User experience and design

3 downloads 0 Views 3MB Size Report
Mar 31, 2014 - What to conclude? ▫ Huge amount of data on consumer habits, choices ... user can access refrigerator several times per day, place or remove ...
User experience and design

Bernard Yannou, Professor [email protected]

How to know if goods and services are well adapted?

Using a home theatre Using a jigsaw for cutting wood Using a refrigerator

Energy and water consuming activities Providing efficient solutions against falls of elderly

2

What to observe? What to conclude? 



 

Huge amount of data on consumer habits, choices and preferences can/could be accessible Capturing data for learning of people motivations, usage and purchasing contexts, perceptions, preferences and satisfactions Which data to measure? How to get innovation insights/leads out of that and use it in the design process?

3

How to mix information streams for specifying and innovating?

Which design platform for gathering information for innovation and redesign? 1) Experiments in labs of prototypes or existing solutions

Azqueta, I. 2014. Global platform for future experimentations and developments of intelligent refrigerators, MSc in IE bibliographical study. Ecole Centrale Paris - Laboratoire Génie Industriel.

2) Surveys and observation of people

3) Innovation processes (need expression, creativity workshops, internal research, usercentred design, crowdsourcing)

4) Design analytics from online reviews, tweets…

5

How to experiment in labs? 

Experiment protocol 





Results    



user can access refrigerator several times per day, place or remove shelves, add, move, expire food. Simulator measures shelves positions, at which shelf food is placed, height of goods. Most of customers use no more than 4 shelves Selve #2 rarely used Front area of shelves is used more actively Bottom and upper shelves are the most populated

Redesign  

Lewis K. and Horn D. Design analytics in consumer product design: a simulated study. ASME International Design Engineering Technical Conferences, 2013

4 shelves only Increase the vertical space of the top slot in order to stock higher objects

Redesign with similar architecture

6

How to relate observable indicators to KPI/Functions/Values assessment? 3) Performance modelling by data fitting or expertise (fuzzy rules)

4) Assessing performances from measures

Intermediate Indicators

2) Observation and data recording

Number of door openings Duration of door opening Ambient temperature Cabinet Load Time examining refrigerator's interior contents Time loading the refrigerator Time spent searching an item Time spent on grocery shopping Time spent cleaning refrigerator Time thinking recipes Quantity of food wasted Usage of shelf space Shelf configuration used by consumers User-friendly interface Size of food items Configuration possibilities Item distribution by size Average cost of electricity Noise

Lushnikova, A. 2014. Hybridizing design analytics techniques on fridges, MSc in IE bibliographical study. Ecole Centrale Paris Laboratoire Génie Industriel.

KPIs Energy Consumption Time

Food Waste

Refrigerator

Comfort of use

6) Changing architecture

Economic Cost

Freezer/Cooler position Volume of Freezer section Volume of Col-section Shelf layout Compartment layout Refrigerant used … … Temp required in F-section Temp required in C-section

Design parameters

1) Functional analysis

5) Adapting design parameters 7

Design analytics from online reviews Data extraction

Preprocessing

Text processing

Sentiment analysis

Sentiment rating

5 4 3

Took ROOT

IT

Average

2

Longer ADVMOD

NSUBJ

Model's Score

1 0 1 3 5 7 9 11 13 15

run DEP

to

wires

AUX

DOBJ

the DET

across room POBJ

the DET



 

did NSUBJ

PREP

than

it

MARK

hook

NSUBJ

to AUX

XCOMP

it DOBJ

up PRT

Amazon Product Code: B003B8VBJ2 Product name: Sony BRAVIA DAV-DZ170 Home Theatre System (Electronics) Review: Well, Sony definitely let me down on this one. First off this unit was easy to set up. It took longer to run the wires across the room than it did to actually hook it up. But the volume on this was sub-par. Even on the max level volume (35) it still wasn't that loud. The main problem was the amount of bass that it produces. The bass is so overpowering that you can barely even hear people talking in the movie, and there is no way to adjust the levels at all.

actually ADVMOD

Automatic sentiment rating, heuristics validated by experimentation Also delivering of good/bad functions, and solutions Next step: using function glossary to synthesize insights

Raghupathi D., Yannou B., Farel R., Poirson E. 2014. Learning from product users, a sentiment rating algorithm, In International Conference on Design Computing and Cognition, 23-25 June, London, UK.

