How do we Better Solve Analytics Problems?

By Rajneet Kaur, Data Scientist

How do I start?

What questions do I answer with data?

What components should the solution framework comprise?

In my experience as an analytics manager/mentor, I often get this question. We handle a quantum of new projects in the analytics and data science space. Problem definition and solution development are key ingredients of being a consultant. Structuring the problem definition phase is critical to project success but may seem like a creative process.

Start with a framework,

my marketing professor had said, when a student inquired on how a case study should begin. A plethora of projects, clients and roles later, I still find this advice relevant.

The start is often the toughest and business frameworks can help make project kick-offs and ad-hoc analytics easier. A lot can be achieved by starting with one or more frameworks and then customizing them. There is a high chance that your clients are aware of these and hence consuming the analytical piece is easier. Story-telling also becomes easier.

While a detailed set can be viewed at this link, here is a starting laundry list of frameworks that can be used:

1. Analytics: RFM, Hypothesis testing, Ishikawa/Fishbone (Cause & Effect)
2. Marketing: AIDA(Customer Buying Journey), 4P, PLC, CLV, Segmentation : Any 2*2 or 3*3 matrix like BCG, Ansoff matrix and others like Pareto
3. General Business: SWOT (Internal), PESTLE (External), Porter’s 5/6 forces, 3Cs

While Ishikawa and 4P can be leveraged for root-cause, cohort/life-cycle analysis can be done well with AIDA and PLC. Optimization is another common analytics area and Pareto and Quadrant analysis can be very helpful.

Let's use some examples to demonstrate 4P usage, among the simplest, and my favorite root-cause framework. This framework can be effectively leveraged for corrective actions for weekly or monthly business health monitoring, like in case 1 :

We find from the above analysis, that the ROAS of overall spends has reduced over the years. On further analysis we find that offline channel has decreasing contribution. Basis these customer behavior patterns, your organization decides to divert marketing budget from offline to online. Offline spends have to be reduced by a third and you need to decide how to still keep the lift maximum possible.

We use Pareto next - Pareto is among the most intuitive frameworks, with focus on the concept of 'Vital Few'. Though it is also called the 80-20 rule, i.e. 20% of drivers should cover 80% of the values, feel free to- change the numeric thresholds.

A few weeks later we decide to optimize spends further. Let's say that we find that different locations have differing response to offline promotions. Let's see how we fine-tune every marketing dollar to the maximum using quadrant analysis:

One or more business frameworks can be used in conjunction and further tuned as per business drivers for effective output. From easy to the most complicated, business frameworks can be leveraged in analytics and in other walks of life. Some of my earliest interviews were cracked through a BCG as part of a self-analytics, with my 'Personal life' as a star, 'academics' as my cash cow, My 'clean' cupboard (sarcastically of-course!) as my dog, and professional career as my question mark.

Do write back about your views, experiences or challenges in problem solving for analytics and data science projects.

 
Bio: Rajneet Kaur is a passionate and result-driven marketing and data science executive with 6 years of business and technology experience.

Original. Reposted with permission.

Related:

• How to Measure Foot Traffic Using Data Analytics
• The Augmented Scientist Part 1: Practical Application Machine Learning in Classification of SEM Images
• How to Sell Your Boss on the Need for Data Analytics