Harmonizing Filipino Household Classification Using Data Science

Nicco De Jesus
January 29, 2019

A fascinating thing happens after you buy or browse anything from Amazon, Lazada, Zalora, and similar e-commerce sites. Related items, items frequently bought together, and items that other people who also considered your original purchase flash onscreen, enticing you to add these to your shopping cart upon check-out.

Buyers on these e-commerce platforms get a push of sorts that whets their appetite to buy more, even making them crave that adrenaline rush of scoring a deal.

This is because the data-enabled recommender apps of Amazon and the like effectively hit the right targets for the happy buyer (read: you), by checking your browsing and buying behavior and then sending you “suggestions” via the tons of interrelated information distilled into your profile.

This is not new. Targeting and retargeting strategies are actioned based on how one’s consumer profile matches a particular demographic or market segment. Marketers heavily rely on profiling info to know what exactly to sell to whom.

Customer profiling in the digital age

The most basic profiling information today falls under the class of socioeconomic (SEC) data. In European countries, SEC data are clustered by occupation. In the US, it’s by income. Theoretically, in countries that practice a form of social democracy, such as Germany and among Nordic nations, there shouldn’t be a marked difference in SEC data. Meanwhile, in most of the developing world, such as Indonesia, the Philippines, Mexico, and India, it’s a mix of measurement systems.

Why is proper profiling of SEC data important? Simply put, it pays to know one’s target market. In this age of hyperpersonalization and creating a seamless omni-channel customer experience, having clarity on the SEC data classes can spell the difference between a successful marketing strategy that resonates with one’s audience, and another that falls flat.

Overhauling Philippine SEC data

If you guessed that there are 5 SEC data classes in the Philippines–A, B, C, D, E–then great. If you then say that there is common agreement on how big or small, say, “A” is from “D”, then you might have some problems there.

The fact of the matter is that there are as many proportions as there are users or doers of research. At the same time, among marketing professionals evaluating market share among, say, “broad C” customers, there is still disagreement as to what defines each SEC data class.

At least, until now.

In 2019, the first massively tested model by the Philippine Statistical Authority (PSA), the Marketing and Opinion Research Society of the Philippines (MORES), and the School of Statistics of the University of the Philippines will be launched.

The following outcomes were derived through the discipline of data science that helped “clean” and harmonize Philippine socioeconomic data (1SEC):

  1.  There are nine, not five, consumer household segments according to socioeconomic class.

    Chart: Consumer Household Segments
    Figure 1. Consumer Household Segments by SEC

  2. Expenditure, not income, is more effective in classifying consumers.
  3. The best strategy to arrive at the SEC data clusters is unsupervised learning approaches. Meanwhile, the best among thousands of models to predict membership in each cluster belongs to a class of models known as Ordinal Logistic regression.

The 2019 1SEC model further features the following:

  1. Aggregate consumption patterns derived using advanced analytics, such as multivariate techniques and machine learning
  2. Learning data involved 2009, 2012, 2015 and the 2018 Family Income and Expenditure Survey  (with n>= approximately 40,000)
  3. Close collaboration between government through the PSA, private sector, academe, and MORES, the professional association for marketing professionals in the country
  4. Set-up as a Continuous Learning Process, and
  5. Involves the use of a secure cloud server.

What drives the prediction model for Philippine consumer segments?  The 1SEC model neatly buckets the variables into nine, namely:

  1. Members of the household, such as the quality of consumers in the household (i.e. number of employed members, level of education, etc);
  2. Energy consumption potential, such as the number of selected energy-using facilities owned (i.e. TV, microwave oven, computers);
  3. Urbanization (i.e. Urban and regional membership in areas defined as “cities”);
  4. Mobility profile (i.e. type of transport owned),
  5. Water source type;
  6. Connectivity, or the number of phones owned and type of broadband subscription accessed, if any;
  7. Living Space Assets (i.e. number of sala sets);
  8. Living Shell (i.e. type of wall and type of roof); and, lastly,
  9. Tenure of Home, which simply asks whether the house is owned or rented.

