Experian

The functions that Aperture Data Studio can perform are detailed in Figure 2. Notable are the enrichment functions that can leverage other aspects of Experian’s data product portfolio for things like geospatial, demographic and credit-based information.

Experian Aperture Data Studio is browser-based, has strong data workflow capabilities and supports user personas. The product has been designed to work within existing technology stacks. Thus, for example, the product’s data preparation capabilities which are largely targeted at operational rather than analytic environments are complementary to both data cataloguing and data governance solutions from third-party vendors such as Collibra and Alation. The software will store relevant metadata defining such integration.

As far as data workflow is concerned, Experian Aperture Data Studio separates data profiling (which is done on all your data, not just samples) from data loading. Thus, data profiling is an explicit workflow step or action during data discovery, and therefore distinct from data manipulation and load processes. Moreover, during profiling you can choose which columns to profile rather than profiling the whole table, and interactive drilldown is supported. Address and email verification can also be defined as steps within a workflow, as can functions you have written using the comprehensive Software Developer’s Kit (SDK). Further, it is worth commenting that you can explore and drill-down from a workflow without having to run a workflow, which is not the case with some other products. This means users can save time by checking the workflow without having to spend time running the file. Other features include the ability to chart results – compare data between runs and undertake version management for workflows. Exhaustive object versioning is a new capability introduced within the latest release and this also supports reversions. Data workflow annotation is supported as well as the ability embed workflows within other workflows.

As mentioned, Experian Aperture Data Studio is persona-based, and there is a particular emphasis on business users and collaboration. In particular, the company there is the concept of “spaces”, where you can define and use functions that are specific to a particular area such as a marketing department. Experian is increasingly adding pre-packaged sets of these functions, for example for credit information quality checking or single customer view templates. Individual functions may come out of the box or you can define your own. As an example, you might have a function that obfuscates credit card numbers. If you want to use something more sophisticated, such as format preserving encryption then there is an SDK that will enable this. Alongside spaces, Experian also supports pre-configured views, consisting of trusted data, that are specific to a particular space or sub-space, such as a “view for marketing” so that the marketing department only sees the data that is relevant to its function.

Fig 3 - An example of duplicate matching rules

From a matching perspective an example of duplicate matching rules is illustrated in Figure 3, These are applied when looking for and evaluating potential duplicates. Experian Aperture Data Studio does not apply machine learning to the matching process itself, but it does apply machine learning to matching rules to obtain the best set of those rules.

After investigating the first release of Experian Aperture Data Studio we wrote that “most vendors, in any market, try to meet modern requirements by bolting on extra capabilities. If that is simply a question of adding on a feature here or there, that is not problem. However, when it comes to fundamentals such as self-service and collaboration, these are not the type of services that are amenable to bolting on. Experian is therefore to be applauded for biting the bullet and developing a product suitable for the third decade of this century.” This is the second time that we have reviewed the product since its original release and it continues to build on its early promise.

The Bottom Line

Experian Aperture Data Studio is a modern application for data management and offers the sort of features that businesses require from such a solution. We said something similar in our first review of the product but were forced to qualify our valediction because of the lack of some capabilities. Since then we have happy to remove that qualification: Experian Aperture Data Studio is a product we are happy to recommend.

Experian has recognised that there are four trends within the market related to operationalising analytics: a demand to use machine/deep learning models for predictive functions; issues around the operationalisation of those models; requirements around the monitoring, management and replacement (or retraining) of models; and governance issues with respect to both explainability and bias. As a result, Experian is in the process of enhancing PowerCurve, which was originally released in 2011, to support these capabilities. For example, it has introduced the ACE Framework. This provides an environment for plugging in predictive models developed within third party environments. Currently ACE supports PMML (predictive modelling mark-up language) based models as well as those developed using R or Python. H2O support is forthcoming and other plug-ins will be introduced based on customer demand.

However, there are some elements of what Bloor Research calls AnalyticOps that are not yet fully implemented, though Experian is by no means alone in this. For example, there are facilities for model auditing today, but it is not an easy and simple process. Experian plans to enhance both its explainability and model governance capabilities, adding model performance monitoring. While there are facilities for automatically deploying and testing models into production there are no workflow/approvals processes to complement this yet. The company is considering whether to implement such processes within PowerCurve or to leave this to third party tools.

AnalyticOps concerns the operationalisation, monitoring, improvement, management and governance of augmented intelligent models. As such models become more and more widely deployed, AnalyticOps capabilities will become more and more vital. Where most companies today have a handful of deployed models, if that, we expect large organisations to have thousands or tens of thousands of models to deploy and manage in the near future.

In practice, there are very few vendors that can address all, or even some, of the AnalyticOps requirements of a modern data science environment. Experian PowerCurve is one of these: it has  a business user-friendly design environment enabling business users to easily include models in decision strategies and is in process of adding and enhancing capabilities for identifying bias, model explainability, model performance monitoring and model management. In this context it is noteworthy that the company supports a methodology known as FACT for fairness, accuracy, (responsibility to the) customer, and transparency (including both explainability and auditability). By way of contrast, most vendors in this space have no model management capabilities at all, few have any capabilities with respect to bias and explainability is by no means common.

