Case Studies | SaaS B2B

De-risking a live fraud platform before it moves clouds

GCP to AWS Migration Assessment

About

A SaaS B2B business working with Cloud Combinator on AWS. The client is anonymised at their request.

Challenge

The challenge had three focus areas, each a reason the migration could not simply be attempted directly.

A real-time stream with no tolerance for data loss

The fraud platform processes live behavioural and payment signals, so any migration of the platform and ML Core workloads had to guarantee that the stream carried no data-loss tolerance during cutover. A validated wave sequence with rollback points was needed before, not during, the move.

A spend commitment that is hard to forecast

With the first paid customers only just onboarded, the client could not confidently forecast a long-term cloud spend, yet a migration of this size needs a costed business case and a funding path. The engagement had to produce a defensible total cost of ownership model rather than a guess.

A platform split across two clouds

Core ML and clickstream workloads ran on Google Cloud while the rest of the estate ran on AWS. Consolidating onto one cloud promised lower operational overhead and a growth-linked funding path, but only if every Google Cloud service had a clean, confirmed AWS equivalent.

Solution

Because the whole platform is already built as infrastructure as code across four repositories, Cloud Combinator could build the assessment from what actually runs rather than from memory. The work runs end to end as a sequence: agree the access model, inventory both workloads from the IaC, give every component a 7Rs disposition and a named AWS target, assemble the target architecture in eu-west-1 with an us-east-1 replica, build the cost model in the AWS Pricing Calculator against current run cost, score readiness against the AWS Migration Readiness Assessment and Well-Architected pillars, then sequence the migration into waves with rollback points and deliver a clear go or no-go.

4 weeks

Indicative end-to-end assessment timeline

2

Core workloads inventoried and mapped to AWS targets

25%

Migration ARR targeted back as AWS MAP credits (projected)

By the numbers:

  • 4 weeks - Indicative end-to-end assessment timeline
  • 2 - Core workloads inventoried and mapped to AWS targets
  • 25% - Migration ARR targeted back as AWS MAP credits (projected)
Changes

The assessment is scoped to hand the client a plan it can act on with confidence: both workloads fully inventoried, every component mapped to a confirmed AWS target with zero unmapped items at sign-off, a costed business case, a readiness score, and a safe wave sequence. The bar for done is that the plan is validated, costed, and safe to execute.

  • Complete coverageBoth the platform and ML Core workloads are inventoried from the four IaC repositories, with each component given a 7Rs disposition and mapped to a confirmed AWS service.
  • A costed business caseA total cost of ownership model is built line for line in the AWS Pricing Calculator, with an indicative annual AWS run cost of around 565,000 USD for the core workload as currently estimated (projected, and to be reconfirmed against the full target).
  • A safe migration planA wave sequence with rollback points is produced for the real-time path, sized so the live fraud stream carries no data-loss tolerance during cutover.
  • Readiness, measured not assertedThe target is scored against the AWS Migration Readiness Assessment framework and the Well-Architected pillars, with high risk items documented and their fixes named.
  • HandoverThe assessment report, target architecture, TCO model, and migration plan are delivered to the client as the foundation for the platform phase.

With the plan validated, the client moves into the platform and the migration itself as separate phases, reusing the assessment artefacts and the infrastructure as code so each later phase is cheaper per unit of scope. The outcome is a platform no longer split across two clouds, but consolidated onto AWS on a validated, funded, and auditable path.

AWS Stack

Amazon Kinesis Data Streams

For real-time event ingestion in place of Pub/Sub.

Amazon Managed Service

For Apache Flink for stateful stream processing and sessionisation.

AWS Lambda

For serverless session and LLM compute that idles down between runs.

Amazon Bedrock

For the Claude Sonnet 4.5 fraud explanation model via the Converse API.

Amazon SageMaker

For model serving, training, feature store, and monitoring.

Amazon DynamoDB

And Amazon S3 for low-latency session storage and the data lake.

Amazon Redshift

With Amazon QuickSight for the analytics warehouse and BI.

YOU MIGHT LIKE

Related success stories

View all case studies

Case Studies | Insights

Utilising Language Recognition, Speed, and Enhanced Security to Make Social Media a Force for Good

  • Here, we take a detailed look at how the Cloud Combinator team collaborated with another cutting-edge AI service provider that provides intelligent systems to “make social media more social” for brands and users alike.
  • Arwen AI is a UK-based startup specialising in AI solutions to manage and enhance brands’ social media interactions. Founded in 2020 by Matt McGrory, Dr. David Cole, and Joel Bailey, Arwen. AI focuses on using AI to automatically detect and remove spam, toxic comments, and other unwanted content from social media platforms.
  • The team at Arwen have three core products. ‘Moderate’ is focused on identifying and removing toxic content from social media channels. ‘Engage’ helps brands identify and engage with meaningful conversations on social media, and ‘Customize’ allows brands to apply bespoke algorithms to their channels - creating an even more effective moderation and engagement.
Read more
CONTACT US

Ready to turn AI into impact?

We'll help you spot the highest-value opportunities, reduce risk around your first AI initiative, and define a clear path to results from day one.

Why talk to us:

Outcome-driven recommendations

AWS-recognised delivery expertise

Risk-aware AI adoption

Clear next step, not a sales pitch

Start with a focused 20-minute conversation about your goals — no pressure, no commitment.

This website uses cookies to enhance user experience and to analyze performance and traffic on our website.

See our Privacy Policy for details.