London Marathon 2025 – TrainAsONE vs Strava

The London Marathon is an iconic event, drawing runners from across the globe. It’s also a fantastic opportunity to look at performance data.

As the pioneers in AI-powered run training, we at TrainAsONE (TAO) are always interested in how data can inform performance. How do runners using our personal, adaptive training system perform compared to the broader running community?

To explore this, we analysed data from TrainAsONE users who trained with us and then recorded their 2025 London Marathon race within our system. We compared their median finishing times against publicly reported median times for Strava users participating in the same event. We looked at results across key demographic groups: Male, Female, and generational cohorts (Boomer: born 1946-64, Gen X: 1965-80, Millennial: 1981-96, Gen Z: 1997-2012). We used the median time (the middle value) as it’s less skewed by outliers (such as our sub 2h30m runners) than the mean average.

Median Finishing Times: TrainAsONE vs Strava

Data Snapshot: Median Finishing Times

Below is the breakdown of median finishing times across demographic groups and generational cohorts.

Population Group	TAO Median	Strava Median	Difference	Top Performer
Female	04:32:12	04:41:57	↑ 9m 45s
Male	03:47:12	04:02:19	↑ 15m 7s
Boomer	04:41:44	04:44:59	↑ 3m 15s
Gen X	04:12:08	04:23:41	↑ 11m 33s
Millennial	03:48:16	04:11:05	↑ 22m 49s
Gen Z	03:59:36	04:24:31	↑ 24m 55s

Key Findings & Discussion

Performance Snapshot

The analysis reveals a clear and consistent trend:

• The median finishing time for the TrainAsONE users was faster than the reported Strava median across all analysed demographic groups (Male, Female, Boomer, Gen X, Millennial, Gen Z).
• While both datasets show expected trends (e.g., Males and Millennials achieving faster median times relative to other groups), the magnitude of the difference between the TAO and Strava medians is notable, particularly for younger generations.

“

“The time difference between TAO and Strava medians grew larger for younger generations.”

”

Interpreting the Differences: What Could Be at Play?

While this data provides a fascinating snapshot, it’s important to consider the factors that might contribute to these observed differences:

• The TrainAsONE Difference: AI-Driven Training Users: Runners who choose TrainAsONE are specifically opting into an AI system designed to optimise training based on their individual data. This suggests a user base potentially more focused on structured training and performance improvement compared to the very broad spectrum of users on a social platform like Strava (which includes many casual runners, cyclists, etc.).
• Potential Impact of AI Optimisation: TrainAsONE’s core purpose is to help runners train smarter and achieve their goals through adaptive, personal plans. While this analysis doesn’t prove causation, the consistently faster median times observed in the TAO users could reflect, in part, the positive outcomes associated with following such tailored training guidance.
• Generational Trends in Differences: An interesting pattern emerged: the time difference between the TAO sample and Strava medians grew larger for younger generations (smallest for Boomers, largest for Gen Z). This raises interesting questions – do younger TAO users potentially engage more deeply with AI training, respond differently, or simply represent a more distinct performance-focused subset compared to their Strava peers?
• Understanding the Strava Comparison: It’s crucial to remember that Strava’s median represents a massive, diverse user base. The comparison highlights how the median performance within a specific, potentially more performance-oriented group (like the TAO user sample) can differ from the median of a much larger, more varied population.

Important Context & Data Notes

• Sample Sizes: The number of users in the Strava sample is unknown, but no doubt considerably larger than the TrainAsONE dataset.
• Single Event: Findings are based solely on the 2025 London Marathon.
• Correlation, Not Causation: This analysis shows an association between the TAO user population and faster median times compared to reported Strava medians. It cannot definitively conclude that using TrainAsONE caused these faster times – other factors (like user self-selection) are involved.
• Strava Data Source: Strava median figures are based on publicly reported data for the event and represent the median for their user base participating. Direct access to the underlying Strava data or their filtering methodology was not available for this analysis.
• Data Curation: Minor differences in how activities are filtered or tagged as belonging to the official event could exist between platforms.

Conclusion: A Promising Snapshot

This analysis of the 2025 London Marathon provides a compelling glimpse into the performance of a sample of TrainAsONE users. The consistently faster median times observed across all demographic groups compared to the broader Strava medians are encouraging.

While acknowledging the role of user dedication and the limitations of sample data, these findings align with the potential benefits of personal, AI-driven training. They highlight how leveraging data intelligently, as TrainAsONE aims to do, might be reflected in strong race-day performances. It underscores the fascinating relationship between targeted training, user commitment, and results on the big stage.

We look forward to continuing to analyse performance data and help runners achieve their best!