-- sales activity level, prospect level, sales representative level and team level to be as specific yet comprehensive as possible. SalesKen Advisory data tells us what needs to be done at a specific sales activity level, and what needs to be accomplished at a macro level, like the team level. In our preliminary trials we have seen using SalesKen Advisory producing as much as a 15% increase in monthly sales figures across teams and pipelines.
The Way Forward
What does the future hold? In upcoming releases, we are focussing on three key areas:
- We want to introduce temporality as a variable in our models, which will enable us to keep the sequencing of sales activities and the ideal conversation graph in mind.
- We want the recommendation and SalesKen advice to work out of unbounded signals. We are pretty much limited by the bias of what we as users of the system deem important as part of the conversion recipe, and we want to minimise this bias to the extent possible.
- Although not all activities involve real-time bidirectional communication, we understand the unique value such channels of communication hold and are actively developing highly scalable real-time systems with low latency that can keep making recommendations as the conversation progresses. Using pitch scoring and lead prediction models will show the most optimised playbook for the representative to succeed.