Reference Architecture for Customer Analytics

Each time your customers interact with you, on whatever channel they choose, they reveal their likes, dislikes and interests, as well as their needs and motivations, leaving behind a trail of data for you to follow. It’s no wonder they expect you have an intuitive understanding of what they need, but are you in a position to deliver?

Acrotrend: Keeping you ahead of the curve
A great customer experience relies on your organisation having the capability to gather customer data efficiently. You also need to know your data is high quality and have the skill to drill down to a single customer view for a complete understanding.
In addition, your customer facing teams need to have deep insights about any given individual, so they know their loyalty level and ongoing engagement with your various brands and how this has evolved over any given period of time. Your team then needs to be able to take relevant action based on those insights, such as communicating with them at the perfect time to deliver the right offer or message.

With the constant shifting of market trends, it’s important for you to be able to anticipate positive and negative movement, share and discuss these insights with stakeholders before experimenting and course correcting rapidly at the right moment. What’s more, you need these capabilities on an everyday basis for it to have maximum effect on your business, and the ability to scale these up and down according to need without wasting time or money.
It’s a big ask.
Acrotrend has the solution. Our customer analytics reference architecture organises your data so you can ask key business questions and generate useful insights, enabling you to act on them at the right time in the most effective way. Our reference architecture doesn’t just address your current needs. It future-proofs your investment across a range of analytics. Our reference architecture gives you the key capabilities to making a successful customer analytics journey whenever you need. You can break down customer data silos to deliver an enhanced customer experience by giving relevant, complete information to those who need it, including:
- Marketeers
- Sales reps
- Customer service agents.
While there are many tools and technology platforms available to fulfill your reference architecture, based on our extensive experience, we’ve picked what we consider the best-fitting ones to make up the stack. This means that not only do you save time trying to make sense of which will work best for you, you know the ones we recommend integrate effectively with each other.


Data Integration and Cleansing – Talend
Talend supplies the data fabric platform covering all capabilities outlined in our reference architecture. This platform expedites the collection and preparation of a wide range of data for any campaign, sales meeting or service, both real time and batch.
In addition, Talend includes:
- Cloud API services
- Data preparation
- Data profiling
- Data stewardship
Since Talend can be hosted on the Cloud, this frees up your IT department to focus on other areas of business instead of needing to provide support and maintenance. In addition, Talend supports machine learning to de-dupe and cleanse customer data, due to its native integration with Apache Spark. This can greatly help speed up the unification of customer records from multiple business systems, allowing you to build single customer view.
Talend is based on an open source stack that allows it to build valuable accelerators so you can ask even the most unusual questions of your data. In addition, Talend’s recent acquisition of Stitch gives you a readymade cloud-based platform to integrate data from a variety of sales and marketing sources. As if that wasn’t enough, the platform works natively with data warehouse platforms for Snowflake and Amazon Redshift, our recommended technologies for Cloud data warehousing.
Cloud Platform – AWS
You need to have a robust cloud platform underpinning your reference architecture. Our choice of AWS was due to 3 factors:
- We have been working with AWS for 7+ years and our consultants have in-depth expertise backed by relevant certifications on data and machine learning.
- All the key components of our reference architecture are directly or natively supported within AWS. This includes Snowflake, Looker and Talend.
- Most importantly, with the enablement of our reference architecture, AWS enables us to achieve our philosophy of data, insight, action to close the loop for our clients.
Besides these factors, the other use cases required for our architecture are fulfilled by AWS. Kinesis provides the streaming alternative directly supported by Snowflake, Redshift and Talend. S3 for storage is natively supported across the stack. Finally, API Gateway enables clients to flexibly define APIs to activate actions from their insights directly into their sales and marketing systems.
The other advantage is the high degree of cost optimisation AWS provides across components. We have been using AWS Trusted Advisor ourselves and across our clients to give a clear and honest picture of what components are being used and where savings can be made. Moreover, AWS shares our philosophy of being customer-centric and provide the best support for our clients.





