Streaming platform TCO audit: the costs your CDN bill misses

Cost reviews at streaming services often start with the CDN invoice, because it has the clearest monthly data. It’s rarely the whole picture.

A streaming platform TCO audit looks at the rest: SaaS licences, cloud infrastructure outside delivery, observability, app codebases, the people who keep it all running, and the contract terms that decide what you can change and when. Some of those costs are worth cutting. Others are worth keeping once you can see what they protect.

Operational savings can also last. ITV’s 2025 full-year results report £63m of permanent non-content savings in 2025 alongside £15m of temporary ones, £253m of permanent savings since 2019, and a further £20m planned for 2026.

What the total cost of ownership of a streaming platform includes

Total cost of ownership covers everything you pay to keep a streaming service running, including the engineering time spent keeping existing systems alive instead of building new features. In practice it splits into seven buckets.

o avoid counting the same money twice, classify each cost by function rather than by invoice. A hosted observability tool belongs under observability, not SaaS; the cloud bill for a QA environment belongs under cloud, and the engineers running the tests under people and run costs. One line, one bucket, one owner.

Delivery is usually the most closely watched area already, and if you route traffic across providers, our guide to multi-CDN setup covers it. If the audit points at the operating model itself, our comparison of managed services and in-house OTT operations covers that decision.

SaaS licences: the line item nobody owns

Zylo’s 2026 SaaS Management Index uses enterprise data across sectors rather than streaming, but it shows how SaaS spend behaves when nobody owns it centrally: 305 applications and $55.7m a year on average, only 54% of licences in use, and $19.8m a year lost to unused ones. Business units control 81% of that spend and IT 15%. The FinOps Foundation’s State of FinOps 2026 points the same way, with 90% of organisations managing SaaS spend or planning to within a year.

Streaming stacks often collect overlap in predictable places. A player SDK reports quality of experience data that a separate analytics tool reports again. A CMS ships a scheduling module while the scheduling team pays for its own tool. An acquisition brings a second subtitling contract that nobody retires.

Consolidation pays when the capabilities genuinely overlap, the data can be exported, the contract ends within your planning window and the team using the tool agrees on the replacement. Where one of those is missing, renegotiating seats or tiers is often the faster saving.

Cloud beyond delivery: staging, logs and archives

Flexera’s 2026 State of the Cloud Report found that organisations estimate 29% of their IaaS and PaaS spend is wasted, the first increase after five years of decline. It’s a cross-industry self-estimate, so treat it as an indication of scale rather than a streaming benchmark.

In streaming infrastructure, waste often sits outside delivery. Staging stacks sized to rehearse a major live event may keep running at that size. Archive renditions can sit in hot storage regardless of how often they’re watched, and a re-transcoded library may keep the old renditions too. Log retention raised during an incident tends to stay there, and Grafana Labs’ 2025 Observability Survey put observability at an average of 17% of compute infrastructure spend.

Spyrosoft’s own cloud work shows which fixes carry across sectors. For a financial services client, fewer test environments, a revised backup policy and managed monitoring in place of a self-hosted stack took the monthly AWS bill from $18,000 to $5,700, a 68% reduction. In a streaming stack, the equivalents are per-release staging for apps and player builds switched off outside release windows, retention rules for mezzanine files and archive renditions, and managed logging for player and CDN data. These habits tend to build up in stacks that moved to cloud streaming in stages.

Published streaming cases point the same way. CANAL+ cut its AWS spend by just over $1m in 12 months, around 30%, while usage grew, and credits the result to engineers trained in cloud cost management. TuneIn, an audio streaming service, saved more than 25% on its AWS bill by running Kubernetes workloads on spot capacity. Both are vendor-published, so read them as illustrations rather than an industry average.

People and run costs: the line that rarely appears on one report

For many streaming platforms, a large share of TCO is the work of keeping them running: OS and SDK updates, device regression testing, certification, L1 to L3 support, incident management during live events, and the interest paid on technical debt. It’s rarely costed as one line, and the FinOps Foundation found that only 28% of organisations are beginning to include labour costs in FinOps.

The useful measure is the split between run and change. Clutch’s 2025 survey of 600 US software development professionals found most teams settle at roughly 60% new features and 40% maintenance. That’s a general software benchmark, and streaming teams carry extra run work most software teams don’t, so the trend in your own numbers matters more. When run work keeps growing as a share, the platform is getting more expensive to own even if no invoice has changed.

This is the work our OTT managed services teams do day to day, and where the operating model matters most: some costs fall when one team runs the same processes across several platforms, and some rise because of handover and governance.

App maintenance: count codebases, not platforms

The device map keeps widening. Omdia’s April 2026 forecast expects TV operating systems that didn’t exist in Europe in 2022, led by VIDAA, Titan OS and TiVo, to reach 30% of the European market by 2030, up from 21% in 2025. Platform owners also change the rules: Roku’s April 2026 certification update made Instant Resume mandatory from 1 October 2026 for US apps above five million streaming hours a month. Our breakdown of Roku certification changes shows how such rules can block a release.

Cost follows codebases and certification cycles more closely than the number of platforms. Depending on architecture and features, an eight-platform estate can often run on four codebases: shared web for browsers, Tizen and webOS, native iOS and tvOS, native Android and Android TV, and Roku. Titan OS and VIDAA also support web apps. Shared code doesn’t remove device work, though: each platform still needs its own testing, performance tuning, DRM checks and certification.

