BUS2002 Customer Analytics: Conjoint Analysis Project Report

Section A: Methodology

This report summarizes the findings from a conjoint analysis conducted for Homer and Co. The purpose of the conjoint method is to measure how customers trade off between different product attributes when they make choices. Rather than asking customers to state which attribute is important, conjoint analysis presents realistic product bundles (combinations of attribute levels) and collects ratings or choices. From those ratings, we estimate “part-worth” utilities for every attribute level. Those part-worths show how much utility (value) customers place on each level. By comparing the range of part-worths across attributes, we can calculate the relative importance of each attribute. Finally, the part-worths are used in conjunction with a simple logit formula to estimate purchase probabilities for candidate bundles and competing products.

From a manager’s perspective, the strength of this method is that it forces respondents to reveal trade-offs — for example, whether they prefer a cheaper product or faster installation — in a way that resembles real buying decisions. It provides an interpretable numeric result (part-worths and importances) that can be directly applied to product design and pricing decisions. The method is appropriate here because Homer and Co. must decide a new product bundle from a small set of clear attributes (life expectancy, installation speed, shingle aesthetics, price) and because the dataset contains respondent ratings for a fixed set of nine bundles. The main weaknesses to be aware of are: (1) the sample used in this mini-case is small (only 20 respondents) so findings are indicative rather than definitive, (2) the analysis relies on the attributes and levels chosen — if an important attribute was not included, conclusions may be incomplete, and (3) linear additive utility and the simple logit assumption may not capture very complex preference interactions. Despite these limitations, the conjoint provides a clear, managerially useful ranking and tells us how to adjust levels and prices to achieve the best market response.

Section B: Interpretation

The conjoint analysis helps to uncover more about customer preferences for roofing shingles in Springfield, based on the estimated part-worth utilities for each attribute level across the 20 respondents. The average part-worth utilities (Appendix A), rounded to two decimal places, indicate that for life expectancy, medium resistance has a utility of -297.30 and high resistance has a utility of -21.90, relative to the baseline of low resistance at 0. For speed of installation, the averages are 1.2 weeks at -13.65 and 1.5 weeks at -22.67, compared to the baseline of 1 week at 0. Shingle aesthetics with minimal imperfections average -4.52, compared to no imperfections at 0. For prices of $9 and $10,the averages are -7.75 and -9.97, respectively, relative to the $8 baseline of 0. The average intercept, representing the baseline rating, is approximately 79.53.

These negative values indicate that respondents generally prefer the baseline levels: low resistance for life expectancy, the fastest installation time, flawless aesthetics, and the lowest price. This suggests that deviations from these baselines reduce overall preference, possibly because higher resistance or slower speeds are seen as unnecessary complications or added burdens in a market facing cost pressures and competition. Critically, the preference for low resistance in life expectancy is counterintuitive, as one might expect customers to value greater durability; however, this could reflect perceptions that higher resistance shingles are over-engineered, heavier, or not worth the trade-offs in price or installation ease, especially in Springfield, where economic factors like rising raw materials and COVID-19 impacts may prioritize affordability over long-term features.

The relative importance of attributes (Appendix B), calculated from the ranges of part-worth utilities and averaged across respondents, highlights what drives decision-making. Life expectancy accounts for approximately 37.67% of the importance, speed of installation for 35.91%, price for 17.48%, and shingle aesthetics for 8.94%. This means that speed and life expectancy dominate preferences, together accounting for over 73% of the decision weight, while price matters moderately and aesthetics have a negligible impact. From a managerial perspective, this implies that investments in faster installation processes or communicating durability could yield significant gains in customer appeal, whereas enhancing aesthetics might only marginally improve perceptions. However, the close importance between speed and life expectancy suggests a balanced approach is needed, as overlooking either could alienate parts of the market.

Segment differences further deepen the understanding, with respondents grouped into three clusters based on their preferences. Averaging importances within segments yields: Segment 1 (with many respondents) prioritizes speed at approximately 50.2%, life expectancy at 24.8%, price at 15.1%, and aesthetics at 9.9%; Segment 2 emphasizes life expectancy at 49.3%, price at 18.5%, speed at 26.1%, and aesthetics at 6.1%; Segment 3 values price at 22.1%, life expectancy at 33.3%, speed at 30.9%, and aesthetics at 13.7%. These distinctions are meaningful and suggest diverse customer needs: Segment 1 appears speed-oriented, likely representing time-sensitive users such as contractors who minimize job disruptions; Segment 2 focuses on durability, fitting long-term homeowners willing to trade convenience for longevity amid rising costs; Segment 3 is more balanced, with higher weight on price and aesthetics, possibly value-seekers who appreciate affordable, visually appealing options without extreme trade-offs.

The bar heights are different, as seen in the Appendix F, between segments, such as the green bars of Segment 3, which are more homogeneous, and this balance suggests that a segmentation strategy might be effective. Homer and Co. might develop variants: a quick-install product for Segment 1, a robust one for Segment 2, and a budget-friendly, aesthetically focused option for Segment 3. Critically, ignoring these differences risks a one-size-fits-all approach that underperforms in a competitive market.

Bundle preferences and predicted market shares provide actionable insights into product configurations. Among the nine candidate bundles, the average purchase probabilities (Appendix E) across respondents show Bundle 1 as the strongest, at approximately 0.89 (89%), followed by Bundle 7 at 0.53 and Bundle 2 at 0.50, with the others being significantly lower. Bundle 1, which offers low resistance, a life expectancy of 1 week, no imperfections in aesthetics, and a price of $10, dominates even when split by segments. This is surprising, as it combines the highest price with the baseline (low) resistance, yet the data indicate that the fastest speed and perfect aesthetics outweigh these drawbacks for many.

It suggests that customers place a high value on convenience and appearance, potentially viewing low resistance as sufficient for local needs rather than requiring premium durability. Appendix H, depicting purchase probabilities by bundle per respondent, illustrates this variation: the blue sections (Bundle 1) often form the largest portion of the stacked bars for most respondents, confirming its broad appeal; however, some respondents favour other bundles, such as the black (Bundle 7), highlighting individual differences.

When considering competitors, Appendix G shows average utilities with Flanders the Neighbour highest at around 52, followed by Skinner the Great at 49, Moe's the Best at 48, and Krusty Unbreakable at 46. This positions Flanders as the toughest rival, likely due to its alignment with preferences for low cost and quick features, suggesting Homer and Co.'s bundles must compete on speed to erode its share.