The Data Science Procurement Dilemma — Build, Buy, Or Outsource?


Singapore, January 31, 2018


Organisations are increasingly convinced that Data Science will be a critical component of their business and are clamouring to adopt it. Nevertheless, it is vital to keep in mind that the wrong approach may result in catastrophic consequences. It is therefore essential that organisations understand what their options and respective implications are.

To address the concern, Gartner established a list of Data Science solutions procurement guidelines in its recent “Machine-Learning and Data Science Solutions: Build, Buy or Outsource?” article.

When making Data Science solutions procurement decisions, organisations often experience the following challenges:

Machine-Learning and Data Science Solutions

Source: Machine-Learning and Data Science Solutions: Build, Buy or Outsource? by Gartner

To determine the most suitable procurement option, organisations should take following into consideration

·         The business problem to be resolved

·         The organisation’s Analytics maturity

·         Availability and scale of staff with relevant skill

·         Time required to solution

·         Solution implementation urgency/business criticality

·         The expected (or required) return-on-investments

·         The amount of flexibility and control needed

·         The organisation’s appetite and budget for Analytics

·         The availability of specialised and readily available tools (COTs)

The diagram below summaries the available procurement in face of the various circumstances.

available procurement in face of the various circumstances

Source: Machine-Learning and Data Science Solutions: Build, Buy or Outsource? by Gartner

Option 1: Build

To “Build” a solution using Data Science platforms means developing the capability in-house.

To adopt this option, organisations should possess skilled staff with deep understanding of how to build Advanced Analytics solutions on the selected Data Science platform.

This option comes with the assumption that the organisation already has an in-house Data Science team, and has decided on a selected Data Science platform. Neither deciding whether to develop an in-house Data Science team, nor deciding which Data Science platform to work on, are simple tasks.

To build a successful Data Science team, it is important to recognise that attaining the right talent composition is not a trivial matter. Our previous post on “How to Build a Data Science Team in the Public Sector?” provides some guidelines for consideration.

Once the organisation has decided to develop an in-house Data Science team, the next natural question will be “which Data Science platform should we adopt”?

Let’s put aside open-source solutions like R and Python for the time being.In Gartner’s “Magic Quadrant for Data Science Platforms, 16 vendors were evaluated, using the 15 Data Science platforms capabilities evaluation criteria:

1.    Data access: How well does the platform support access and integrate data across sources and types.

2.    Data preparation: Does the product have an extensive range of coding or noncoding features to prepare data for modeling?

3.    Data exploration and visualisation: Does the product support data exploratory and visualization?

4.    Automation: Does the product facilitate automation of feature generation and hyperparameter tuning?

5.    Interface: Does the product have a coherent and usable user interface?

6.    Machine learning: How extensive are the product’s machine-learning approaches (pre-packaged or accessible from the product)?

7.    Other Advanced Analytics: How are other methods of analysis (involving statistics, optimisation, simulation, and text and image analytics) integrated into the development environment?

8.    Flexibility, Extensibility and Openness: How well does the product support the user’s needs (e.g. open source, development of own functions)

9.    Performance and Scalability: How well can the deployments and configurations be controlled for performance and scalability?

10.  Delivery: How well does the platform support the ability to create APIs or containers that can be used for faster deployment in business scenarios?

11.  Platform and Project Management: What project essential capabilities does the platform provide (such as for security, compute resource management, governance, version management of projects)?

12.  Model Management: What capabilities does the platform provide to monitor and recalibrate models (e.g. K-fold cross-validation, champion/challenger [A/B] testing)

13.  Precanned Solutions: Does the platform offer “precanned” solutions (e.g. social network analysis) that can be integrated and imported via libraries, marketplaces and galleries?

14.  Collaboration: How do users with different skills work together on the same workflows and projects?

15.  Coherence: How intuitive, consistent and integrated is the platform to support an entire data analytics pipeline?

Option 2: “Buy” commercial off-the-shelf (COTS) solutions

Specifically in Asia Pacific/Japan (APJ), in Gartner’s “Market Guide for Data and Analytics Service Providers, Asia/Pacific and Japan”, Gartner reached out to a sample of renowned APJ-based Data and Analytics service providers to respond to a short survey, about their capabilities in Data and Analytics services in the region.

Data and Analytics services in the region

Source: Market Guide for Data and Analytics Service Providers, Asia/Pacific and Japan by Gartner]

Market Guide for Data and Analytics Service Providers

Source: Market Guide for Data and Analytics Service Providers, Asia/Pacific and Japan by Gartner

Option 3: “Outsource” building a solution to an Analytics service provider or freelancer

Similar to the “Buy” option, should the organisation decide to embark on the “Outsource” option, Gartner has identified and interviewed the following list of APJ-based analytics service providers. This list of COTS and service providers serves only as a starting point for evaluation.

building a solution to an Analytics service provider

Source: Market Guide for Data and Analytics Service Providers, Asia/Pacific and Japan by Gartner

The above illustrations give a high-level overview of APJ-based Analytics landscape. For evaluation, organisations will need to consider many other dimensions, including the following

·         The problem(s) to be solved

·         The service provider’s records of accomplishment

·         The service provider’s ability to provide good local support for implementation and maintenance

·         The service provider’s ability to understand your industry and business needs.

·         The service provider’s ability to customize the offerings to suit the organisation’s variable/future needs.

·         The service provider/COTS’s ability to integrate their offerings with your existing architecture (e.g. programming languages, APIs)

 

Keen to arrange a Data Analytics demo or to find out more?
Drop us a note at ai@ncs.com.sg