![\"\"](\"https:\/\/predictly.se\/wp-content\/uploads\/2024\/07\/binoculars-1024x491.jpg\")
<\/p>\n<\/p>\n
<\/p>\n
For retailers, maintaining the correct inventory levels is crucial: too few items can lead to stockouts, while too many can lead to obsolescence. Our client needed a sales forecasting solution to accurately predict demand and thereby generate purchase orders for new releases and replenishment of sold goods.<\/p>\n
The client operates nearly 100 brick-and-mortar stores across Sweden, in addition to managing two distinct e-commerce sites from a central warehouse. As part of their IT platform modernization, they have recently started using Google BigQuery as their new data platform and are adopting a Data Mesh strategy supported by a microservice and event-driven architecture.<\/p>\n
To meet the client’s needs, we created a BigQuery dataset dedicated to forecasting, containing sales data with daily, weekly, and monthly aggregates and product information. Initial data exploration and preparation were conducted using Jupyter Notebook, followed by demand prediction using BigQuery ML and its built-in S-ARIMA support.<\/p>\n
Integrating BigQuery APIs into the microservice architecture enabled the easy deployment of trained demand forecasting models in the replenishment process, combined with other data points to create purchase orders. This approach allowed for rapid development of sales forecasting capabilities, enhancing the return on invested capital in stores and warehouses.<\/p>\n
<\/p>\n
![\"\"](\"https:\/\/predictly.se\/wp-content\/uploads\/2024\/07\/books-1024x602.jpg\")
<\/p>\n
Several of our client’s stores, located near a few Swedish universities, had fostered strong relationships with the student body. Store staff manually scanned public university web pages to ascertain the courses and required literature for the upcoming semester. They ensured this literature was available as students commenced their new semester, aiding students with prepackaged bundles and consistently achieving higher sales margins.<\/p>\n
The manual process of scanning web pages for each course’s literature was time-consuming and often detracted from customer service. By integrating a large language model, specifically Google Gemini Pro, we automated the scanning of each course page, identifying relevant courses and literature swiftly. This automation significantly reduced the time required to maintain this business model, subsequently boosting profitability.<\/p>\n
<\/p>\n
![\"\"](\"https:\/\/predictly.se\/wp-content\/uploads\/2024\/07\/customer-satisfaction-1024x602.jpg\")
<\/p>\n
A Swedish insurance company known for excellent customer support required a method to enhance information sharing within the organization. They offer tailored pension and health insurance solutions and services to Swedish companies, but addressing customer queries and managing cases was time-consuming due to scattered information across different systems.<\/p>\n
By integrating advancements in Generative AI and Google Gemini’s large language models with Retrievable Augmented Generation (RAG), we consolidated various information sources into a searchable and interactive chat-based system. This innovation significantly reduced the time needed to respond to customer inquiries and issues, thereby maintaining a high-quality customer experience without compromise.<\/p>\n
<\/p>\n
![\"\"](\"https:\/\/predictly.se\/wp-content\/uploads\/2024\/07\/happy-girl-1024x602.jpg\")
<\/p>\n
A Swedish retailer analyzed their customers’ shopping behaviors and identified nine distinct categories that matched their patterns. This segmentation became a cornerstone of their weekly communication and offers strategy, executed by the marketing department to ensure that each customer received the most suitable offerings, thus enhancing the customer’s lifetime value.<\/p>\n
As part of their IT platform modernization, the routine for ensuring accurate customer segmentation was migrated to Google BigQuery, which supports clustering via KMeans through BigQuery ML. The existing model was transferred to BigQuery, and using its open APIs, updated segmentation data is continuously exported and integrated into the client’s marketing automation platform, Voyado.<\/p>\n
By leveraging Google BigQuery, we seamlessly transitioned the customer segmentation process to the new platform without disrupting customer communications.<\/p>\n
<\/p>\n
![\"\"](\"https:\/\/predictly.se\/wp-content\/uploads\/2024\/07\/postman-1024x602.jpg\")
<\/p>\n
A large Swedish retailer with over 2 million addressable customers needed to determine the profitability of their annual Addressed Direct Mail (ADM) campaign, especially with rising costs in physical mail transport looming. The customers were randomly assigned into different groups: one receiving the annual catalog and a control group that did not. As customer profiles were already integrated into the client’s data platform on Google BigQuery, grouping information was readily incorporated into the campaign’s business processes and address lists sent to the printing company.<\/p>\n
After the anticipated effect period of the annual campaign had elapsed, it was crucial to analyze whether the ADM campaign had made a significant impact. This analysis was straightforward, using the sales transaction data available on the client’s data platform.<\/p>\n","protected":false},"excerpt":{"rendered":"
A summary of the most interesting AI Use Cases we have implemented.<\/p>\n","protected":false},"author":1,"featured_media":13419,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"inline_featured_image":false,"cybocfi_hide_featured_image":"","footnotes":""},"categories":[141,116,110,106,114,113,137],"tags":[],"class_list":["post-13436","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-en","category-big-query-en","category-data-en","category-google-cloud-en","category-jupyter-en","category-machine-learning-en","category-retail-en"],"yoast_head":"\n