Company-wide operational analytics
公司规模
1,000+
产品
- Looker
- Amazon Athena
- Firebolt
- Databricks
- BigQuery
技术栈
- SQL
- Spark
- Cloud Data Warehouse
实施规模
- Enterprise-wide Deployment
影响指标
- Cost Savings
- Customer Satisfaction
- Productivity Improvements
技术
- 分析与建模 - 大数据分析
- 分析与建模 - 预测分析
- 平台即服务 (PaaS) - 数据管理平台
适用行业
- Professional Service
- Software
适用功能
- 商业运营
- 质量保证
- 销售与市场营销
服务
- 软件设计与工程服务
- 系统集成
关于客户
Appsflyer is a leading mobile marketing analytics and attribution platform, the source of truth for many leading brands when it comes to attribution. Over 12,000 companies rely on Appsflyer to help them track, analyze, and understand buyer journeys, from tracking mobile installs to ad interactions across mobile and other channels. Appsflyer works with over 6,000 partners to make tracking buyer journeys a reality. The company has grown significantly over the years, expanding its internal analytics capabilities to support a wide range of functions including sales, support, marketing, finance, and HR. With a user base of 1,000 Looker users and 60 Looker experts, Appsflyer has been at the forefront of leveraging data to drive business decisions and improve customer experiences.
挑战
Over the last few years, Appsflyer had pushed its internal analytics to the limit. Three years ago, Appsflyer started using Looker. Over time, adoption grew to 1,000 Looker users across sales, support, marketing, finance, and HR, including 60 Looker experts who supported other users by building and maintaining the various dashboards. Appsflyer was running their analytics on Amazon Athena, but as usage had grown, so had the challenges. First, the data had become too big for Athena to handle. Appsflyer has 35 petabytes of terabytes of raw data about customers and their activities, with multiple tables consisting of billions to hundreds of billions of rows. Athena could not handle more than 5 billion rows of data in any query. So they had to aggregate the raw data using Databricks to a smaller data set that Athena could handle. Athena also could not handle more than 20 concurrent queries, which had become a problem by the time Appsflyer reached 1,000 Looker users. Second, the ETL process was slow, expensive, and inflexible to change. A spark job often took weeks to change, and days to run if they had to rebuild all the dimensions and history. This was not only too long, since users often required new reports in days, but also very expensive. Updates also became a problem with GDPR regulation. Each time an entity opted out, Appsflyer had to rerun entire Spark jobs to remove them, which not only took a long time to run but was very expensive as well. Third, even with the dramatically reduced data sets, Athena was still too slow. Some queries took minutes to run. Most interactive analytics, such as the executive dashboards, required query times of a few seconds at most. When the company needed a new report that required the detailed results from a join of two multibillion row tables, Appsflyer decided it was time to replace Athena.
解决方案
Appsflyer started to look at a cloud data warehouse that could store all of their data and aggregations, deliver the performance they needed, and provide greater self-service so that various teams could deliver new analytics faster. Appsflyer evaluated several vendors in a proof of concept (POC) with 3 stages: joining a big and small table, joining two big, multi-billion row tables, and a dashboard use case. They were already using BigQuery, and evaluated Snowflake, Vertica, Dremio, and SQream. They had evaluated Redshift a while ago and had ruled it out. They also used Druid with a customer Web interface for their customer-facing analytics and decided against using it because they needed a more general-purpose data warehouse. In the end, they chose Firebolt based on its much faster performance, and much better price-performance. With Firebolt, all of Appsflyer’s 1,000 Looker users are now able to run any analytics, at any scale, in seconds, and deliver new reports within days. As part of evaluating Firebolt, Appsflyer took the same schema and queries from Athena syntax and ran them on Firebolt, performing joins with 3 billion and 26 billion row tables. Before Firebolt, any new data or schema needed to be processed using Spark. It could take weeks to develop and run the Spark jobs. With Firebolt, data engineers can now do most of the ELT in hours using 100% SQL, and run it in Firebolt at any scale. The first project took 2 weeks or so of development. Most work now takes a few days or less.
