NEWS, FANTASY SPORTS & STREAMING

Everyone arrives in the same four minutes. That is the whole business.

Q: What does video actually cost to run?

It scales with watching, not with subscribers — delivery, storage and transcoding. We put those figures on the table before the build, because for some products they decide whether the model works at all.

Q: Our live data feed is unreliable. Is that fixable?

It is handleable, which is not quite the same thing. Scoring written to be idempotent copes with a feed that repeats or arrives late, so a stutter upstream does not become a wrong leaderboard.

News, streaming and fantasy sports share one problem: traffic that does nothing for a week and then arrives all at once. What breaks is never the feature — it is the minute the match starts.

Tell us your setup See the work

Peak

THE NUMBER WE DESIGN FOR

Idempotent

SCORING, WHEN A FEED REPEATS

Batched

PUSH, NOT ALL AT ONCE

Costed

VIDEO, BEFORE IT IS BUILT

LIVE SCORES ◆ FANTASY TEAMS AND CONTESTS ◆ VIDEO DELIVERY ◆ SUBSCRIPTIONS ◆ PUSH AT SCALE ◆ EDITORIAL WORKFLOW ◆

01 — THE PARTICULAR HEADACHES

Four things that break at kick-off.

01

The spike is the product

A system that is comfortable all week falls over in the four minutes before a match. Average load is a meaningless number here; the peak is the only one that matters.

02

Live data arrives late or twice

Scores and events come from a feed that stutters, duplicates and occasionally lies. Handling that badly means a leaderboard everyone can see is wrong.

03

Video costs more than anyone budgeted

Delivery, storage and transcoding scale with watching, not with subscribers. It belongs in unit economics before it belongs in a feature list.

04

Push to everyone, at once

A notification to a large audience is a self-inflicted load test. Without batching it takes down the thing people are being notified about.

02 — WHAT WE BUILD FOR THIS TRADE

Three routes for a media or sports product.

FASTEST ROUTE

Publishing that holds up

News and editorial where the traffic is spiky and the content is the product. Cached properly, so a story going wide is good news rather than an incident.

Editorial workflow with scheduling and roles
Caching built for a story going wide
Push batched rather than fired in one go

LIVE DATA

Fantasy and live scoring

Contests, teams, points and leaderboards driven off a live feed — with the duplicate and late-arrival handling that decides whether the scores can be trusted.

Scoring that is idempotent when a feed repeats itself
Leaderboards that update without re-reading everything
Contest and wallet rules kept in one place

FULL BUILD

Streaming and subscriptions

Video delivery, entitlements and billing, built with the running costs on the table from the first conversation rather than discovered in month three.

Delivery and transcoding costed before it is built
Entitlements that survive a shared password
Subscription billing with the failure cases handled

03 — WHAT WE HAVE ALREADY SHIPPED HERE

What this work looks like.

PLATFORM

News and editorial

Publishing systems built for spiky traffic — scheduling, roles and a caching layer that treats a story going wide as the point rather than the problem.

APP

Fantasy and live scores

Contests, team selection, live points and leaderboards, with the unglamorous work that keeps a score correct when the data feed misbehaves.

STREAMING

Video and subscriptions

Delivery, entitlements and billing — designed around what it costs to run at the audience you expect, not at the audience you have today.

04 — WHAT OWNERS IN THIS TRADE ASK

Four questions, answered before the call.

How do you plan for a traffic spike?

By designing for the peak rather than the average, and by testing against it before the day rather than discovering it on the day. The number we need from you is the worst minute you expect, not the monthly total.

What does video actually cost to run?

Our live data feed is unreliable. Is that fixable?

05 — YOUR SETUP