Implementing and integrating advanced marketing analytics is one of the toughest decisions facing many businesses over the past few years. The first movers are well and truly gone; marketing analytics is no longer innovative, and it is no longer a comparative advantage.
The acceptance of data-driven marketing has spread across the industry, but businesses still face the question of whether to truly become a data-driven business, and many are paralysed by the challenge that lies ahead. So what are the challenges that face businesses and why do they halt these initiatives moving forward?
1. Buy vs Build
Often this is a question that is asked after a business decides to implement marketing analytics, however can end up being a barrier if the wrong choice is made. There is no general rule as to whether you should buy or build as it completely depends on the business objectives, but often the decision is not well thought out. If you are thinking of building inhouse, you must consider a few things:
- Do we have capacity? How much will this drain our resources, and divert these resources away from profitable activities?
- What are the competing projects currently being considered?
- Do we have the skills inhouse for this to be successful?
- If successful, is there any key man risk associated with the analytics build?
Of course, buying also has its pitfalls:
- Who is the right supplier? Are they capable of meeting our business objectives?
- Is what I am buying really suited to my needs?
- Will this continue to work once the consultants leave and implement the system?
In many cases, the business case is rejected based on the wrong choice of buy vs build. Considering this choice carefully is vital to getting the project off the ground.
2. Data Quality
Often data quality is not as good as you think it is and is a result of incremental changes as data models evolve. Businesses evolve and so too does the way data is collected, so on first inspection data quality can be underwhelming and even disheartening.
However, this neglect should not stop action but rather encourage it. Believing that marketing analytics is too hard due to data quality is only going to make the inevitable task harder down the track. It is like not going to the gym because you are unfit – sure, it will be painful at the start, but putting it off only exacerbates the problem.
3. Misinterpretation of Statistics
Just because someone is predicted as having a high probability of doing something is no guarantee they will actually do it. Any stochastic model is based off imperfect information. Furthermore, rarely is any model calibrated when made operational, often resulting in inflated probability scores from the training of a model.
In addition, humans are bad at understanding probability. We gamble, and make poor investment decisions. Giving non-technical staff a probability to work with has no meaning and only makes them disengage with the system. Instead, providing a rank (i.e. "customer X is in the top 5% of churn risk) and a prescriptive analytic (i.e. "Offer customer X a special discount") will provide non-technical staff with actionable intelligence.
4. Traumatic Experiences - Everyone has a horror story
"Does it work?" "We tried this before and it didn't work" "Employee X built a system that did this but then they left and it no longer works". These are all too familiar and highlight a lack of integration, and maintenance. Automation does not mean set and forget, it just means you do not need to reinvent the wheel every time.
A client of ours recently told us that she feels that often analytics is seen as "optional" and not a vital part of the process. This lack of attention is causing these projects to slip through the cracks and leave a huge opportunity begging. Analytics, even weak analytics based on minimal data has huge upside if done at scale.
Given most people agree analytics is the future in marketing, why are we putting this off? We are all risk averse, and we don't want to be the ones to stick our neck out in case it fails. However, where there is risk there is also reward. Making this step towards marketing analytics sends a powerful message across the company that you are now truly data driven.
5. Lack of appreciation for the process
Quick set up gimmicks and the promise of easy use always have a dirty little secret. Unfortunately, the secret that sales does not want you to know is that their demo of their slick dashboard or the ease in which they can build a predictive model using point and click software has a fatal flaw: that was the easy part. The hard part is actually getting there, building the right analytics environment that can tap into all of your available data sources is challenging and time consuming.
Consultants flogging off the shelf solutions are not much better. Sure, they hired some big guns to build some fancy models, but it is trained on someone else’s data. There is no guarantee that these models, regardless of how strong they were, will run effectively on your data. Models need to be trained, tested and validated on as close to the real production environment as possible.
Deploying advanced predictive analytics into a business production environment is a problem, which has been solved before, and will continue to be solved by many businesses in the near future. There are well defined processes in academic and industry literature which even explain how to do this, yet the marketing industry still wants to believe it can't be done. But we know it can, we have seen it done, so we need to implement.
The time is now.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Follow Kristen on LinkedIn: www.linkedin.com/in/kristenturnbullcoredata
This article was written by CoreData's data scientist, Mark Brackenrig, and originally published on CoreData's insights blog: http://www.coredata.com.au/blog/