ZDF shows the pattern at scale. Its apps run on more than 25 connected TV platforms. Before consolidation, platforms each needed dedicated maintenance, features and fixes were built several times, and QA multiplied with every target. ZDF moved to a shared core with small per-platform adapters and a single CI/CD pipeline, and reports faster releases and lower development costs, with each new platform becoming an adapter task. No before-and-after figures were published.

The audit questions follow: how many codebases you fund, how many certification cycles you run a year, and what share of viewing each platform delivers.

Metrics that show whether the spend is working

Knowing where the money goes is half the audit. Five ratios show whether it buys enough.

Published benchmarks for these ratios are rare, because each operator draws the boundaries differently, so the trend is what counts. Getting the data is often the harder part: TuneIn only gained a cost per stream once it could see spend per Kubernetes service, which its cloud bill alone didn’t show.

How to run an OTT platform cost audit

1. Inventory every contract, environment and codebase

List every SaaS and cloud contract with its owner, renewal date, notice period, commitment and actual usage, plus every environment, app codebase and supplier. A contract nobody can name an owner for is the first one to question, and an environment nobody can explain is a compliance risk as well as a cost.

2. Allocate cost to services, not departments

Map operational costs to what viewers and advertisers use: live, VOD, each app, ad sales, editorial tooling. Report unallocated cloud spend as a number on its own. The FinOps Foundation found that 78% of FinOps practices now report into the CTO or CIO, which is where cost control becomes part of normal management.

3. Decide what to keep, optimise, consolidate, renegotiate or switch off

Not every line needs cutting. Give each one of five outcomes.

Then score every change on the annual saving, the one-off cost of making it, and the operational risk to viewers, advertisers or certification. A modest saving with no risk often beats a large one that needs a migration during the live season.

4. Build a contract calendar

Plot renewal dates and notice deadlines against your budget year: a tool with a December notice deadline goes first, even if a bigger saving is available in June. Switch-off savings usually land within a billing cycle; consolidation savings arrive once the old system is retired.

Use the EU Data Act at your next renewal

Under Article 29 of Regulation (EU) 2023/2854, from 12 January 2027 providers of data processing services can no longer impose switching charges, including data egress. The rules cover infrastructure, platform and software services.

The exceptions matter. Article 31 exempts services mostly custom-built for one customer and not offered at broad commercial scale, plus time-limited test versions, and providers must say before signing which obligations don’t apply. The ban also doesn’t cover your own migration work or parallel running, and minimum terms still apply. Treat it as a stronger position at renewal rather than a free exit, check each contract with legal counsel, and see what breaks in an OTT platform migration for the practical side.

What a TCO audit won’t fix

An audit won’t touch rights or content costs, and it won’t fix the business model: if a service loses money on every subscriber, lower infrastructure costs only slow the loss. Consolidation also costs money first; when ITV raised its 2025 savings target, it guided a one-off £40m to achieve it. And cutting monitoring without knowing what it protects can move cost into outages.

Where to start

If you can only look at three areas first, start with the contract calendar for the next 12 months, because notice deadlines decide which savings are still available. Then review non-production environments and observability, where changes are quick to make and easy to measure. Finally, cost each app platform against its share of viewing, because that’s where long-term run costs build up and where you can reduce TCO for years, not months.

Spyrosoft BSG’s 200+ video specialists have spent 13 years building, maintaining and running streaming platforms and multiscreen apps for broadcasters. If you’d like an outside view on your cloud infrastructure, your app estate or your maintenance model, talk to our experts.

FAQ

What is included in streaming platform TCO?

expand_more

Video delivery, SaaS licences, cloud infrastructure outside delivery, observability, app codebases, people and run costs such as maintenance, QA, support and incident management, and the contract terms behind each tool. To avoid double counting, assign each cost to one bucket by its function rather than by the invoice it arrives on, and give every bucket a named owner.

Which metrics show whether OTT platform spend is efficient?

expand_more

Five ratios cover most of it: cost per streaming hour, cost per monthly active user, cost per platform compared with that platform’s share of viewing, the share of engineering capacity spent on run work, and operating costs as a share of total cost of ownership. Track the trend over several quarters, since published benchmarks rarely use comparable boundaries.

Should every cost line in an audit be reduced?

expand_more

No. Some costs protect revenue, reliability or compliance and should be kept, because they reduce hidden costs and operational burden elsewhere, or even increased, such as monitoring for live events. Sort each line into keep, optimise, consolidate, renegotiate or switch off, then weigh the annual saving against the one-off cost of the change and the operational risk to viewers, advertisers and certification. This is especially true when automation or a managed platform can replace recurring manual work.

Does the EU Data Act make switching cloud or SaaS providers free?

expand_more

Not entirely. From 12 January 2027 providers can’t impose switching charges, including data egress, but custom-built services not offered at broad commercial scale and time-limited test versions are exempt. Your own migration work, parallel running and contract minimum terms still cost money. Its practical value is as a bargaining point at renewal, checked contract by contract with legal counsel.

How can you reduce OTT app maintenance costs across TV platforms?

expand_more

Start by counting codebases and certification cycles. A unified platform or shared core can lower vendor management and operational complexity across app estates. Then compare each platform’s maintenance, QA and support effort with its share of viewing. Where the architecture allows, a shared web codebase can serve browsers and several smart TV platforms, but each device still needs its own testing, performance tuning and certification. Consolidate low-viewing platforms first. Keeping every platform in-house gives full control, but it may not be the lowest-TCO option for organisations without substantial engineering resources.