运营影响
数量效益
Case Study missing?
Start adding your own!
Register with your work email and create a new case study profile for your business.
相关案例.
Case Study
Factor-y S.r.l. – Establishes a cost-effective, security-rich development environment with SoftLayer technology
Factor-y S.r.l., a web portal developer, was faced with the challenge of migrating its development infrastructure to a reliable cloud services provider with highly responsive technical support. The company needed a solution that would not only provide a secure and reliable environment but also support its expansion by providing resources to create and deliver innovative offerings.
Case Study
UBM plc: Taking the pulse of the business and engaging employees with a far-reaching strategic transformation
UBM, a leading global events business, was undergoing a significant strategic transformation named 'Events First'. As part of this transformation, the company was preparing to complete the largest acquisition in its history - Advanstar, a US-based events and marketing services business valued at more than USD970m. The company faced the risk of human capital flight if it was unable to effectively engage top talent with the new strategic direction. UBM needed to make significant structural, process and systems changes, uniting its previously autonomous regional businesses. The challenge was to ensure all of its employees were engaged and aligned with the new future vision.
Case Study
Darwin Ecosystem: Accelerating discovery and insight through cutting-edge big data and cognitive technologies
Darwin Ecosystem was founded with a unique vision of harnessing chaos theory mathematics to uncover previously hidden connections in unstructured data. The company’s algorithms can look at all the data generated by any source (such as news, RSS feeds and Twitter), and analyze how a specific set of concepts within that data are evolving over time. This is particularly valuable in situations such as business and competitive intelligence, social research, brand monitoring, legal discovery, risk mitigation and even law enforcement. A common problem in these areas is that a regular web search will only turn up the all-time most popular answers to a given question – but what the expert researcher is actually interested in is the moment-tomoment evolution of the data available on that topic. Darwin’s algorithm is computationally intensive, and the sources of data it correlates can be vast. To bring its benefits to a larger commercial audience, Darwin needed to find a way to make it scale.
Case Study
Wittmann EDV-Systeme launches IT monitoring services
Small and medium-sized businesses often lack the know-how and resources required for thorough IT system monitoring. Wittmann EDV-Systeme wanted to launch a solution to plug the gap – enabling it to improve its own competitiveness and that of its customers. IT landscapes are becoming ever more complex and outsourcing is gaining popularity, IT systems must nonetheless remain easy-to-use and extremely reliable at all times. Automated, round-the-clock system monitoring therefore represents an immensely valuable proposition for companies: downtime for business-critical applications can be avoided, and IT systems remain available at all times.
Case Study
Zend accelerates, simplifies PHP development
Zend Technologies, a major contributor to the PHP open source community, needed to keep pace with emerging trends such as mobility, agile development, application lifecycle management and continuous delivery. The company needed to provide the right tools to the worldwide community of PHP developers. The challenge was to support enterprise-class capabilities from end to end, including mobile, compliance and security. The pace of business required developers to show results fast across a variety of devices without compromising quality or security.
Case Study
Delivering modern data protection with cloud scale backup from Cobalt Iron and IBM
Organizations are struggling to modernize their legacy data protection environments in the face of growing demands around new infrastructure, new applications, and budget consolidation. Virtualization and modern application development processes have significantly outgrown legacy backup architectures. In response, infrastructure teams have created multiple backup solution types to handle the varying SLAs (performance, scale, cost) required by their business sponsors. However, the sheer number and variety of solutions in this uncontrolled expansion creates huge amounts of work, threatening to overwhelm the IT team in many organizations. Today, developers may add new applications and virtual server instances by the hundreds per day without accounting for the restrictions of the existing backup infrastructure. They leverage the cloud for immediate compute and storage resources, yet rarely communicate succinctly with corporate IT to ensure that the appropriate data protection services are in place.