Data Warehousing – Snowflake and Amazon Redshift
With modern technology allowing data and customer actions to be recorded in near real time, traditional data warehouses can no longer keep up. We firmly believe your data warehouse should be intelligent enough to be self-recuperating without the requirement of an army of personnel overseeing daily maintenance.
This is why a Cloud based data warehouse is the only solution.
Snowflake is a data warehouse that is genuinely cloud native. It supports standard SQL, a favourite with analysts and data scientists alike. It can be commissioned in minutes without the requirement to build an infrastructure. It can scale up or down based on usage and concurrency, giving you a great way to spread the load across teams. With support available for both structured and semi-structured data, it can support a high number of use cases.
Alternatively, choose Amazon Redshift, one of the first scalable data warehouses released on Cloud.
Redshift’s architecture differs to Snowflake’s, since it is primarily based on PostgreSQL database. If you are embedded deeply on AWS with a focus on its integration solution using AWS Glue platform, databases on RDS & querying using Athena, we would recommend Redshift, as it provides the out of the box flexibility you need for integrating data.
Ultimately, the choice you make depends on your needs and we always recommend an assessment and PoC to ensure you make the right choice, since the data warehouse sits at the heart of a data driven organisation. When carrying out a data warehouse assessment, we consider several parameters, such as:
- Alignment to business and enterprise architecture objectives
- Cost benefit analysis
- Concurrency
- Performance and accessibility
Machine Learning and Data Science – AWS Machine Learning, Python and R
Having an ML and data science capability gives you further opportunities to carry out targeted cross-selling and up-sell revenue.
Python has been the number one choice for years, given its excellent frameworks to work on different use cases on machine learning combined with extending that language knowledge into areas of web application. In addition, it helps with data prep and training the model.
Though Python is our go-to language, R, with its community and packages, is the usual option for our data scientists to solve ‘math-heavy’ problems. Both Python and R based packages provide you with the ability to model advance predictions for churn, LTV and drive strategic decisions to increase revenue and growth.
AWS provides an excellent platform to deploy our ML models in a secure and scalable manner. The ease with which Sagemaker can do this was incredibly exciting for our engineers. It doesn’t require any manual provisioning of containers behind the scenes or managing the performance – Sagemaker does that. This gives you high cost savings, with engineers focusing on value added activities.







Data Visualisation and Reporting – Looker and Microsoft Power BI
With a Cloud-based architecture and strong governance via its mark-up language, LookML, we believe Looker is well placed to replace monolithic BI platforms. It provides the semantic layer missing in self-service platforms thanks to its strong Git based integration for collaboration and rapid changes. With the rise of Cloud-based data warehousing platforms such as Snowflake and Amazon Redshift, Looker fits perfectly on top of these scalable platforms to discover and explore your data.
While Looker gives you a data platform, Power BI offers an excellent self-service alternative. Its ease of use and packaging with Office 365 gives you a powerful alternative. Power BI supports a wide range of sources, so you can connect and directly start working with them. We’re confident that Microsoft will continue to deliver a wide range of visualisations and innovations, making Power BI a superb alternative to Looker for businesses getting started on their data driven journey.
The key differences between Looker and Power BI are:
- The strong governance layer
- Native integration with version control
- The native action component to close the loop
Marketing Automation & Segmentation – Salesforce.com and Eloqua
Data and insight are useless unless they’re acted upon. For instance, if ML algorithms can discover segments using the sales, marketing and transactional data, these segments have no value unless the marketing teams can execute campaigns based on them in Salesforce Marketing Cloud or Eloqua. This requires integration of the segments with your marketing system.
Our domain expertise with sales, marketing and customer service, alongside technology accelerators to integrate critical insights back into the systems, can help you close the loop from the insights discovered from your data.
Action is the critical part of this process that is frequently left out and, in our experience, never executed effectively. These platforms are not meant to be data platforms. Although they are excellent in offering operational analytics within their domain, integrating and analysing data from your different silos into such platforms can be cumbersome, leading to higher costs.
Our approach of managing the data and insights within specialised platforms to support appropriate action provides benefits to your organisation, leading to the delivery of a cohesive and harmonious customer experience, resulting in higher NPS/CSAT scores and brand loyalty.



Delivering the level of service you need
Our reference architecture, together with the technology stack, supports our core tenet of building a culture of curiosity by asking unusual questions of our data, enabling the full cycle of data to insight to action. We are confident the right stack will power the next generation of digital and data transformation so your business can fully benefit from everything these tools have to offer.
While you might not initially need all capabilities, we’ll design a roadmap that considers all possibilities at-build based on your priorities, giving you the greatest chance of success.