No Products in the Cart
A mosaic of AI generated images depicting Analytics dashboards, charts, reports and analyses
Oh boy, where to even begin with this "analytics" thing? First off, let's define it as the process of using data to make informed decisions. Sounds simple enough, right? It’s not. It's a never-ending cycle of identifying a business need, collecting data, cleaning and preparing that data, analyzing it, making sense of it all, discovering insights, applying those insights to business decisions, only to realize your new business need has already changed because your decisions are having an effect - and that's on a good day. Add to that conflicts between the "business" and the "technical" folks, the politics of decision making. Oh, the drama.
Rinse and repeat.
Let's start at the beginning. The first step is identifying a business need. Do I want more sales or leads? Conversions or brand awareness? Process optimization or fulfillment speed? Not as easy as it sounds, considering corporate politics will always get at least a secondary role in this. Now add to that the complexities of coordinating the interpretation of KPIs across business functions. Is a sale the same as an acquisition? Is an order fulfilled when it’s delivered or accepted?
With a bit of luck, you've identified your business need and the associated KPIs that provide a glance at how that need is being served. If you have selected more than 10, you have a problem: KPI stands for Key Performance Indicator, can you really have more than a few Key indicators? Could you be confusing it with a metric? But that's for another post. Let's asume you have a few KPIs that describe the issue you're addressing. Now what?
Next up, data collection. Oh, you mean like trying to wrangle a group of wild animals? Yeah, precisely that. Data is messy. It comes from different sources, in different formats and frequencies, with different interpretations. Raw data quality is often a reflection of the way it was gathered, rather than what it was intended to portrait. Meaning: it’s freaking hard. The sexy term these days is data wrangling – do yourself a favor, Google the definition of wrangling, chuckle and then come back.
Enter data engineers. To many, a consultative role with the key function of converting raw data into usable information that resembles the way a business behaves. The starting point for a data engineer must always be the business. They are the first to interpret how business logic is captured in data and hence its sources and nuances, how to display it, aggregate it, drill it. Once sources and business logic are identified, it's time to get your hands dirty.
This is where you get to clean, transform, and prepare your data for analysis. In a world where data is created everywhere, all the time, you can no longer afford to work with snapshots. You work with data pipelines. We’re not getting into detail on that on this post, let the analogy sink in and move on once you’re ready.
Now we get to what many call the "good stuff" - data analysis. Enter data scientists and data analysts. Enter fancy statistical and Machine Learning techniques to extract insights from the data. Only not so much. Even with all the prior work to get data “ready for analysis”, the various types of models and their requirements are rarely plug and play to anything. There’s still a ton of data cleaning and preparation to be done. Clean out outliers, interpret blanks and N/As…some say that a data scientist will typically spend 60-80% of their time just data massaging till it’s good to, well, try just one approach over hundreds possible. Long story short, the data scientist figured out a format that works, the data engineer puts it together in an easier structure, the data scientist runs and fine tunes models (rinse & repeat a few times) and you finally come up with something that can actually help make a decision. Metrics that can be associated with those KPIs you identified at the very beginning.
Now the cherry on top: decision making. Enter the CEO, takes a look at the model, provides some guidance, gets a polished version, runs some simulations and voilà! You've made an objective, non-controversial, comprehensive, data-driven decision.
NOT!
In reality, enter BI developers, graphic designers, business analysts, product managers, and more. All a merry band of caffeine-soaked multi-disciplinary professionals that can actually convert all that beautiful math and stats into a coherent insight. This is, a C-level digestible one-liner, maybe a bar chart too, that, after the political dust has settled, tells decision makers what they wanted to hear in the first place.
All is not lost. Enter the analytics translator to save the day. The unicorn expected to convert a science puzzle into toddler wording with cartoons. A Master of Politics, diplomacy, engineering, math, stats, design, product development, astrology, tarot and tea-leaf reading, the analytics translator can often soften up the tensions between functional areas and convey an objective output of all the efforts made before. Out of dynamic dashboards, Machine Learning models, advanced statistical inference, state-of-the-art cloud data engineering, a 3-slide PowerPoint presentation is born. And then maybe, just maybe, a decision is made through objective data interpretation and robust logic.