8

What to observe on users / customers / purchasers?

Usage modelling is needed, design by usage is useful for creating values Data on usage patterns are needed

Design Optimization framework Candidate product

Performance evaluation Objective function

Functional analysis (simplified and averaged description of expected functions and performances in lifecycle situations)

Representation of averaged preferences (assessed by polls, marketing experts, design workshops…)

Preference aggregation model

Multiple objectives

Pareto optimal solutions set

Competing products

Degree of absolute satisfaction

Averaging customer preferences is awful! No consideration of competing offers

Degree of performance dominance (distance to Pareto frontier)

Performance evaluation

Yannou B., Yvars P.-A., Hoyle C., Chen W., (2013), Set-based design by simulation of usage scenario coverage. Journal of Engineering Design, vol. 24(8), p. 575-603, doi. 10.1080/09544828.2013.780201.

10

Choice and Market Share modeling framework

Market segmentation

Primary selection of attributes influencing the choice

Questionnaires

Choice model (e.g.: EUVT Conjoint analysis Discrete choice analysis)

Competing products

Market share model Market part

A market segmentation and survey is possible only if the product class exists

Candidate product

Not adapted to an innovative offer

Yannou B., Yvars P.-A., Hoyle C., Chen W., (2013), Set-based design by simulation of usage scenario coverage. Journal of Engineering Design, vol. 24(8), p. 575-603, doi. 10.1080/09544828.2013.780201.

11

Design by usage coverage simulation framework Collecting usage situations is more independent of existing classes of solutions Candidate (parameterized) product Space of usage scenarios Collection of typical usage contexts (or lifecycle situations)

Customer demographics, User skills, Performance bounds, Preference profiles

Simulated customer desired attributes/performances (Physics-based models and/or Human appraisal experiments)

Mapping models between simulated performances and usage scenarios

A product is considered as a mean to deliver a service. The service quality depends on the user skill

Feasible usage scenarios at an expected quality level

Absolute usage coverage indicators

Competing products

Relative usage coverage indicators

Yannou B., Yvars P.-A., Hoyle C., Chen W., (2013), Set-based design by simulation of usage scenario coverage. Journal of Engineering Design, vol. 24(8), p. 575-603, doi. 10.1080/09544828.2013.780201.

12

A design by usage approach based on physical models of performances and constraint programming 

Which existing solution is the best for my needs?

13

A design by usage approach based on physical models of performances and constraint programming

Bosch family of jigsaws

50 equations of physics (efforts, velocities, cutting law…) Usage coverage simulation

Ergonomics user-related data Survey on usage scenarios sets of a set of representative users



The service delivered depends on the user capabilities as well as 14 the products and usage scenarios !!!

Simulation of market share and optimization of a scale-based product family

Wang J., Yannou B., Alizon F., Yvars P.-A., 2013. A Usage Coverage-Based Approach for Assessing Product Family Design. Engineering With Computers, 29 (4), 449-465.





Development of Usage Coverage Indicators for a product family Simulation of market share in different market contexts

15

How to get information in case of diffuse situations?

Falls among the elderly (65 years and older)



A painful situation, in France per year (in 2010):     



3.5 million falls/year 5,000 deaths/year directly caused by falls As a comparison: 1,800 deaths by car accidents 80,000 hospitalizations/year because of falls 2 Billions € of financial costs of hospitalization

A diffuse situation   

Multiple fall (usage) contexts Multiple causes No commercial interest to organize data collection

17

A design by usage approach based on learning of usages scenarios segments and usage coverage simulation Bekhradi A., Yannou B., Farel R., Zimmer B., 2013. Building of usage scenarios space for investigating the fall situations of the elderly people, IDETC Conferences, Portland, Oregon. Male: 1

60 medical publications

Female: 1,7

Probability of falling

More than 30 contextual variables e.g. Gender={Male, Female}

Male 42% Female 58%

Modality (segment) size

Experts' heuristic-rules declaratives

Usage scenario space (Oracle)

Performance vector of pain points

Bekhradi A., Yannou B., Farel R., Jena S., Zimmer B., 2014. Simulating global utility of design solutions to elderly falls by building relevant usage segmentation, In International Design Conference, Dubrovnik, Croatia.

