Pricing analytics – disregarded, underused and misunderstood

Last week we heard back from a key client of ours, a listed company – they were happy to confirm that their 50,000€ investment in pricing analytics had already generated 2 MEUR increase in revenue in the matter of months. We were thrilled to hear that we had been able to create a powerful profit lever for them, but simultaneously I could not help but wonder why price optimization still remains as one of the most untapped business opportunities. Price modeling insights help to forecast demand, improve  revenue and strengthen customer satisfaction, but given the technical complexity of the required analysis, unconvincing business benefits and doubted relevance of findings in ever-changing market conditions, it keeps being pushed lower down the companies’ priority lists.

We discussed with our team of data scientists, Arto Sakko and Ville Suvanto, to analyze the reasons why pricing analyses are not much used today. Through our work in various data-intensive industries we have identified three main concerns on data-based (as opposed to cost-based) pricing regardless of company size and clientele. By sharing our views and learnings we hope to debunk the myth of pricing analytics being merely an academic curiosity, and to encourage businesses to take advantage of one of the most underdeveloped chances of success.       

1.    Our data is insufficient or of not good enough quality

Rarely is any organization lucky enough to have an abundance of good-quality data on multiple price points and all possible market scenarios. Sufficient input is obviously a prerequisite and certain level of quality always helps, but it does not have to be perfect – a good data analyst does not get stuck on shortcomings but proactively identifies new and enriches existing sources of data depending on the business problem the company wants the analysis to address.

Developing a mathematically grand and universal model of the real world complexity is problematic, and due to the inevitable element of approximation always present, most statisticians and economists have not even attempted to tackle it. However, by focusing on scientific perfection we risk missing on significant insights. The future success stories are written by those who manage to strike a perfect balance between robust scientific methods and creative sourcing and interpretation of big data.

2.    Recommendations from the analysis are vague and impractical

Delivering understandable and actionable recommendations is the expert team’s responsibility. Back in the days statisticians, consultants and other subject matter experts had the luxury of working behind a curtain, coming out with some cryptography even Alan Turing would have had difficult time understanding and leaving after a hasty high-level explanation. Fortunately, democratization of knowledge has forced the modern day experts to come become transparent and to help the rest of us understand.

The analysis is tricky – the pricing models are never universal, they must always be attributed to certain tolerance limits and they rarely accommodate psychological or other qualitative parameters that are inarguably significant in all socio-economic contexts. These uncertainties can be partly addressed by market simulation, price elasticity and customer surveys, and a good expert analyst will always help you to investigate them further. That’s how it is folks – our world is chaotic and to understand it requires a joint effort.

3.       The analysis becomes soon outdated

The only thing permanent is change, and it would be naïve to claim long-lasting relevance for any business analysis. A litre of milk can change prices every week, train tickets once a year at most. The more volatile the market, elastic the price and larger the quantities sold, the more frequent is the need for remodeling.

The good news is that the once-built pricing models will actually improve with time – once the static analysis parameters have been identified and put in place, full focus can be turned towards the variables, their trends and anomalies. Repeating the analysis with new transaction data is quicker and easier than first time around, and a robust analysis process allows the pricing model to update as frequently as needed. Also when there are clear dependencies – such as sunscreen sales peak simultaneously with beach towels – the same analytics can be applied with high certainty to all interdependent products.