Google and Facebook make billions each year by analysing their users’ data and then selling it back to consumers and businesses in various forms. Whilst not necessarily expensive, the data is often useless and tends to have little relevance to your own business needs.
The alternative, of course, is to look at engaging a management consultant to spend time analysing your specific business. But this can quickly become expensive and difficult to demonstrate any measurable return on investment.
So we decided to try and do this ourselves using whatever easily accessible data we had available to see what it could tell us. And it turned out that we had a lot of data – mainly in the form of emails. With the average employee sending and receiving around 130 emails a day our small business of 10 people, was generating over 6,500 emails a week. That’s over 3 million emails in the last 10 years and that’s a huge amount of information about our customers (and potential customers) which was simply going unrecorded or even ignored.
In order to try and filter, sort and organise the data, we adopted a 3 stage process. Aggregate our data, analyse it and then act on the findings.
1. Aggregate
To begin analysing your emails, you first need to aggregate all of the data that you have available in one place. This can be tricky as, one of the major problems with email is that it is a personal tool and key information often end up trapped in an individual’s email account – a problem exacerbated further when employees leave a business.
Several companies have tried to solve this problem by offering shared inbox solutions. And whilst these can definitely help, they often require users to change business practices or to manually organise the data – which can be difficult to implement and maintain. It will also likely ignore any emails that were exchanged prior to the date of implementation.
A better solution is to pool all of your company’s digital information, such as emails, phone calls, SMS, tweets etc into one single database and then apply machine learning to extract information. Indeed, some of the best examples of what can be achieved when you pool communications from your entire company can be demonstrated by the Enron corpus. The Enron dataset includes over 600,000 emails and phone calls from the entire company, not just one person’s email. We learn a lot more about what was actually going on at Enron by looking at everyone’s emails rather than just those of one individual.
2. Analyse
Once you have aggregated your data, you then need to decide what you are interested in analysing. This will ultimately depend on what your business might find useful, but here are some examples of the analysis that we did:
Frequency Analysis
This usually involves analysing the frequency of emails sent and received and can pinpoint whether there are any particular times of day or days of the week when activity spikes. In our own data set we could quickly see that the majority of our leads came at the end of the week, on a Friday or over the weekend.
Project Analysis
Identifying emails that pertain to a specific project or action can provide a deeper understanding of the timeline from start to finish to identify pick up common patterns. For example, we noticed that in relation to support renewals, if we had communicated with the customer in the 3 months prior to their support renewal date, they were 47% more likely to renew than otherwise.
Sentiment Analysis
Sentiment analysis is type of natural language processing (NLP) and can be used to determine the tone of text or speech. By analysing emails for positive or negative keywords and counting the relative frequency of these words, an email can be scored and a decision taking as to whether it contains positive or negative content. Whilst we don’t include intercompany emails in our own data set, a number of people have sought to apply sentiment analysis to the ENRON data set to analyse staff morale before, during an after the scandal.
Word Count Analysis
A simple measure of productivity can be undertaken by word count analysis. In certain situations a brief email might suffice whilst in others a more detailed response is necessary. Neither one is right or wrong and will depend on your own business, but applied to our own data set, we found that our potential leads responded more positively to brief email communications than ones containing excessive detail.
3. Use Your Analysis
So what did we learn from this exercise and what changes have we made to our business as a result of it?
We decided to employ a part time intern to respond to and manage our leads over the weekend. This was a cost effective way of ensuring that each and every enquiry was responded to at a time when our offices were usually closed and it helped ensure that we didn’t lose a potential customer.
We also adjusted our pay per click (PPC) campaigns to ensure that the majority of our budget was focussed on capturing customers during these key time periods. This had an immediate effect and in the first month we saw an uplift in the number of conversions by over 50%.
We began initiating conversations during the 3 month ‘sweet spot’ prior to customer renewal dates. Our support renewal rate has gone up by by over 45% so far. We still feel that there is more that we can be doing here so we expect this to increase as we hone our proposition.
We further investigated those customers that had engaged in the higher frequency of emails and noticed that these were often correlated to customers who were trying to find a suitable time for a demo. On the back of this we decided to invest in a new calendar tool to allow our customers to book demos when they suited them without the need for us to get involved. This has significantly improved our demo request conversion rate.
Conclusion
There is so much data hidden in our email inboxes and the emails or our colleagues that goes untapped. By applying some simple analytics we were able to quickly identify areas where things were going well and not so well and implement or change our strategy accordingly. And we don’t intend of stopping there. We will continue to find new angles and ways to analyse our data to make sure that we are making the most of the information that we already have, rather than paying someone else to tell us what we already know.
We also now routinely record and transcribe our phone calls with customer as well and aggregate this with our emails. It gives us valuable context around the emails as nearly 40% of our communications with customers are done on the phone and it was vital that we didn’t lose this valuable data resource. In the same way that we have been able to analyse our historic emails for insights we are able to do this for phone calls. More on what we found from our phone calls to come in a future blog post.
So before you start looking outside your company for ways of improving your business, look no further than your own data. It’s all there for the taking.
If you have carried out a similar task or want some help getting started with carrying out this exercise, then feel free to get in touch with me at francesca@threads.cloud.