Data Alchemy Case Studies

Videogame Revenue Model

Brief Overview:

Game makers, aiming to optimize their business models, often employ conjoint analysis. This approach is designed to tap into the gamers' choice process, providing insights into how different game modes, like "Capture the Flag," should be packaged and priced — as part of the base game, downloadable content, or via subscription.

Problem Highlight:

Standard conjoint analysis tools were not programmed to foresee issues arising from a game mode appearing in multiple formats within the same analysis. This oversight led to substantial problems with the statistical model, inducing multicollinearity, which in turn resulted in inconsistent and incoherent outputs. Essentially, the model was not correctly interpreting the data from gamers who encountered "Capture the Flag" in various purchase options.

Data Alchemy Solution Snapshot:

In response to this challenge, the Data Alchemy (DA) approach involved developing an advanced model that could identify "Capture the Flag"—and all other game modes presented in the analysis—as a single entity, regardless of its delivery method. A new multiplicative parameter was introduced to accurately measure how preferences for each game mode varied depending on whether it was part of the base game, a DLC, or a subscription. This allowed for a nuanced understanding of gamer preferences across different delivery platforms.

Results:

The Data Alchemy model provided decision-makers with a coherent and consistent data set, revealing the true valuation of each game mode across all purchase methods. This clarity empowered them to make profit-optimal decisions, aligning their offerings with the preferences of their target market.

Conclusion:

Unlike the rigid, one-size-fits-all approach of standard conjoint analysis, the Data Alchemy method adapted to the complexity of consumer choice. It harnessed this complexity to deliver a refined analytical strategy, one that effectively decoded the nuances of consumer preferences. Consequently, the Data Alchemy approach enabled game developers to strategically place their game modes in the market, optimizing engagement and profitability.

Fundraiser Snack - Price Pack Architecture

Brief Overview:

A snack company facing the need to adjust prices amidst rising costs sought to determine the best pricing strategy for their range of products, which vary primarily by flavor and pack size.

Problem Highlight:

Using a standard conjoint analysis, the company faced a significant issue: pack size and price were so intertwined in the analysis that the resulting model was at risk of suggesting impractical pricing strategies, such as recommending lower prices for bulk packs than individual packs, which could undermine the credibility of the research.

Data Alchemy Solution Snapshot:

The Price Pack Architecture-conjoint (PPA-conjoint) method was introduced to tackle this. It took into account economic principles that recognize not only the value of more product but also the diminishing utility of larger pack sizes due to factors like increased cost, storage concerns, and the desire for variety.

Results:

By incorporating these consumer considerations into the model, PPA-conjoint delivered a sensible pricing strategy that maintained the integrity of the company's pricing structure. The approach provided clear, logical recommendations that made sense for both the consumer and the company's bottom line.

Conclusion:

In contrast to the rigid standard conjoint analysis, which could have led to bizarre and counterintuitive pricing suggestions, the PPA-conjoint model embraced the complexity of consumer behavior. It led to a trustworthy strategy that enabled the company to adjust prices with confidence, knowing that their decisions were based on a solid understanding of their customers' preferences and needs.

Best Practice Disaster Averted

Brief Overview:

A well-regarded motorized vehicle manufacturer was encountering illogical forecast results from their choice modeling, which not only jeopardized the reliability of the forecasts but also threatened the research team's credibility. The issue was traced to an unexpectedly high correlation between two unrelated vehicle features.

Problem Highlight:

The industry-standard Inverse-Wishart (IW) distribution used in their hierarchical Bayesian choice model was identified as the culprit. Its intrinsic limitations were confounding variance and correlations, leading to irrational results. Proceeding with these findings risked eroding the team's credibility and could result in misguided strategic decisions. The outrageous correlations were a function of the model not the data.

Data Alchemy Solution Snapshot:

By substituting the LKJ distribution for the conventional IW in their Bayesian model, the research team eliminated a significant source of unexplainable anomalies, ensuring that variance and correlations are accurately reflected in their data. This careful refinement of their statistical approach yields forecasts that are not only more reliable but also more defensible in strategic decision-making.

Results:

The switch to LKJ distribution recalibrated the model's correlation estimates, bringing them back to realistic levels. This correction safeguarded the research team's reputation, as they were able to provide trustworthy and defendable forecasts.

Conclusion:

Persisting with outdated methodologies, much like using an old, worn map, can lead businesses astray. The switch from IW to LKJ is akin to adopting cutting-edge navigation technologies over traditional maps. This not only enhanced the precision of the model but also reinforced the research team's standing as a credible and insightful contributor to the company's strategic planning. Updating to the latest analytical methods isn't just about accuracy; it's about sustaining trust and confidence in a competitive industry.