Tools owerful ls P t Mode Conjoin

Tools owerful ls – P t Mode Conjoin e Sales aximiz to M

COURSE OBJECTIVES This course is designed to: • Build an understanding of what conjoint models are, what they can do and how they are used • Know what options are available and how to evaluate the most appropriate conjoint approach • Share real-life conjoint and simulation examples in Home Improvements and other categories This is not a course in: • Survey design • Mathematical or technical aspects of conjoint • Forecasting 2

What is conjoint? • A method of understanding and predicting how people select one product or service over another • Uses “trade-off” exercises that require people to choose or hypothetical offerings over alternatives • A way to measure the impact of price and features on offering selection • The basis for a “simulation” model that can estimate offering adoption and revenue 3

What does conjoint help us do? • Gauge the importance of price and various product/service features • Determine sensitivity to price and elasticity with other features • Predict uptake given specific features and price points • Identify optimal feature and price configurations • Model impact of service-level changes • Anticipate competitor responses 4

Conjoint can be used to test: • • Entirely new product or service concepts prior to launch New or additional features of existing products/services Pricing strategies Branding and positioning strategies § New brands and JVs § Repositioning of existing brands • Discontinuance of a product, service, line or brand § Where will the business go? • Service-level improvement initiatives 5

Why do we need a “trade-off”? • Alternative to reliance on articulated choices • Purchasers (rationally) want the best of everything at the lowest cost • All product/features may seem equally important or attractive • Price is always important - especially if unbounded • Trade-off exercise forces purchasers to choose what’s most important to them • Features, price and brand • Consistent patterns across trade-offs determine “importance” of features and price points 6

Example: Will consumers pay more for a Lead-Safe faucet? Will they pay: • 5% more? • 10% more? • 15% more? • Will they buy it all? 7

With only two options and four price points on one option, four simple choices will answer our questions Which faucet would you choose at each price premium for lead-free brass waterway? Standard brass at $100. 00 Lead-free brass at $100. 00 Lead free brass at $105. 00 X Lead-free brass at $110. 00 X Lead free brass at $115. 00 X X NOTE: We can adjust the $100 base price and all the offset prices depending on how much the customer plans to spend. For example, if the customer is shopping in the $300 range, the offset prices would be $315 (+5%), $330 (+10%) and $345 (+15%). This is more realistic than: • Having customers make choices in price ranges they aren’t shopping, or • Presenting percentages that require mental-math Tip: Realism is critical in trade-off exercises! 8

Results show expected volumes at each price point FICTITIOUS DATA! 9

But it gets more complicated by adding just one more waterway material… 16 choices with 3 options each • Becomes unmanageable with any more dimensions or options Which faucet would you choose at each price premium for lead-free brass or plastic/PEX waterway? Brass Brass Brass Brass Lead-free Brass, same price as brass Lead-free Brass, 5% more than brass Lead-free Brass, 10% more than brass Lead-free Brass, 15% more than brass Plastic/PEX, same price as brass Plastic/PEX, 5% more than brass Plastic/PEX, 10% more than brass Plastic/PEX, 15% more than brass Conjoint efficiently organizes more trade-offs than could otherwise be tested 10

Actual Faucet Example • Six brands • Five features with 3 -4 options each • “Supply line material” feature only asked with “Supply lines included” option Tip: Keep features consistent and logically ordered! • 6 tested price-points (including “same as standard”) • Two numeric “benefits” derived from feature option combinations 11

Conjoint Exercise: Feature Definitions Tip: Clearly define all features especially if testing new concepts! 12

Conjoint Exercise Screen Tip: Conjoint exercise should be easy to read ! Tip: Change color schemes to highlight differences across screens and indicate progress to avoid respondent fatigue! 13 13

Conjoint and Simulation Process BASELINE CONJOINT SIMULATOR PROJECTIONS OUTPUT 14

Simulator “Baseline” • • Required to forecast actual share-shifts, unit sales and/or revenue Not required simply to identify configuration that maximizes uptake FICTITIOUS DATA! Tip: Baseline information should closely match what we are trying to forecast: e. g. , current year sales (baseline) to forecast next year sales, not installed base – may be very different from recent sales 15

Simulator Input: Baseline product configuration Tip: Baseline configurations should come as close as possible to what each competitor actually offers 16

Simulator Input: “Other Brand” offering Tip: Program simulator to easily recover baseline after simulated changes 17

Simulator Output: “Other Brand” offering, no price change Tip: It’s a good idea to display the scenario on the same simulator page as the output to keep track of simulations easily 18