Simulation of usage segment coverage and global perforrmances

18

A new innovation method and tool for simulating in advance the utility to create in targetting painfull situations Bekhradi A., Yannou B., Farel R., Jena S., Zimmer B., 2014. Simulating global utility of design solutions to elderly falls by building relevant usage segmentation, In International Design Conference, Dubrovnik, Croatia.

Value buckets

19

How to organize feedback to people and incent them to well behave?

The case of energy and water consumption of households

Eco-designing buildings 

Zaraket T., March 31 2014. Stochastic activity-based approach of occupantrelated energy consumption in residential buildings. PhD dissertation. Ecole Centrale Paris, Laboratoire Génie Industriel.

Presence at home Issue: Energy performance of a residential building

Location Altitude Surface area Orientation Energy-consuming actions

HVAC

Temperature

Lighting Hot water

Humidity

Fabi et al. (2012), Yun et al. (2011), McLoughlin et al. (2012)



Occupant behavior (activities, usages) is the most complex phenomenon (Robinson 2006, Page et al.2008)

21

Data model for a systemic view of energy consumption in residential building

18/03/2014 22

SABEC simulation scheme (Stochastic Activity-Based approach of occupant-related Energy Consumption ) Activity Patterns For each activity

Diversity of user profiles Income Number of occupants

Family types

Activity quantity per individual

Socioprofessional class

Aggregation

Education level

Activity quantity per household Individuals attributes

Age

Appliance(s) use pattern

A specific household Household attributes

Influencing parameters

Energy consumption per activity

Appliance Appliance(s) ownership Appliance(s) characteristics

Interesting notion of cascading of service units Overall energy consumption for all activities 23

Simulating “Watching TV” consumption TV’s technology TV ownership

2

1

Given Household

TV’s energy efficiency

2,3,4

CRT

LCD

Plasma

Stochastic

Stochastic

Stochastic

Energy consumption (KwH) Statistical data 1

2

3

1 4

Individual TV watching duration

REDOMECE

Activity sharing heurisrics Stochastic

Model fitting

Simulating “Washing clothes” consumption Given Household

WM’s ownership

WM’s characteristics

WM: washing machine/

Energy efficiency

Capacity Clothes weight worn/individual/day 1,2

3

Dirty clothes weight /month

Number of cycles /month

Consumption Kwh/month

White clothes

White clothes

White clothes

Colored clothes

Colored clothes

Colored clothes

3

3

Clothes changing rate/month/individual

White/colored %

4

WM filling %

Statistical data 1

2

3

4

5

Energy consumption (KwH/month)

6

Survey Stochastic

Wahsing temperature (white/colored)

Stochastic

Static or dynamic analysis of residential energy consumption? Dwelling

Household variables

Domestic appliances

Smart-metering National statistical data SABEC-Prediction (Static Model)

Parametric predictive modeling

SABECmonitoring (Dynamic Model)

Dwelling

Clustering of households’ profiles Dwelling

Regional & Building National network

Regional Building & National network

SABEC-Prediction Database Energy consumption allocation per activity

Real Consumption Database

Specific household ?

Human-Machine Interface Learning Oracle

Static or dynamic analysis of residential energy consumption?  Which HMI for managing residential energy consumption and incenting positively occupants to save energy and water? Human-Machine Interface Recommendations and Action Plans Diagnostics, Recommendations according to Occupants’ Profile

Data Cloud Consumption analysis

Reference points

Household Consult

Inform

Specific Consumption

Visualization Diagrams, Datasheets, Audio-Visual Alerts

25

20 20 17 14

15 12,16 10

9 8

7,29

5 4

5

2 0 Laundry

Entertainment

Real consump on of HH1

Food

House-caring

Self-care

Na onal average consump ons

Prediction database

Building Consumption Regional Consumption

National Consumption

Monitoring database