# Reducing Lag on a Live Odds Feed

> TechSupport Team supported RapidLine’s Kafka and Flink pipeline, which supplied live odds to six sportsbooks. The client’s Grafana records show p95 feed lag falling from 9 seconds to 400 ms over five months.

- Publisher: TechSupport Team
- Canonical page: https://techsupport-team.com/case-studies/real-time-odds-feed-kafka-flink/
- Updated: 2026-11-16

Client: RapidLine · iGaming & Betting · UK real-time odds feed provider

## Client and engagement facts

- **Client:** RapidLine
- **Industry:** iGaming & Betting
- **Business context:** UK real-time odds feed provider
- **Delivery model:** Data platform support
- **Application:** Real-time odds and trading feed pipeline
- **Supplied geography:** UK

## Application stack and tools

### Application stack

- Kafka
- Flink
- ClickHouse
- Python
- AWS

### Tools used

- Grafana
- Datadog
- Kafka monitoring

Related services: [Application support](https://techsupport-team.com/services/application-support/)

## Why Support Was Needed

A checkpoint failure delayed odds during a major match. The original pipeline team had left, and the remaining documentation offered little guidance on checkpoint state or stalled consumers.

## Technical Work

- Mapped the Flink pipeline and checkpoint state in week one.
- Rebuilt the checkpoint strategy with dead-letter handling for messages the pipeline could not process.
- Added a Grafana panel showing feed lag by customer.

## Representative Incident

The first checkpoint redesign failed a simulated Kafka broker restart in staging. The engineers revised how checkpoint state was persisted and retested broker failover before deploying the second version.

## Results and how they were measured

Table: Reported results and their measurement scope

| Metric | Baseline | Result | Units | Measurement window | Source and scope |
| --- | --- | --- | --- | --- | --- |
| P95 feed lag | 9 s | 400 ms | Baseline: s; Result: ms | 5 months | Client Grafana |
| Dropped messages | 0.7% | 0.02% | % | 5 months | Client Kafka |
| Data-quality alerts | 19/day | 3/day | /day | 5 months | Client Datadog |

## Evidence and Limitations

The comparison covers five months. Grafana measured feed lag, Kafka supplied dropped-message counts and Datadog tracked data-quality alerts.

## Ongoing Feed Reliability

TechSupport Team continued to support pipeline reliability. RapidLine owned the customer-facing service levels.

## Scope a Similar Engagement

Describe your stream processor, how delays reach customers and the failure cases your runbooks cover. [Request a support plan](https://techsupport-team.com/contact/) or explore [application support](https://techsupport-team.com/services/application-support/).

## Related pages

- [Application support](https://techsupport-team.com/services/application-support/): The related service for this engagement.

## Contact

Request engineers from TechSupport Team or ask a question: https://techsupport-team.com/contact/