Simulator Output: “Other Brand” offering with 10% price increase 19

Simulator Input: Competitor response 20

Simulator Output: competitor response and “Other Brand” 5% price decrease (from 10%) 10% “Other” Price Premium 5% “Other” Price Premium 21

Percent share of sales Simulator Output: competitor response and “Other Brand” at 0 -10% price premium TIP: Tested price points should bracket range under consideration. Interpolation more valid than extrapolation. “Other Brand” Percent Price Premium over Standard Brass - Delta fixed at 5% - Moen fixed at 0% 22

Types of Conjoint Exercises • • • Ratings Discrete choice Rankings Constant sum Volumetric 23

Example New Home Rating Conjoint (ca 2001) 24

Ranking combines strengths of discrete choice and ratings “Discrete” 1 st choice 2 nd-6 th choices provide additional data But strength of preference not always clear and ranking might not seem realistic 25

Constant Sum and Volumetric Allocation both distribute fixed sums (per respondent) across offerings • Constant sum allocates “points” or probabilities to gather data on all offerings while forcing a trade-off • Volumetric asks, “how many of each would you purchase? ” • Volumetric is required for any frequent or multiple purchase situations, e. g. : § Multiple units or volumes of products (trucks, gallons of paint, parts…) § Frequent purchase occasions (visits to stores, delivery services, etc. ) Tip: In the faucet project, we used rankings for consumers who typically buy only one faucet and volumetric for plumbers and channels 26

Summary: Types of Conjoint Exercises Type Ratings Discrete choice Rankings Pros Cons Captures data on each rating Can result in ties, flat ratings, truncated scaling Simple, intuitive, forces a Thin input data; may choice overestimate adoption Maximizes discrimination Can be difficult to rank all while capturing data on each choices; interval among rankings offering uncertain Constant sum Captures data on all offerings, can produce interval results Can be tedious/time consuming for respondents; less realistic Volumetric Highly realistic, captures data on all offerings, essential for multi-purchase situations Not appropriate for single or infrequent purchases 27

Conjoint Alternative: Max Diff • Determines relative importance (rank ordering) of attributes via repeated subset trade-offs • Works best with long lists of binary (yes/no, present/absent) attributes • Answers questions like “what are top 5 drivers of shopping here? ” • Not appropriate for product configuration or pricing Of these four features, which is least likely and most likely to get you to shop at a hardware store or home center? 28

Hybrid Conjoint: “Nested” example Requirements Identification & Provider Performance Ratings Constant Sum Volumetric Conjoint 29

Hybrid Conjoint Simulator Baseline Scenario Modeled Scenario 30

Complete hybrid model: Over 100 testable features in 9 product/service touch-points Example Modeled 31

Hybrid Conjoint Simulator used to prioritize strategic service-level improvements (shipping category, ca 2002) “What if” brand performance was significantly improved? Share change as result of Service Level improvement BASELINE 32

Full profile vs. Adaptive Conjoint • Full profile shows all attributes and all levels of each attribute § However, every respondent sees a subset of combinations • Adaptive shows partial profile, often based on preliminary screening § Ostensibly can handle more attributes § But concern is that attributes initially “dismissed” as unimportant might be evaluated differently in a full-profile § And a hybrid might be a better alternative if the attributes are “nested” 33

Types of Conjoint Exercises – Other Considerations • Branded vs. un-branded § Single or multiple-offerings per brand • Ability to back in-out brands or products § New brands, co-branding § Entirely new products § Retiring brands or products § New/old brand or product co-existence TIP: Decide up front if you will need to back specific brands or offerings in and out. They should “appear and disappear” in the choice sets respondents actually see. 34

Example truck JV “super-simulator” (ca 2007): • • Multi-products per brand Cross-branded offerings Product back in-out Brand back in-out 35

A few final tips: • Keep trade-off exercise realistic and manageable § Sufficient but not excessive number of attributes and options § No illogical combinations if some attributes depend on others § Keep pricing consistent with offerings on same screen • Baseline and calibration are critical forecasts § Baseline configurations and shares should be empirically based § Simulator sensitivity often in need of adjustment § Proper data weighting critical to gauge movement off of baseline • Purchase timing can be indeterminate unless built-in § Can specify purchase period – e. g. , “within next 12 months” § Can adjust forecasts based on actual incidence of purchases, cycles • Conjoint assumes “perfect market awareness” § Therefore, simulations often over-estimate actual uptake § Back-end adjustments often required 36

What have we learned? • What is conjoint? • What does conjoint help us do? • What are the various conjoint approaches? • How can we use conjoint to predict purchase behavior? George H. Leon, P h. D. Senior Vice Preside nt gleon@nationala nalysts. com Tel: 215. 496. 6892 37