About
PT Telekomunikasi Selular is a public Indonesian mobile operator established in 1995 and a subsidiary of Telkom Indonesia. With its headquarters located in South Jakarta, Telkomsel is the largest mobile operator in the country, with more than 160 million customers.
The Problem
Telkomsel encountered a set of challenges unique to the retail sector. Namely, the difficulty present in traditional retail stores regarding obtaining insights into the competitive landscape.
The process of gathering intelligence proved to be labour intensive and produced varying levels of accuracy. In addition to the excessive and inefficient semi-manual work hours, the management dashboards failed to display data in real-time, thus complicating and interrupting the predictability of sales, specifically in more remote areas.
The Goal
Devoteam G Cloud experts suggested the implementation of Google Cloud products to help Telkomsel achieve their goals of efficient retail intelligence.
The recent years have seen tremendous progress in the area of deep learning and neural networks. Various AI’s have surpassed expectations and demonstrated the potential of machine learning technologies. Countless AI applications have dramatically improved the performance of everyday life, in our homes or our day to day work, be it how to predict and analyse customer churn or apply intelligence to our retail. Google is investing heavily in these technologies, so naturally, we can expect significant innovations coming from Google soon. This is fun for us at Devoteam as we get to work together with Google on some of the most compelling business cases and mature solutions within machine learning. We can offer more innovative solutions to customers like Telkomsel.
We have seen the potential of democratising ML through AutoML (now a part of Vertex AI). Combining data-centric AutoML Vision and Vision API, we can extract information from marketing materials, such as posters, banners, flyers. We serve the intelligence dashboard in Looker from BigQuery storing the information. This is a breakthrough for understanding competition in retail. Google is investing heavily in these technologies, so naturally, we can expect significant innovations coming from Google soon. This is fun for us at Devoteam as we get to work together with Google on some of the most compelling business cases and mature solutions within machine learning. We can offer more innovative solutions to customers like Telkomsel.
The Solution
While Devoteam G Cloud brings cloud and AI knowledge to Telkomsel, Devoteam and Telkomsel chose a joint-development approach to maximise the impacts and accelerate the project timeline.
App Engine presented Telkomsel with a serverless platform for the back-end of web and mobile applications.
The joint team addressed the company’s storage needs by implementing cloud storage which provided a data lake for raw photos. The implementation of Vision API OCR can further extract data from pictures by capturing the relevant text, and AutoML Vision assists analysts in image recognition and identification.
Cloud Functions directs the right person’s attention to events occurring in the cloud environment that may require action.
The implementation of BigQuery and Looker provide both a modern data warehouse for retail intelligence and intuitive visualisations via centralised dashboards.
The Results
App Engine has freed up developers’ time at Telkomsel, allowing them to get down to the crux of what matters. With no server management or configuration deployments, developers can scale applications up or down without the need for infrastructural management. App Engine provides serverless back-ends, real-time data processing and opportunities to create more intelligent applications.
Additionally, Cloud Functions offers automatic scaling and high availability. When watched events are triggered, an alert is created, and this encourages greater efficiency for developers responding to events in the cloud environment that may require direct action.
Cloud storage allows Telkomsel employees to work from anywhere and access the relevant data from authorised devices at any time. Since adopting Google Cloud Storage, Telkomsel’s digital data has benefited from Google’s robust security protocols, ensuring data loss prevention, high availability and disaster recovery.
Telkomsel uses Cloud Storage primarily as a data lake for raw photos.
Cloud Vision API and AutoML Vision enable engineers to derive insights from raw photos and data. The use of ML creates opportunities for better understanding by creating structure out of unstructured data. The OCR capabilities of Vision API automatically detect and class unlabeled images with human-like aptitude without human error. Engineers can gain further insight into raw data with OCR text detection in more than 50 languages. Both GCP products can process millions of documents in record time.
Telkomsel can now use AutoML Vision to train machine learning models to categorise raw photos through object detection and logo and poster brand recognition according to its predefined labels.
BigQuery and Looker work hand-in-hand to leverage the underlying power of Telkomsel’s data as the company grows.
Data is loaded directly from cloud storage into BigQuery. This data can be further analysed and presented intuitively to allow for more intelligent business decisions and improve customer experience.
Looker uses visualisations and dashboards to deliver easy access to the relevant data needed for users to understand Telkomsel’s revenue, margins, and sales on a deeper level.
By accessing the dashboard in Looker with analytics in BigQuery, we’re able to see patterns in our data that were previously invisible to us. Although we only recently implemented Google Cloud, we’ve already seen significant benefits: both in terms of our campaigns and being able to analyse our data in ways that were simply not possible before. In one solution, we have all the tools we need.
