The analytics sector has grown exponentially as terabytes upon terabytes of data become available every day. As technology becomes increasingly efficient, machine learning models have heavily automated data analysis. Is Big Data and Machine Learning enough to make business decisions based on your data?
I would argue, absolutely not – without data expertise, industry knowledge and the willingness and ability to question the results, output from automated models is as meaningless as raw data.
Big Data Isn’t Useful Data
From data collection to results reporting, a data expert understands that collecting and manipulating the right raw data lays the foundation for sound analysis and clear reporting. For example, if you want to explain sales of a supermarket product and you attribute any rise in sales to media campaigns, you are ignoring the effect of all other factors like promotions, display, or the economy. In practice, this can be difficult if the data isn’t clean, complete, or even available. While data experts aren’t magicians, they’ve seen it and can handle it all: missing data, incomplete data, and generally “messy” data.
The Only Constant Is Change
It is important to have continual test and learn approaches imbedded in your analytics function in order to adapt and account for changing market conditions. Forecasts will only go so far and won’t account for any fundamental changes in the market. For example, while the results from a 2013 report might have been robust at the time, it doesn’t account for events such as Brexit that have notable political and economic implications. In addition, each new marketing innovation and campaigns can will have a different effect on your KPIs. The faster your business and markets move, the more essential frequently updating your world view to your marketing strategy.
No Model Is an Island
Although a carefully tuned model can help you understand what is affecting your KPI, it is also important to combine this with other insights. MMM is able to determine what factors affect sales for example, attribution modelling would be able to determine how users are directed to websites. Combining both MMM and attribution would therefore provide you with an overall picture of what is the impact of marketing on the KPI as well as how users are drawn to the website to potentially make a purchase. It is important to look at the full story that the data is trying to tell you, which requires you to look at more than one model or type of analysis.
Sales Reports Can Be Misleading
Sales reports tell you what is happening with your KPI, but not why. Sales reports fall into the fallacy of confusing correlation with causation. As an example from the Journal of Chemical Information and Modelling demonstrates, just because greater fresh lemons imported from Mexico is correlated with lower highway fatalities does not mean that the American Association of State Highway and Transportation Officials should lobby the government to actively pursue greater import of Mexican lemons. Businesses need to take the same level of care to understand what drives their KPIs so that they can optimise their media spends and maximize ROI.
Undoubtedly, there is a great potential to understand what drives KPIs better than ever before given the technologies and amount of data available. However, it is important to have a data expert that understands how to collect, analyse, and translate statistics to plain English in order to derive the most value from your data.
For more information on how we can turn your raw data into actionable insights, feel free to contact us.
By Jonathan Ramirez, Brightblue consulting, London UK