05 / 06 / 2019
Truata Nominated for two National Fellowship and Industry Awards
THE NATIONAL FELLOWSHIP AND INDUSTRY AWARDS NOMINEE
Truata are delighted to have been shortlisted in two categories for The National Fellowship and Industry Awards. We have been shortlisted in both the Implementation of AI Award and Emerging Tech Product categories.
Companies, across all industries, are under increasing pressure to become more data-driven. They’re analysing customer data to understand and improve their processes, products and services to make decisions and drive their businesses forward. These data-driven decisions are critical to maintaining a competitive advantage.
Meanwhile, as a result of the GDPR, many companies have been forced to curtail or cease valuable analytics programs. Others have engaged in in-house anonymisation. However, since these companies hold the original source data they are likely only producing pseudonymised data. Thus, falling short of the threshold set under the GDPR.
The solution is a set of processing steps designed to identify and mitigate privacy risks through a combination of manual and automatic actions.
The Trūata Anonymisation Solution is depicted in Figure 1 below.
Truata’s Anonymisation Solution brings customer data on the following journey, with each step below corresponding to the numbered step in Figure 1:
1. Customer removes personal data and tokenises direct and indirect identifiers then transfers to Truata
2. Truata tokenises direct and indirect identifiers
3. Truata runs risk assessment routines to identify privacy risks and applies anonymisation techniques to data
4. Truata moves data to customer specific data storage and performs analytics to generate reports and model code
5. Aggregated reports and model code provided to customer. Customer can use those insights to improve their business processes and better serve their customers
6. Truata analyses the anonymized data using tools selected based on each individual customer’s analytical needs
7. Customers receive analytics, algorithms and aggregated reports back from Truata for use in their products and solutions
8. Customers apply those analytics, algorithms and reports to unlock powerful, value-generating insights from their data
Along this data journey, data science and advanced machine learning techniques are employed.
2. Personal data identification. As part of the risk assessment process the entire dataset is scanned to find personal data. For certain types such as email or IP addresses, well known patterns are used while for types such as postal addresses or the names of locations or people, AI is used. This is especially true for unstructured data such as comments or user provided data. Natural language processing techniques are used to find personal data so that it can be anonymised.
3. Dynamic Outlier handling. Outliers are data fields or records whose value differs from the majority of others to a significant degree. Outliers are more distinctive and thus have a higher reidentification risk. Basic approaches to outlier handling involve calculated fixed thresholds and removing or adding extra processing to fields whose values lie outside the threshold. The Trūata solution takes a more dynamic approach, recognising that for any given query what is considered an outlier might change. Trūata uses AI to dynamically decide what represents an outlier when a query is made so that appropriate thresholds are used and privacy is preserved while optimising analytic output to the highest degree possible.
4. Privacy-conscious analytics tools. To empower customers to make data driven decisions as normal, Trūata provides tools that customers are familiar with, such as data science notebooks. Using these, customers can train and build machine learning models over their data. Supporting these AI applications while also preserving privacy requires specialised tools using machine learning to analyse a dataset and make intelligent recommendations around features that can be utilised, for example, in building models.
Truata’s solution powers client’s predictive analysis so their acquisition and retention programs can be optimised. It facilitates longitudinal studies, as data retention limitations don’t apply to non-personal data, allowing developmental trends to surface. It ensures companies can analyse larger portions of their data universe, resulting in less biased, more powerful analytics. All of these increase the accuracy of insights leading to more efficient marketing spend and the strengthening of brand preference translating to higher margins and market share.