Task2

26/02/20

1

Measuring Consumer Behaviour Part II

Week 6

Required readings In-class handouts Ehrenberg, A. S. (1995). Empirical generalisations, theory, and method.

Marketing Science, 14(3), 20-28. Uncles, M & Wright, M. (2004). Empirical generalisation in marketing,

Australasian Marketing Journal, vol. 12, no. 3, pp. 5-18. Ehrenberg, A. S. C., Uncles, M. D., & Goodhardt, G. G. (2004).

Understanding brand performance measures: Using Dirichlet benchmarks. Journal of Business Research, 57(12), 1307–1325.

Learning objectives •  To understand the importance of measuring consumer

behaviour and the different approaches to it

•  To explain how to measure consumer buying behaviour through purchase panel data

•  To discuss the most important known patterns and expectations in consumer buying behaviour

•  To discuss the importance of empirical generalisations and mathematical models of consumer behaviour

•  To practice the analysis of purchase panel data

26/02/20

2

Purchase panel data

•  Records of individual consumers’ purchases over a certain period of time for all products/brands in a market

•  Large samples of consumers (thousands) •  ‘Revealed’ choice (real purchases made by

consumers)

In the 1950’s Andrew Ehrenberg, marketing scientist, started analyzing UK purchase panel data and noticed that:

•  Consumer buying behaviour showed great variation at individual consumer level, yet was very stable recurring trends/patterns at aggregate level

•  These characteristics of consumer buying behaviour strongly affected the effectiveness of business strategies (and vice versa)

Usefulness

So…Why using this?

•  Objective and scientific (facts, not ‘stories’)

•  Helps identifying measurable trends (expectations), which leads to forecasting and setting of practical guidelines

•  Useful to originate data-driven theories, i.e. empirical generalisations

26/02/20

3

Empirical generalisations

•  Recurring trends/patterns emerging across a wide range of conditions or contexts (e.g., across a wide range of different markets)

–  Supported by empirical evidence (analysis of purchase panel data over time)

–  Can be explained through quantification (mathematical expressions or formalized statistical models)

–  Can be used for benchmarking and forecasting

Total sales for each individual brand (sum of the values in each column)

Total category purchases by each shopper (sum of values in each row)

Total CATEGORY sales

Snapshot of panel data (e.g., spreads):

Measures obtained from panel data

Brand Size Brand Loyalty Market Share (%) Penetration (%)

Average Purchase Frequency Category Buying Rate Share of Category Requirement (%)

Derived from the counts of individual purchases over a specific time frame – e.g., one year or one month

26/02/20

4

30

26 20

16 8

Market Share % Canola Sun

Dairy Blend

Olive Grove

Golden Spread

To calculate MS simply take the total brand sales (values in the bottom row of the table) divided by the total category sales

MS (%) = sales of a brand / sales of the category * 100

Market Share (%)

PEN (%) = number of buyers / number of shoppers * 100

•  Buyers are ACTUAL customers of a brand (people that have bought the

brand), whereas shoppers are all POTENTIAL customers in the market

•  Penetration is a measure of the proportion of people that bought the brand AT LEAST ONCE, in the given time period

To calculate penetration, you need to work out how many ‘buyers’ a brand has, so counting the number of people out of the sample of 30 potential buyers who purchased a specific brand That number divided by 30 gives the penetration of each brand

Penetration (%)

APF = number of purchases of THE BRAND / number of brand buyers NB: it's a FREQUENCY (there is no %) It explains how often - how many times on average brand buyers have bought the BRAND

To calculate average purchase frequency you simply need the total brand sales (values in the bottom row of the table) divided by the number of brand buyers

Average Purchase Frequency

26/02/20

5

CBR = number of purchases of THE CATEGORY made by brand buyers / number of brand buyers NB: it's a RATE (there is no %) It explains how often - how many times on average brand buyers have bought the CATEGORY

To calculate category buying rate you simply need the category purchases by each individual brand buyer (pick values from total category purchases column) divided by the number of brand buyers

Category Buying Rate

Calculating CBR:

E.g. for Golden Spread: Customer 3 bought GS 4 times, but bought from the category 7 times, it’s the 7 that we’re interested in. Customer 5 bought GS once, but bought from the category 11 times. Again, it’s the 11 that we’re interested in To calculate the top part of CBR formula, you’re simply adding each brand buyer’s category total... 7 + 11 +5 + 4 + so on…

SCR (%) = (APF / CBR) * 100 It measures the proportion of category needs/purchases satisfied by a certain brand (it is, in fact, the proportion between how much brand buyers have bought on average of the category and how many of those category purchases were dedicated to a specific brand)

SCR is extremely simple if you have correctly calculated APF and CBR J

Share of Category Requirement (%)

26/02/20

6

What can we understand about consumers buying behaviour through the analysis of the values of these measures over time, in a given market?

