Merging multiple streams of data to improve portfolio strategy

Merging multiple streams of data to improve portfolio strategy


Reading time: 5 mins

Market research contributes to many of the
tactical decisions, but the wealth of new data sources and forms, combined with
various analytics, means MR can meaningfully contribute to more strategic decisions
regarding the products and services being offered. What markets do we enter?
What markets do we exit? What products and services do we add, refine or
emphasise to meet the needs and wants of individuals, households, and firms – and
to grow the business profitability?

One of the most useful structures for
framing product portfolio strategy discussion is market structure or
competitive landscape. What products or services compete with each other, to
what degree do they compete, and why? When populated with historical sales
data, a number of important questions can be answered.

  • What product/service segments
    are growing faster or slower?
  • Is the firm’s portfolio aligned
    to growth?
  • How do the firms’ growth
    prospects compare to competitor firms?
  • Where are the greatest new
    product/service opportunities, and what size of business can reasonably be
    expected?
  • Are there flanker and extension
    opportunities, and for which brands in which product/service segments?
  • If new products are not a
    viable way to enter an attractive product/service segment, what acquisition
    targets are attractive?

Approaches

Identifying product/service segments
usually involves some direct or indirect measurement of the degree to which
items are substitutable or similar. Econometric approaches usually involve
price elasticities and/or brand switching. Survey approaches usually involve
simulated trade-offs or measures of perceived similarity. A partial list of
approaches:

  • Analysis of price elasticities
    based on continuous syndicated sales data
  • Analysis of brand switching
    based on longitudinal purchase patterns
  • Analysis of claimed past
    purchase (survey), often based on correlation or covariance
  • Sorting into dynamic groups
    (survey)
  • Analysis of brand perceptions,
    usually based on survey data

Each approach has its own strengths and
weaknesses, and recommendation is based on context and specific objectives.

A case
study

The client is a global food and beverage
fmcg predominantly focused on breakfast and morning eating occasions. Their
syndicated sales databases were organised on the basis of a market structure/competitive
landscape that had been created many years earlier. Consumer tastes and
behaviors had evolved and it was times to update their understanding. A
simplified example might look like:

The most critical project objectives were
to:

  • Identify the relevant
    competitive structure/product segments in several of their most important
    markets, which would then be used to realign the syndicated sales data base
  • Delineate in detail “how” and
    “why” brands cluster into their various product segments, which would ease
    transition from strategy to tactics

 Challenge #1; a framework that serves both strategy
and tactics

Strategies often fail in implementation
because they are too abstract. In discussions with client, the most attractive
framework for organising detail was brand architecture, and the critical elements
were:

  • Target
  • Occasions
  • Functional benefits
  • Emotional benefits
  • Key attributes

If brand profiles were available for all
(or nearly all) brands in the respective markets, the key objectives could
theoretically be achieved;

  • Brands could be clustered based
    on their profiles (in roughly the same way that consumers are clustered in
    consumer segmentations)
  • “How” and “Why” brands are
    similar or different would be clear from both the analysis and aggregated brand
    profile comparisons

Additionally, a well articulated market
structure, combined with sales data and the appropriate analytic tools, could
answer important foundational questions:

  • What elements of brand
    architecture are critical to growing existing brands or to creating new brands?
  • Where do we focus; target,
    functional or emotional benefits, occasions?

Challenge
#2; data paucity, especially in developing markets

In the age of big data, data paucity in
certain markets might be a surprise. In developed markets syndicated sales data
coverage might be close to 100%, but in some relevant developing markets it is
less than 50%. Household purchase panel data is typically more limited. Surveys
with large sample sizes could be used to augment, but even here there are issues.
The relatively low penetration of relevant product categories and brands negatively
impacts what can be achieved with either household purchase panel data or
survey data. In some markets the majority of brands have annual penetration of
1% or less.

Merging
multiple streams of data

The solution was found in a qualitative
form of data combined formally and holistically with the aforementioned
traditional quantitative forms of data; a synthesis of syndicated sales,
household purchasing behavior, quantitative survey and qualitative semiotics.

The table below describes the different
data streams, and what each could contribute to informing the brand profiles.

Semiotics

In this case a semiotician interpreted readily
available commercial iconography – packaging, advertising, websites for all (or
nearly all) brands in all markets to infer the various elements of the brand
architecture; who the target is, what the functional and emotional benefits
are, what the occasion is, etc. Semiotic inference was especially useful for
identifying the emotional benefits.

The chart below displays how this works and
how it is validated. The brand is Energen – a breakfast cereal drink that
originated in Indonesia. You can see the inferred target, occasion, and
functional and emotional benefits. Since this is a fairly ubiquitous brand, there
is adequate quantitative data that can be used to validate the semiotic
inferences. Here, the semiotic inference for target and occasion was not wrong,
but it was too narrow; the brand was consumed by a broader demographic across a
wider set of occasions. Across the various markets, this kind of validation could
be done for 10% to 70% of the brands, on various brand profile elements.
Indeed, predictive models could be created for various visual and text icons.

Proof
of concept

Given the forms of data and how they were
merged, proof of concept/validation was important to build faith in the final
product. Brand profiles built semiotically were validated against brand profile
elements from quantitative data where those existed. Some markets had market
structures built recently from household purchase panel data (considered by
client to be the gold standard). Fusion based market structures were compared
to these.

While the deliverable met objectives, attention
areas were:

  • one important market failed to
    embrace final conclusions because of the inclusion of qualitative data despite
    efforts to build credibility
  • it was difficult in certain
    markets to find semioticians
  • the semiotic coding was time
    consuming, but we hope that as AI develops, there will be better ways to “code”
    visual iconography

Fusion approaches promise to provide flexible,
holistic, and end-to-end solutions. The key in this case study was a framework
for merging multiple streams of data. While the unstructured data flowed from
semiotics, we think there is opportunity for other forms of unstructured data
such as social media, biometric data, geo-location, and passive metering.



Source link

?
WP Twitter Auto Publish Powered By : XYZScripts.com