Chart: 1SEC Model Variables

It is expected that these drivers will vary in importance as consumer lifestyles rapidly change alongside shifts in the environment, given technology advancements, culture shifts, new family structures, developments in retail, and the like.

Consumer Distribution by 1SEC Clusters
Figure 3. A Quick View of PH Consumer Distribution by 1SEC Clusters.

Implications of harmonized SEC data for marketing

Because of the 2019 1SEC model, the application of big data in the field of marketing can rapidly advance as more marketing executives openly compare SEC data wins (or losses) on each of the standardized SEC data cut.

What does this mean for the average business and/or marketer? Properly prepared data that is put to good use has now become the holy grail. Companies and marketers who actively track, manage, and make use of their existing and prospective customers’ data relative to the new model, and apply the emergent disciplines of data science and artificial intelligence, will reap the fruits of the digital age.

Other countries have relooked their respective consumer clusters as the digital age has made targeted marketing a necessity. Take for example the second biggest market in the world, India. The Indian socioeconomic system shows grades of high SEC using the 5-letter system with the highest-spending clusters starting from “A1”.

The household classification for both countries when standardized in 2020–India is reported to again revise their SEC data–will serve a basis for marketing recommender strategies for home-based consumers.

Sharing SEC household classification with at least one billion other people opens up not only synergy across multi-country marketing strategies, but it also makes possible other benefits, namely:

  1. One ‘Scoreboard’.  Enables a standard definition—or loosely, one scoreboard—of SEC data that marketers, product developers, business decision-makers, and policy makers can refer to in planning and in making action standards.    This eventually paves the way towards an equivalent or individual (per country) SEC; one that is more relevant in the age of digital commerce.
  2. Granularity in Marketing Segments Analysis.  Allows finer analysis of economic and social groups, especially in the context of product consumption and regional location.  Urban-rural, regional, and cluster groupings form potentially rich ground for cross-discipline research (i.e. migrant worker groups, impact of education, poverty alleviation, effective media and advertising content, life stages and culture, population density, and social cohesion, etc.)
  3. Ability to Track the New (and evolving) Philippine Middle Class.  This facilitates more detailed groupings of the middle class and transition groups, i.e. from extreme low expenditure to mid-tier, and from mid-tier to the highest expenditure groups.   The HSBC 2040 Study projects the Philippines to be within the top 20 ranked economies in two decades. Imagine the swelling of the middle tier clusters from the present year onwards.
  4. Spawns a Strong Marketing Information Ecosystem. Allows for the spread of the culture of data usage among field research professionals, e-commerce marketers, retailers, statisticians, and data engineers and scientists across all B2C domains. This motivates sharing of best practice and continuous improvement programs, eventually benefiting the broader public.
  5. Harmonization of Marketing Knowledge.   A coherent household classification system sets the stage for a marketing industry that can be empowered to keep its members updated and current. Economic and marketing bodies who use the harmonized data would be perceived to be more trustworthy and consistent. An offshoot is increased reliability for research-based decisions, nationally and regionally.

Both the Philippine measurement sector (data analytics, research, media metrics) and the B2C marketing sector, as well as the government itself through the PSA, have embraced the need to join forces in harmonizing SEC data standards.

Expect more partnerships like this as technology, data science, and marketing rapidly converge in this age of disruption that on one hand upends traditional knowledge, but, on the positive side, inspires more artful integration.

About the author:

Nicco de Jesus has over 25 years of experience in data analytics, marketing research, and business analytics in local and multinational corporations in the Philippines. He was formerly a business director of one of the world’s leading research, data and insights brands.

Nicco has actively participated and led industry associations and societies, including Marketing and Opinion Research Society of the Philippines (MORES), UP School of Statistics Alumni Association and Digital Measurement Board (DMB).

In Pointwest, Nicco provides thought leadership on Data Science, continuously develops the capability to work on projects requiring Data Science, and works on business models using data-centric business concepts for various markets.