A streaming platform's engineering team optimises for latency and availability while the business measures watch time and creator retention. How should these connect?
Show the full answer Hide the answer
Why they must connect
Technical metrics are only valuable as proxies for business outcomes. Optimising a proxy without knowing its relationship to the outcome means investing where it may not matter — and, worse, being unable to argue for the investments where it does.
How to connect them
1. Establish the relationship empirically. Measure how a technical metric affects a business one: how much does startup latency change the probability a viewer stays past thirty seconds; how much does a stream interruption reduce a creator's next-session likelihood.
Once quantified, "reduce startup latency by 200 ms" becomes an argument with a number attached rather than an engineering preference.
2. Define SLOs from the business impact. Availability targets derived from what unavailability costs rather than from convention. This also settles cost-versus-reliability arguments with arithmetic instead of opinion.
3. Put business metrics in the operational monitoring system, on the same dashboards, with the same alerting. Concurrent viewers, streams started, chat messages, creator payouts. These catch failures that technical monitoring structurally cannot — where the system works correctly and produces the wrong result.
4. Segment both by what concentrates value. For a streaming platform, a small number of large streams and creators account for a disproportionate share of watch time. A metric averaged across all streams hides a failure affecting the ones that matter.
The finding this usually produces
The technical metric that matters most is rarely the one being optimised. For short-session platforms it is frequently startup latency rather than sustained quality, because the user's decision happens in the first seconds and recurs on every stream — so a small improvement compounds across a session in a way a bitrate improvement never does.
Without the empirical link, an engineering team will reasonably optimise sustained quality, because that is what a video metrics dashboard emphasises.
The organisational benefit
Shared metrics end the argument about whether reliability work is worth doing. When an availability improvement can be stated in terms of retained watch time, it competes on the same terms as a feature — which is the only way reliability work gets funded consistently rather than after each incident.