Monzo reduces in-app support requests by 50% with BigQuery

December 2018

Monzo is a leading UK challenger bank with a customer base of over a million people, and growing at a rate of 3,000 a day. The bank refines and optimises its fast-developing product with accessible, zero-maintenance, high-powered BI analytics based on BigQuery and GCP.

About Monzo

Financial Services

Founded in 2015

Leading UK challenger banks


Implement a single point of reference for all data

Make data accessible across the company

Empower staff to use and understand analytics


High-powered BI analytics based on BigQuery


50% less in-app support requests over 10 months

Non-technical staff self-serve 85% of BI queries

Automated analytics with minimal maintenance.

Every day, 3,000 people open accounts with Monzo, one of the leading challenger banks in the UK. Founded in 2015, Monzo now helps more than a million users handle their finances through easy-to-use current accounts, the Monzo debit card, and its intuitive smartphone app. That success has put the company at the forefront of the UK’s fast-growing and highly competitive FinTech sector.

“We are the only challenger bank right now that’s building its whole stack in-house,” says Dimitri Masin, Head of Data at Monzo. “That means that we don't rely on third parties, unlike our competitors, and our microservices architecture means we can maximise availability even while upgrading our systems. Customers don’t care why something does or doesn’t work, they just need it to work. It’s our responsibility to make it happen.”

Monzo uses an open-source database management system for its microservices, but needed another solution for business analytics. Looking to create a sole reference point for all analytics, Monzo searched for a solution ready to receive messages, logs and events from all of its applications and microservices in a single place.

“As a tech team we value simplicity, and with BigQuery we’ve created the simplest and most scalable setup we could imagine. For four or five years, banking conferences have discussed the need to bring all data together in one place for analysis. With BigQuery, we’re actually doing it.”

– Dimitri Masin, Head of Data at Monzo

A single source of truth, available to all

Monzo runs over 600 microservices on Kubernetes, using Apache Cassandra as the transactional database. Without additional tooling, that setup could complicate the task of BI analysis by making it impossible to create snapshots of the production database. For analytics ready to inform decision-making throughout the business, Monzo looked to record all of its operational or user-related data in a single place, from transactional information to credit checks.

To do that, Monzo stores enriched events from its microservices, apps and website in BigQuery, as a single source of truth available in real-time. In keeping with the company’s focus on simplicity, the solution operates without intermediate storage solutions, contributing to a streamlined stack. At the same time, straightforward tooling means new arrivals familiar with SQL can already be productive in their second week at the company. And because BigQuery is a managed service that scales automatically, Dimitri and his team don’t have to worry about capacity, despite rapid user-base growth.

“In other companies of our size you need a data engineering team of at least two to four people constantly on-hand to maintain and run day-to-day analytics infrastructure,” says Dimitri. “BigQuery doesn't need a dedicated team to maintain it. In the two and a half years since we set up the solution, it’s been so robust and so scalable that it’s required no maintenance work whatsoever.”

Meeting the three criteria for effective analytics

As Monzo’s Head of Data, Dimitri picks out three key areas that he sees as vital for effective analytics: autonomy; granularity; and automation. “I want every analyst or data scientist that joins our team to work autonomously on data, without depending on anyone else,” says Dimitri. “That’s why we use Looker on top of BigQuery to visualise results and make analytics accessible to everyone in the company.”

Those results will then be most valuable if data can be analysed with a high degree of granularity. “To derive deep insights about the business, you need to have data available in as granular a form as possible, so that you don’t waste time creating a new ETL for each new aggregation,” continues Dimitri. And by automating as much as possible, Monzo teams avoid wasting time and energy on repetitive work. “Because Google Cloud Platform is easy to use, even for non-technical people, everyone on our team is comfortable using Google Compute Engine to automate jobs.”

The combination of accessibility and low-maintenance of BigQuery means teams run analytics when and how they want, helping the company stay agile and responsive to its customer base.

“We do all of our analysis on the fly because BigQuery can execute such gigantic joins of tables at speed. That’s an incredible advantage. We define and analyse segments as we think of them, instead of creating an ETL process and realising the next day that we want something else.”

– Dimitri Masin, Head of Data at Monzo

Focusing on the analysis that really counts

Empowering Monzo staff to run their own analytics is a key goal for Dimitri, who set the target that 85% of business data questions should be answered through self-serve analytics. Releasing the data team from work on small issues means Monzo can apply its expertise to the issues that matter most for the company’s development.

“One of the main cost drivers for our business is the chat tool inside our app, which connects people with customer service in real-time,” says Dimitri. “We set up dashboards driven by BigQuery in the customer support room that show trending issues on chat and other information around that process, and we identified the most frequently recurring problems by looking at customer behaviour immediately before they put in a support request. Because we were able to understand how our users behave and where things go wrong on a granular level, we reduced the number of support requests we received through our app by 50% over 10 months.”

In another example, the Monzo team addressed international ATM fees, which constituted the biggest cost driver for the company. Using their analytics solution, they discovered that 60% of the fees were attributable to only 5% of customers. Because the behaviour of a minority of customers was driving up costs for the other 95%, Monzo put a £200 cap on free withdrawals abroad, creating a more cost-effective, equitable system.

Streamlining for the future

Today, Monzo’s BigQuery table contains over 70TB of data and grows by 150GB a day, with no performance or maintenance issues reported. Already, Monzo has achieved its goal, with 85% of day-to-day business questions answered by staff directly, without consulting the Monzo data team.

A long-time G Suite user, Monzo now plans to expand its TensorFlow fraud prediction model using GCP tools. “The Cloud Machine Learning API is something we’re particularly excited about,” says Dimitri. “We are in the early testing stage, but the initial results look promising. Our original model took three weeks to train, but with Machine Learning APIs it’s almost as simple as choosing the number of machines you want to run the training on.” Using BigQuery, Cloud Dataflow and Cloud Datastore to extract and store features for its model in real-time, Monzo has already reduced its rate of fraud to an order of magnitude lower than the industry average.

“Analytics play a big part in streamlining our system for the future. The speed and usability of BigQuery means we can really understand our user segments, so we know which are profitable, which aren’t, and why. That’s vital to defining our company strategy, now and in the future.”

– Dimitri Masin, Head of Data at Monzo

AI: Machine-learning increases Tajawal’s campaign revenues by 44%