Reports, Analytics, Insights, Recommendations…What does an OEM truly need?
OEMs have a unique set of discovery needs, just as every industry with its own language, jargon & terminology does.
The good news is that these discovery needs are fundamental & have seldom changed since the advent of modern technology. Take installed base data as an example – all you want to do is categorize your accounts, drill down from a 40K feet account view to a 10 feet line-item level. You want to find outliers or see how revenue trends over time. But when you dig deeper, what you really want to do is understand how you can exceed your sales target or closeout service requests successfully in the given time and act on it seamlessly…it’s that simple. Data & tools are just a means to an end. With the right data, any tool should be able to get you to the finish line, right…?
The challenge with basic Reporting
Think of a fitness tracker for a moment – I have owned many in the past, and one of my earlier trackers would simply report my step count. Just pure raw data. You walked & the steps would show up immediately on my dashboard. However, it could not distinguish a slow bike ride vs. a brisk walk & would mistake a steep climb in my neighborhood with a flight of stairs.
Basic reporting in the OEM world follows the same logic. Data flows in from so many directions (think tools, people), and in so many shapes (think incomplete, dirty, unreliable data) and in volumes that run into millions of records, dating back decades. These challenges are characteristic of transactional data.
Correcting transactional data requires changing behavior at the point of data generation, which in essence implies changing human behavior. Despite all the emails one sends out, process one sets up, and education that happens, you will still find missing location addresses, orders reconciled incorrectly, missing warranty expiration dates, incomplete contact details, amongst other things.
So, you end up doing the next best thing – put a reporting tool in place. It’s highly likely that your CRM or ERP ships with one. The only problem here – these are reporting tools at best, isolated from each other & built to serve transactional data. They merely aggregate data using “joins”, “queries,” & “languages” that would make an IT engineer out of you. They don’t offer insights – insights that should just come naturally. The reporting tools follow what’s known in the technology world as – garbage in, garbage out. Sure, there are dazzling charts, beautifully animated to load when you fire that dashboard, but what good is this data visualized with fancy charts if it is unreliable?
Remember my fitness tracker- I had to start walking slowly & deliberately so my fitness tracker could record my steps right. Needless, to say I didn’t change my behavior but promptly tossed the tracker away.
Note: Let me also go on record & say that if your data is crystal clear, then a robust, interconnected reporting tool will get the job done. What are the chances of an OEM data being clean to begin with…?
Where does Analytics fit into the picture?
Analytics is the art of making the static report make sense with business rules, definitions & exceptions. Consider it a step above basic reporting, which I consider mostly science because it’s already refined & optimized. Analytics means connecting the dots & understanding which accounts & locations are tied together, which product segment needs an exception, which hundred records are redundant or duplicates & which product line is no longer actively serviced. With the right analytics, you could onboard a new sales employee or a service team member in rapid time & make your customer teams much smarter & self-reliant.
There’s always a ‘but’ so here it goes – analytics requires dedicated time & effort. More importantly, it requires domain expertise & in-depth knowledge of everything that transpired in your sales history. This is where the vast majority of powerful BI tools, CRMs with in-built dashboards, or the ‘machine-learning driven; always learning’ tools will fail right at the door. They expect you to do the heavy lifting. They expect you to inform the tool about your business model & nuances which if not purpose-built takes a lot of time and resources, clean your own data first which is a challenging project, fill in the gaps, establish relationships & supervise the machine.
There’s no such thing as a free lunch you can eat, and there is no such thing as a self-learning tool that you can practically use. However, if someone tells you a machine can learn everything, they are absolutely right – but remember, the machine will learn what’s fed to it. Again, the same principle of garbage-in, garbage-out applies to this learning too.
