Using Speech Analytics to Drive Sales Revenue, Better Target Marketing and Increase Employee Engagement
Many early adopters of Speech Analytics in Australia have been, in the main, disappointed with what the technology has delivered versus what was promised.
Recently, three major Australian companies have written off their speech analytics acquisitions. In each case, executive sponsors were less than enthralled by expensive ‘word boosting’ add ons required on top of the initial software investment – only to be presented with customer word clouds and graphs that raised more questions than they answered.
Many multiples of the original licence or rental fee has been invested in trying to get a result from the system. For those that have struggled through and have identified specific words, the reward from the business is often “So the customer says this word or phrase, so what?”
This paper will outline how through an intelligent application of KnowledgeSpace analytics solution, a third party outsourcer was able to:
- Increase sales conversion from 21% to 35.5%.
- Provide value added insights into what consumers were calling about, why they didn’t buy and what campaigns could be designed to convert these prospects.
- Identify which campaigns they managed generated the most interest and why.
- Reduce agent turnover by building a culture of success.
Where did we begin?
Our experience has proven that even the best reports produce no measurable changes if the solution is not owned and actioned by the business. We worked with our client to set up three levels of project governance with weekly meetings at the:
- Executive Levels – to set the strategic direction, manage stakeholder expectations and communicate the project objectives and anticipated outcomes.
- Management Levels – to communicate across the frontline, iron out issues and manage the ongoing deployment of the reporting outcomes
- Frontline Levels – made up of analysts and client team leads to ensure that insights are reported, understood and actioned.
Defining the problem
We began by having clearly defined and agreed objectives for the analytics project. This was a complex process because there were two parties involved, our client the outsourcer and their client. It was complex because neither party agreed on what the actual sales conversion rates were simply because neither party had a full dataset they trusted to base their decisions upon and both parties had different definitions.
The outsourcer believed particular campaigns received a large proportion of non-sales related calls where customers complained, “My service isn’t working and I can’t get through to customer service”. This issue caused staff to be demotivated when they were rostered on particular campaigns. Our client tried to articulate the size of the issue by reporting on agent post call wrap codes.
The client, on the other hand, did not trust the post call wrap codes the agents entered because some of the transactions were inconsistently recorded. They believed that agents were incorrectly applying codes, either intentionally or unintentionally. Both parties needed a clear source of truth from an independent source about what was occurring within the contact centre.
The best determinant of this source of truth was the call recordings. We commenced our service by gaining an understanding of which calls were potential sales and which were not, by campaign. Analysing a sample of over 18,000 calls, we identified and gained agreement between both parties about:
- Which calls were not sales opportunities and why
- Which issues generated the highest number of calls by product type, area and campaign type.
- Which campaigns generated the lowest opportunities overall (proportionate to the campaign reach)
We found this by manually listening to a large sample of calls to identify standard phrases used during the interaction, programmed our speech engine to identify these phrases, manually checking for results to confirm the findings and then producing graphical representations of the data that both parties could easily read and interpret. This report provided both parties with the first clear and independent evidence of what was occurring in the centre. Once all parties had agreed upon a set of definitions we commenced analysis of the actual sales process.
Sales Analysis – A Blend of Speech and Statistics
With definitions and terms agreed, we then used our 3 week structured delivery process to develop a sales call flow that could predict sales success and be communicated simply to front line agents. Through listening to calls we hypothesized that consumers called this particular client with three basic psychological needs:
- Can I save money, time or effort?
- Which offer is right for me?
- Should I buy this?
Based on our manual analysis we were able to identify precisely what agent behaviours and phrases were used in a successful call and which were not. These phrases were loaded into our phonetic speech analytics engine and the results were then measured against a sample of calls.
Using our findings, we then conducted regression analysis on our model outputs to predict which calls would result in a successful sale and which would not. The statistical findings were then compared with the actual sales results to test the predictive power of our model. Refinements were then made based on our findings.
The phrases were then divided into 11 categories that mirrored the sales process. Reliability analysis indicated that each of the phrases in these categories provided a high level of consistency and reliably measured the category. From this, a sales call flow model was produced that, when used, would increase the propensity for a consumer to purchase for this client and industry.
How was the data actioned?
Finding results was only part of the journey for our solution. Our experience with Speech Analytics and Insights Solutions is that unless reports are actioned, they produce little value to organisations. To ensure our findings resulted in measurable and changeable results our results needed to be communicated and coached on the floor. Our call flow provided a simple and effective structure to provide feedback to agents that would illustrate what they could do better in a specific, measurable and observable way.
We began by measuring daily agent adherence to the call flow for 100% of calls with a next day turnaround. These reports are generated electronically and are displayed by team, agent and across the centre. Weekly ‘coach the coach’ sessions were conducted with team leaders. These sessions enabled Team Leaders to directly question the KnowledgeSpace analytics team about what they found, raise issues they may have coaching agents and input into refining the reports.
An executive insights report about campaign, team and centre effectiveness is also produced weekly. This report is intended to provide our client, the outsourcer and their client with meaningful marketing insights so that future campaign planning can be more effective.
What are the results?
- Over the last three months, our client has increased sales conversion from 21% to 35.5%. This has meant that their sales revenue has increased significantly and the project now makes a profit.
- More importantly, the client has developed a more collaborative relationship with their clients and can now provide value added insights into what consumers were calling about, why they didn’t buy and what campaigns could be designed to convert these prospects. This has resulted in an expansion of their service contract.
- Identify which of the 123 campaigns they managed generated the most interest and why.
- Reduce agent turnover by building a culture of success.
Our client has gone on to become a raving fan of the KnowledgeSpace speech analytics and customer insight solution and has recommended our services to others. We continue to have a close relationship with the client and continue to provide services.