General kMatrix Sector Methodology

The kMatrix research methodology started as an international research programme within Harvard University and was subsequently developed in countries across Europe. The core research processes within the methodology are used to track technology and market change and employs “big data”, analytical industrial data grids, business case studies and industry- level research to provide probabilistic and measurable evidence about how companies and economies adapt to and perform within changing market conditions. 

 

This multi-source research methodology is most often employed within the public or private sector when:

  • An emerging or newly “discovered” sector needs to be defined and quantified
  • An existing sector needs a more flexible and detailed segmentation than SIC can provide
  • International market information and national benchmarking are required

 

In these circumstances the methodology provides:

  • Internally consistent global research, based upon market intelligence sources and rules rather than national statistics, that can be used to measure economic sectors that are not well represented by Standard Industrial Classification (SIC) codes or industry bodies
  • Flexible/adaptable data structure with multiple levels of detail and multiple measures for benchmarking competitiveness and identifying international markets and market opportunities.

 

The methodology reflects best practice in private sector research processes like market research, competitive and competitor intelligence gathering and analysis - as well as more traditional public sector economic assessment methodologies. It has developed from long term experience of selecting, monitoring, evaluating, triangulating and then analysing multiple unstructured, semi-structured and structured data sources to produce quantified values, forecasts and economic/industrial indicators with explicitly measured levels of data confidence.

 

The fundamental concept of the process is to use triangulation, similar to that used by satellites for navigation, to look at, in and around a sector, using robust and recognised data sources, including unusual sources such as recruitment and investment information (some in the public domain, some not) which are then used to calculate new values for sales, exports, number of companies, employees and over 65 other metrics. This process is described and demonstrated in detailed methodology documents for various sectors including security, low carbon, green economy, digital media, construction, healthcare, climate services etc. and, despite the innovative nature of the methodology, was awarded UK statistical status in 2013. Analysts from Greater London Authority have assessed the process and data and approve its use. 

 

The flow chart in figure 1 gives an outline of the kMatrix profiling system, the same basic process is used, with slight variance depending on whether it is a full sector being mapped, or a single technology being mapped through a supply chain.  Some sectors are easier to map than others, for example, Climate Services is an emerging sector, albeit using existing competencies, which is difficult to measure, with sparse data sources available.  In contrast, construction is a well-established sector, with a wide range of high quality and robust data sources.  Sources selected varies, with never less than 7 sources used per data value.

 

Figure 1

Why this methodology?

 

There are three distinctive approaches to industrial sector analysis and each have their own merits. They are:

 

1. The national statistical approach - can be generic, slow moving, structurally rigid and objective as befits its role as a national score-keeper. It has longevity, consistency and credibility and a widely experienced base of analysts and users familiar with its interpretation and application. But it suffers the limitations and inaccuracies of self-classification against a list of generic product/service descriptions leading to an inherent level of inaccuracy.

 

2. The Trade/Industry approach - generally has the advantage of agreed sector boundaries (that may be different to SIC definitions), higher levels of accuracy within those boundaries (is internally consistent) and some benchmark standards for performance against competing nations and their flagship companies (can be outwardly focused). The limitations are that data collected in this way takes many years to achieve consistency and credibility, is only available where strong associations exist, is subjective (collected and reported for the benefits of its membership) and are inherently conservative or self-interested.

 

3. The multi-sourced approach - has the advantage of being flexible, rapid, objective and can operate in areas where the above two approaches either do not exist or are poorly aligned. The limitations are that this approach is not directly comparable with traditional statistics (numbers have to be created, not just collected), is inherently complex and requires mapping to and reporting out through SIC or ISIC codes if a direct comparison is required.

 

The kMatrix Sector Research methodology falls clearly within the third category of multi-source analysis, triangulating data from multiple sources, taking into account past reliability or sources and creating new values. It provides flexible, rapid, objective data that can be easily updated. 

 

 

How does the methodology correct deficiencies in the source data sets and what makes the evidence base unique?

 

The methodology illustrated in figure 1 includes a source selection process, part of which involves assessing the past reliability of the source.  We use multiple sources for each point of data, dismissing outliers and triangulating a value, which is then allocated a confidence level.  The confidence level is a mathematical function of the spread of values across a range of sources that we include in our analysis.  Confidence levels can be applied to all metrics, but are usually applied to sales, number of employees, transaction levels and growth forecasts.  Confidence levels vary by activity, geography and by forecast year.  Typically, a confidence level of above 80-85% is achievable.

 

Two key strengths of the kMatrix' approach are its taxonomies' flexibility to report the activities within any agreed sector definition according to one or more organisational structure and for sector definitions to be expanded (horizontally or vertically) or reorganised easily, to keep pace with industry change through its lifecycle. Together, these features ensure that both old and new definitions of a sector can be reported on consistently for time series trend analysis.  Other differentiators include the ability to report the performance of a sector at global (226 countries and territories representing more than 90% of world trade), country and (for larger countries including the UK), regional and sub-regional level.

 

Ensuring robust data

 

Our proven global “big data” approach relies upon using multiple but robust sources of data to create values where relevant government or industry statistics do not exist. We do not take in data through the SIC system, however we have the ability to report out through the SIC classification if required. Sources include published reports, sector analysis, patents, papers, ONS information, other government information, websites, social media and grey publications but it will includes data from legal, financial, investment, academic and other sources. Research methods include:

• Desk research to define sector content and determine sector boundaries

• Industrial templates that identify core and supply chain activities in detail for inclusion in sector definition

• Data discovery tools to identify new data and sources relating to the defined sector activities

• Data coding systems to ensure that sector, company-level data and other classification systems are aligned for analysis and reporting purposes

• Data management techniques and systems to maintain existing source libraries and integrate them with new source materials

• Software systems with defined (but flexible) rulesets to filter source content

• Semi-automated processes for modelling and calculating data values from selected source lists

• Knowledge base of case study materials that can be accessed to help fill data gaps and provide industrial performance benchmarks

• Quality assurance processes and tools that check all values against a range of international, national and industrial comparators.