On a practical, day to day basis, this is how it would play out. Your organization would end up asking you to spend time fixing bad data, effectively making you a part-time data analyst, or if you are fortunate, they will end up hiring data analyst(s). A BI tool or reporting tool that you already use will be re-purposed for analytics. The logic is – if it comes with the box, why not make use of it. Plus, it says it can do machine learning to help you get faster, efficient, smarter, etc. So, there’s a strong case for you & your teams to focus on enabling this machine to learn. Now, you will spend countless hours writing business rules, defining exceptions as well as correcting the relationship between equipment, locations, parts, components, services, warranties, and contracts. You will also work with implementation teams, IT teams, Vendor teams, training teams to feed the machine. And every step of the way, just when you think you’ve got it all figured out, the machine would throw a curve-ball putting you right back to work. Ask the vendor & they’d blame it on the machine learning model, precision, accuracy, and other technical jargon, essentially blaming the human who sits between the keyboard & the computer. I’ve seen more employee burn-out because of bad data & attempts to fix that data than due to hitting a rough patch at work. There’s just something that feels inherently wrong in cleaning what should’ve been done right in the first place. All of this doesn’t even get to the main purpose of the exercise – actually leveraging the analytics to drive your business and sales efforts. Even when companies get close enough to have valuable analytics, sales teams don’t act on it and enable it to drive growth, profitability via aftermarket transactions, and revenue across your installed base. What if you could have a solution that not only automated analytics purpose-built for OEMs but also was built for sales action and results?
In my fitness tracker example, I eventually moved onto another tracker, which could figure out a jog from a walk, a slope from a staircase. It required me to inform the tracker before I started an activity so it knew what to expect, time and again go in and correct its mistakes, and did it’s best to tell me which days I was making healthy choices, and which days I wasn’t. It mostly got it right but made me wonder why won’t this technology just work. I know I am brisk walking; my neighbors know I am brisk walking; then why won’t this tech simply understand this? (face-palm).
Insights – the holy grail of investigation
Insight, in my opinion, is a transformation beyond being art and science bordering onto intuition. An insight is derived from the very mental model that you operate in. Both you & I might look at the same data, the same graph, the same number, or even the same painting & walk away with two different interpretations, and two entirely different courses of actions. And therein lies the key – an insight is always contextual, and it should lead to action. The insight could be a customer is feeling neglected & you need to reach out to them. Or an insight that a customer is overwhelmed and you need to take a step back. Similarly, there could be insights into an under-performing extended warranty segment or a part that is cannibalizing another service segment.
If you visualize the path carefully, your Insights decide your priorities which in turn lead to actions & outcomes. Now imagine, if there were a faster means to get to those insights, a solution that provides recommendations for you to quickly reach the finish line or start asking more questions such as what’s going on with this account. This solution can predict as well as explain those predictions so you could prioritize them. Wouldn’t you be smarter & efficient, spending time where it matters, and on those whom it matters?
Now, let me make something clear – no machine or algorithm can ever replicate your intuition nor replicate the mental model you operate in, including those of your customers. But the right solution can provide recommendations that can shorten your path to those insights – a solution that is driven by a subject matter expert with an in-depth understanding of your industry & understanding of data reliability issues, aided by the right tool & the right machine learning.
This subject matter expert would know what rules to set up, relationships to be defined & create a process for a good “rinse, scrub & clean” of your data. They would draw upon the benefit of having worked with countless OEM teams such as yours. Finally, close the loop with a learning machine optimized for OEMs & you get the right blend of a purpose-built tool + subject matter expert + learning machine that can deliver what could be the closest thing to the insights that you need. So next time, when you do a quarterly plan, you won’t be staring at data aggregation as the first step – you’d already have the most optimal actions laid out for you to achieve the quarterly target. Or when you are on a call with a customer, you wouldn’t have to fetch information from multiple systems, you’d already be looking at a comprehensive Customer 360 reassuring your customer with your intimate knowledge of their set up.
I go back to my fitness tracker example again – I ended up buying a third tracker with built-in insights. It knew my height, weight & BMI & had already calculated the most appropriate weight for my body type. The next time when I had logged more than 10,000 steps at a customer location & had slept for less than 5 hours because of a red-eye home, the tracker sent me a cautionary message on my phone. It predicted – given the activity & sleep cycle & my weight goals, I would end up binging on fatty foods and high-sodium foods today. It recommended that I up my water intake & helped me increase my hydration goal for the day, along with a reminder to stop coffee by 3 pm so I could get to bed on time. I religiously followed those insights because it made me a better, efficient person & made me realize something I didn’t know I needed.
To sum all of this, you as an OEM, need actionable insights that work out-of-the-box` & you needed them yesterday. It is possible only with the right combination of purpose-built tools, human supervision & machine learning. Anyone promising to solve this for you without all the three is just deluding themself as well as you.