Five common data stream problems

4 minutes read
30 November 2020

Today, supplying end-users with truly useful and efficient digital services depends on the ability to immediately collect and process data streams generated by customer interactions with the brand, through applications and contact channels. This data must be organized and made available to applications used in business or consumer environments in real-time, in order to generate precise and timely responses to customer needs and continuously improve the value generated.

Every moment, the multiple data sources – applications, physical and digital contact channels, IoT devices, etc. – quickly and continuously generate large volumes of Fast Data, that – if not correctly and immediately managed, filtered, and made available – lose their value. For this type of data, the solution is to create data streams.

For example, in the GDO context, where huge volumes of data are administered on a daily basis, it is essential to know how to manage a continuous flow of data to obtain valuable information.


What is a data stream?

The term data stream means a flow of heterogeneous data, coming from multiple sources, produced in real-time, and aimed at making the information immediately available and usable.

A data stream is generated by adopting the Fast Data paradigm and aggregating information in Single Customer View: this is possible by creating an event-driven architecture based on event streaming platforms such as Apache Kafka.

Single Customer Views are unique, always updated, 360-degree views of the profile information and history of purchase and interaction of each individual customer. By building Single Customer Views, data streams allow keeping updated all contact channels and applications, both internal and external. Furthermore, they make it possible to make business decisions based on accurate and updated information, from which it is possible to extract the maximum value when needed.




Common technical and business problems with data streams

The implementation of an architecture capable of managing data streams is a complex project because data is continuously generated from a large variety of sources (IoT sensors and devices, IT systems, websites, social networks). These sources dispense data using many different types of formats. Among the technical problems that can slow down data streaming projects there are some main ones:

Errors caused by duplication, anomalies, data inconsistency

For example, the data acquired by IoT sensors can often be ‘dirty’, and contain ‘out of range’ readings, null or duplicate values, syntax errors, which require cleaning.

Problems of non-homogeneity of data present in different systems

The fragmentation of data across multiple business systems creates standardization difficulties. Even applications, produced by different manufacturers, could operate using different schemes and data structures, which end up corrupting the data pipeline.

Disaster recovery challenges

In case of catastrophic events, it is necessary to organize a disaster recovery plan to restore the functioning of the data streaming application, typically implemented on server clusters, which can be distributed across multiple data centers.

Architectural approaches for data streams

Mia-Platform offers a Fast Data solution that allows you to create a data streaming platform capable of anticipating or solving the technical problems just mentioned, by applying, for example, rollback mechanisms and management strategies aimed at ensuring that data is always clean and consistent throughout the system.

Mia-Platform solution allows decoupling contact channels from IT systems, through a digital layer, composed of different services, which acts as a Digital Integration Hub.

The data flows into a stream and is immediately aggregated into JSON Single Customer Views (SCV) by a series of specially created microservices. Consequently, SCVs are exposed to various applications and channels.

Channels can call APIs to access unique views, or receive a push notification for changes, following the CQRS (command query responsibility segregation) architectural approach.

The data is saved on a high-performance, low-latency database, which always keeps the information updated and accessible to the channels, regardless of the availability of the underlying systems. All the operations on the channels involving the IT systems are mediated and performed asynchronously.



A stream of constantly updated data and a single view of all customer information allow us to provide timely, accurate, and valid answers to customer needs. Choose Mia-Platform Fast Data to build your data streaming platform and unlock the real value of your data.

Back to start ↑
What is a data stream?
Common technical and business problems with data streams
Architectural approaches for data streams