…consumer s are

‘heterogene ous’ in

their choices and

preferences

…con sume

rs buy

infreq uently

and

irregu larly o

ver tim e

…consumers buy different

product/services categories at different rates, but brands

within a specific category at similar rates

…consumers buy more than one brand within a specific category (repertoire buying)

KEY PATTERNS

KEY PATTERNS

Reliable abstractions to predict consumer buying behaviour (output) by relying on these

expected patterns (input)

1 + 2 always = 3

Mathematical models

26/02/20

7

Key regularities of consumer

buying behaviour

observable in the measures derived from

purchase panel data

Predictions of expected

brand performance

measures

Benchmark of current vs. expected trends

Statistical distributions simulating

these regularities

Forecasting future trends

INPUT: observed measures derived from purchase panel data analyzed over time ‘MATHS’ (software) OUTPUT: theoretical measures (to be compared against observed for managerial insights)

The Ehrenberg and Goodhardt repeat purchase model

How the software looks like

Input data

26/02/20

8

o  Mean Absolute Deviations (MADs) are the averages of the deviations (differences) between observed data from the panel and theoretical outputs from the model, once withdrawn the sign

o  Reasonably small deviations and Mean Absolute Deviations (MADs) represent ‘good fit’ of the model – i.e. representativeness of real purchase patterns and brand performance for the market analysed

Output data

Resulting empirical generalisations 1 of 2

•  Double Jeopardy

Bigger brands (greater market share/more popular) have more buyers (greater penetration), who are also slightly more loyal (greater purchase frequency and SCR)

Smaller brands (smaller market share/less popular) have fewer buyers (lower penetration), who are also slightly less loyal (lower purchase frequency and SCR)

How can you spot this in the data?

0.00#

0.50#

1.00#

1.50#

2.00#

2.50#

3.00#

3.50#

4.00#

4.50#

5.00#

0.00# 0.05# 0.10# 0.15# 0.20# 0.25# 0.30# 0.35#

Purchase) Frequency)

Penetra/on)%)

Double)Jeopardy)line)

26/02/20

9

Exceptions to the Double Jeopardy

Excess loyalty brands (very high market share) – i.e., brands that have many buyers (very high penetration) and very high loyalty (very high purchase frequency and SCR)

Niche brands (small market share) – i.e., brands that have few buyers (low penetration), but these buyers are quite loyal (high purchase frequency and SCR)

Change-of-pace brands (high market share) – i.e., brands that have many buyers (high penetration), but these buyers are not very loyal (low purchase frequency and SCR)

Resulting empirical generalisations 2 of 2

•  Duplication of Purchase Law

Brands share customers with other brands (repertoire buying) in line with their market share (smaller brands are once again penalized)

Exceptions to this rule are market partitions, i.e. subgroups of brands that are highly substitutable and catering for specific needs (e.g., diet soft drinks, or gum protection toothpastes)

To evaluate the Duplication of Purchase count the proportion of customer sharing of customers, i.e. for each brand’s buyers, see how many have bought the other brands as well

BRAND 1 BRAND 2 BRAND 3

BRAND 1

BRAND 2

BRAND 3

Average

Who also bought ..

T h

o se

w h

o h

av e

b o

u g

h t.

.

% who also bought = number of buyers of a brand who also bought another brand out of the total number of that brand’s buyers * 100

Brands need to be reported in line with their size (i.e. bigger brand first row/column, then second biggest brand and then smallest brand)

How can you spot this in the data?

26/02/20

10

HOW TO DETERMINE IF THERE IS DUPLICATION OF PURCHASE: (1) ROWS: numbers must decrease (getting smaller) from left to right (2) COLUMNS: numbers must be pretty much in line with the average value in the bottom Note: exceptions from (1) and (2) (e.g., large difference from the average) indicate MARKET PARTITIONS = brands sharing more (or less) than expected customers, given their market share

Limitations of panel data •  Expensive and hard to access

•  Data may be chain or retailer specific •  The approach takes into consideration only the

objective nature of consumer behaviour – all other factors and contingencies are assumed to be exogenous and somewhat irrelevant

•  Longitudinal approach required (i.e., analysis must be carried out regularly over multiple time periods otherwise is pointless)

Team activity 2

•  Use the Excel data set given to calculate all measures for all brands in the given market (cereals)

•  Report your results in the template given (see FLO)

•  Write 150-200 words describing the patterns in the data, highlighting if you see any exception from these patterns

26/02/20

11

Teaching break and mid-semester revision

Next…