In 2016 we built a customer service analytics proof of concept application. That same year came the AI wave. Artificial intelligence has had some big wins since we made this application in mid 2016. I believe it will completely reshape analytics. Big data + AI + enormous business opportunity makes this inevitable.
Jumping The Curve
In Twelve things I learned from Steve Jobs, Guy Kawasaki talks about jumping the curve. The curve is being in the right place at the right time. In 2017, the right place and time is AI. Using consultants or analytics to arrive at the aforementioned KPIs will soon be replaced by AI figuring it out and adapting on the fly.
So we are jumping the curve, shuttering Scormi-the-app, and starting work on a virtual agent for a different vertical. That said, it was an enlightening exercise…
The idea was to shift the analytics paradigm from reporting random metrics to first setting performance goals, then scoring progress toward those goals. In other words, shift mindset and workflow from mining to refining.
This project came from discussions and questions I had with a support manager and a digital marketing media buyer who were having trouble using their data to move the needle. The support manager worked at a start-up, was new to the role and did not know where to begin. The media buyer worked at a major consumer products company and was flooded with more analytics reports than she knew what to do with. Consequently, those wonderful PowerPoint, Excel and PDF analytics reports sank into the abyss of her 200/day email inbox.
The common denominator to which they both readily admitted: neither one knew which KPIs most mattered and what their values should be.
The light bulb went off; why not build an analytics dashboard with KPIs baked right into it?
If you don’t know what you are looking for before you start measuring things then you will end up tracking random metrics that have no business impact.
Just 7 KPIs
For customer service then, which KPIs matter the most?
After some investigation I found that other companies had performed extensive benchmarking of customer service KPIs. Using the 80/20 rule, it turns out that just seven KPIs account for 80% of a service desk’s performance level. While there were 30 KPIs in total, 7 mattered most.
You can read about these KPIs in the linked article. Suffice it to say that the 7 KPIs most definitely have a clear line of site to net income.
Be clear in your mind about the difference between a KPI and a metric. Think about this phrase for a minute “key performance indicator.” It explicitly says some metrics are key to performance. The rest are not that important.
If you’re in a nfunction other than customer service how can you know which KPIs will most move the needle? I once read a great shortcut for determining this. “My business would implode if this happened.” Whatever this is, is a KPI.
In Win With Web Metrics: Ensure A Clear Line Of Sight To Net Income! well-known digital marketing evangelist Avinash Kaushik expanded on Professor Ken Wong’s work about the need for business activities to have a clear line of sight to net income.
Having a clear line of sight means that you are able to map every metric you report on…every single day directly to the strategic objective of the company.
Building a Web Application
That’s the backstory and rationale for our analytics app. Our goal was to gauge commercial interest in a dashboard built around these seven, core KPIs and explore receptivity to the Mining vs Refining concept.
The semi-lean startup. We wrote code before market validation, which lean startup practitioners caution against. I get that. It’s better to convince a few early adopters to pay for development than it is to potentially build something no one wants. That said, “do you have a demo?” was the number one reply to my prospecting activities.
Building the demo took about 60 days of our time and minuscule monthly hosting charges. It seemed a reasonable compromise versus pitching vaporware.
We settled on a bare bones version of the application that used 4 of the 7 KPIs. This is a screen capture of the dashboard. If you are in support I think you can see they each have an effect on net income.
Development. We set up a VPS on Digital Ocean to host the app and Alex, my co-founder, built it using PHP, MySQL, and AngularJS. We opted to connect the demo to Zendesk as they are the big dog in SMB customer support. It was also very easy to discover which companies use Zendesk as their support platform and to then contact with them to gauge interest.
Algorithms. Since we don’t have a large inflow of support tickets ourselves, we created an algorithm that automatically generated new tickets every day and added some random noise to give the charts some variability. Once everything was working properly we shifted to running the demo as a simulation on our server instead of using the API. It’s faster that way and we don’t have to make changes to the demo if Zendesk alters the API rules. Hat tip Alex for that. As of this writing there are over 28,000 tickets in the database, which are the data underlying the charts above.
Try it. Anyone can explore the demo (update: retired effective 10/20/2017). In the settings section there are sliders to set KPI goals. We added a default, per agent daily support budget that you can adjust to your own circumstances. For simplicity, Cost per Resolution is based on a single support channel (telephone). In a production environment that chart would be broken down by channel, such as email, chat, phone, social, etc.. With the charts above being nearly real-time one could quickly see the individual and holistic impact of any changes. For example, the impact of adding a virtual agent (a chatbot) would quickly flow though all KPIs.
We also have a live version that you can try and even use for your own reporting if you are on Zendesk. Contact me if you think that app has potential in your organization.
Scormi app settings section
Update 11-2017: AI can easily identify which KPIs most move the needle and this is starting to be built into analytics software. Ultimately, however, analytics software as we now know it is doomed. With AI linked to automated business process the system will make its own adjustments without the need for an analyst.