The Bottom Line

However, PowerCurve has not historically been marketed as a generic decisioning platform but only as providing underlying capabilities to support the various decision management solutions offered by Experian. While the latter remains a valid strategy, it means that PowerCurve remains a well-kept secret. Given that many of the well-known data science platforms cannot match the capabilities of PowerCurve when it comes to AnalyticOps – at least at this time – we would like to see PowerCurve marketed more aggressively to a wider market.

Fig 1 - Experian IntoZetta coverage

At the heart of the offering is a data dictionary which stores information about business data entities and data flows. Technical metadata can be ingested from relational database catalogs and other sources such as Sharepoint, and used as a basis for developing a catalogue of business terms such as “customer” and “asset” and their attributes. These business entities can be assigned data owners and be classified as desired, for example to note whether the data contains personally identifiable information that may require greater care than other data. The dictionary allows users and roles to be defined, such as data stewards, data owners etc and have rules developed around these roles. Data profiling is carried out to further flesh out the metadata, and data quality targets can be established.

There is an attractive visual interface to show the data at an overview level. Data flows can be defined and displayed in a visual form too, with basic data lineage shown. Customers can, for example, see all things associated with an entity like “customer”, such as related data, who is the assigned owner for it etc. The tool can interface with specialist data lineage tools for customers who need this. A link exists to Octopai and one is planned for Manta. The technology has a range of dashboards and reporting capabilities. For example, data quality progress can be monitored, as can the coverage of rules and ownership. There is a high-level report that neatly captures overall progress of a data governance initiative at a high level, drawing on the underlying information in the data dictionary. The user interface is intuitive and aimed squarely at business rather than technology staff.

The IntoZetta technology already has some linkage to the Experian Aperture Data Studio suite, so for example some data quality functionality from Data Studio can be invoked from within IntoZetta and the results returned into the IntoZetta Data Dictionary. In addition, Experian Aperture Data Studio can push profile metrics out into IntoZetta, as Experian can with its data governance solution partners. Over the next year or so it is likely that the integration will go deeper.

Currently, the data governance market leaders at the high end of the market have a top-down approach to corporate data governance, as seen in tools like Collibra and Alation. While Experian partners, and will continue to partner, with several of these solution providers. It is likely that Experian will position their data governance tool so as to be suitable for a more bottom-up approach. This may be more suitable for projects within business lines or departments. It will also be well suited to small and medium-sized entities. This would mirror the Experian approach to other software such as its data quality suite, and so would fit easily into its existing sales and marketing channels.

The Bottom Line

Experian is a large and long-established data quality vendor, and given the linkage between data governance and data quality it makes sense for Experian to bring out some sort of data governance offering. By purchasing IntoZetta they can build on an established product that has proven its worth in a number of customers and integrate it into their existing Aperture Data Studio product line. Customers looking for a business-focused data governance tool that is aimed at either SMEs or is suitable for departmental data governance solutions should take a look at the new Experian offering.

Fig 02 - Experian Visual associations

The Experian data quality software has a broad range of functionality. It can profile data to carry out statistical analysis of a dataset to provide summary information (such as the number of records and the structure of the data). The tool can also discover relationships between data and check on anomalies in the content. For example, profiling would detect whether a field called “phone number” had any records that did not match the normal format for a phone number or had missing data. The software can set up data quality rules that can be used to validate data, such as checking for valid social security numbers of valid postal codes. It can detect likely duplicates and carry out merging and matching of such fields, for example, merging files with common header names.

The software can enrich data too. For example, you might have a company with an HQ at a business address. As well as just validating the address, the software can add additional information that it knows about that business. At a basic level that might be something like adding a missing postal code, but it can be much more elaborate. Experian in particular has a rich vein of business data in its wider business, so could add information about the company such as the number of employees.

Data quality is a long-standing issue in business, with various surveys consistently showing that only around 30% of executives fully trust their own data. Poor data quality can have serious consequences, from the inconvenience of misdirected deliveries or invoices through to significant regulatory or compliance fines.

GDPR non-compliance can result in fines of 20 million euros or 4% of annual turnover in the EU, and there are numerous other regulatory bodies in the USA and elsewhere that apply fines for poor data quality in various industries such as finance (for example the MiFID II regulations) and pharmaceuticals (CGMP rules in the USA).

Another recent trend is likely to raise data quality higher up the corporate executive agenda. Companies, inspired by the success of ChatGPT and others in the public consciousness, are widely deploying generative AI technology in assorted industries. However, they are increasingly finding that they need to train such large language models on their own data (and operate private rather than public versions of such AIs for security reasons). It turns out that the effectiveness of such models is highly dependent on the quality of the data it is trained on, echoing the old “garbage in, garbage out” rule of computing. It is therefore important to assess and usually improve the quality of corporate data before a broad roll-out of generative AI technology.

The bottom line

Experian has a well-proven and comprehensive set of capabilities for data quality. It has expanded significantly in functionality in recent times, with its data governance capability the latest manifestation of this. With the excitement about deploying generative AI in corporations turning out to be heavily dependent on good data quality, vendors such as Experian should see renewed interest in their technology.