by Dana Lee on 15th Apr 2026 in Video Interviews

In this episode of A Coffee With… Vijay Kumar, founder and CEO of Mile, joins John Still over a flat white to discuss the unique challenges facing the sell-side.

Mile rebranded in 2022 after a decade building for publishers. The company positions itself as an independent sell-side technology provider. Its objective is to improve publisher yield using custom machine learning to optimise the ‘last mile’ – the final auction before an impression is served.

Vijay explains why traditional exchanges and DSPs often optimise for their own transaction rates rather than publisher revenue. The conversation delves into technical solutions like dynamic floor pricing and traffic shaping, which reduces midstream bloat by ensuring bid requests reach the most relevant advertisers.

As the industry navigates the deprecation of third-party cookies and the rise of alternative IDs, these machine-learning-driven insights become essential for publishers to maintain a competitive edge and maximise the yield of their first-party audience attention.

AI & Machine Learning Context

Buy-side entities have long applied machine learning in ad tech. The novelty lies in offering equivalent tools to publishers, with targeted use cases. Mile concentrates on revenue operations and internal enablement, leveraging machine learning to capitalise on publishers’ loyal audiences and first-party data.

“Today, it’s a lot of headwinds for publishers, but that’s always been the case and they’ve always been resilient and adapting … if you have a direct relationship with your audience, and if you’re able to get the attention of the audience across your properties, we’d be able to help you maximise the yield that you can generate from the attention that you generate.” — Vijay Kumar

By layering dynamic floor pricing, traffic shaping, and real-time decisioning onto existing stacks, Mile addresses a longstanding asymmetry where buy-side entities leverage advanced machine learning while publishers rely on static rules.

Traffic Shaping

Vijay traces the evolutionary phases of the advertising ecosystem: header bidding mainstreaming (2013–2017), cookie deprecation, and the proliferation of alternative IDs, leading to signal variability and bid request bloat. DSP capacity constraints and inconsistent signals also complicate efficient transactions.

By analysing bid rate, win rate, and participation to learn what each SSP values, traffic shaping helps to reduce noise and infrastructure strain while surfacing and packaging high-value inventory appropriately. The goal is ‘clean’ transactions – ensuring requests reach buyers intact and actionable, despite complex incentives and limited publisher visibility.

Future Outlook and Expansion

Vijay is excited about bringing sell-side machine learning parity to more surfaces. Beyond display and video, Mile aims to extend into programmatic audio, CTV, in-app, and additional formats. The vision is to productise sophisticated experimentation and machine learning so publishers need not build or staff these capabilities internally.

Vijay reports strong publisher feedback, with unprecedented insights and operational improvements, underscoring a clear market gap. Mile intends to remain a pure-play optimisation partner, without altering publishers’ broader operations.

AttentionOptimisationPublisherRevenueSell Side