Server-side ad Insertion for broadcasters: Take control at the source

In 2026, connected TV won the upfront for the first time. US CTV upfront commitments reached $17.73bn, edging past primetime linear’s $16.98bn, according to eMarketer. The audience and the money have moved. Yet most broadcasters still run ad insertion far downstream, deciding which ad plays, where and for whom in the CDN, or on the viewer’s device. This is the case for treating server-side ad insertion, and dynamic ad insertion more broadly, as a decision you make at the source, the head-end, not just a delivery method you outsource downstream.

The CTV money is moving faster than your ad-tech

The headline growth still points one way. eMarketer puts US CTV ad spend at roughly $37.95bn in 2026, up around 15% year on year, and the IAB’s 2026 ad spend forecast drawn from a survey of 205 buy-side executives, projects CTV growth of 13.8%, second only to social media. The demand for targeted ads on the big screen is there.

The gap is where it gets interesting for you. In 2025, CTV captured 20.2% of the time people spent with media but only 7.7% of ad spend, per eMarketer. That’s the widest attention-to-spend gap in advertising, and one that’s closing fast. Ad-supported streaming now drives the subscriber base too: ad-supported tiers accounted for 57% of gross additions across premium SVOD services in early 2025, according to Antenna. These headline figures are US-centric, but the shift isn’t: Omdia expects global FAST revenues to climb toward $11bn by 2030, with Europe a major driver. It’s why broadcasters like CANAL+ and ITV are already moving ad control to the source.

So the money is arriving. The question is whether your platform is built to capture it, filling every ad break with the right ad creative, or to hand a share of it downstream. If you’re weighing how AVOD, FAST and hybrid models fit together first, our guide to OTT content monetisation sets the scene. This piece goes a layer deeper: the ad insertion plumbing that decides how much of that ad revenue you actually keep.

Client-side vs downstream ad insertion: where the decision leaks

Follow an ad impression backwards and you’ll usually find the decision (which ad, which viewer, which region) being made in one of two places. Both leak.

Client-side ad insertion (CSAI) makes the call on the device. When an ad break arrives, the player fires an ad request to an ad server such as Google Ad Manager or FreeWheel, which returns the selected video ads for playback. It’s flexible and it measures well, but that separate ad request is exactly what ad blockers intercept, and on connected TVs the ad playback is fragile: buffering, failed fills, an inconsistent viewer experience. Every blocked call is ad inventory you can’t sell, and every failed fill is an ad impression you never served.

The alternative many broadcasters reach for is dynamic ad insertion at the distribution layer, handled by the CDN or a downstream partner. It survives ad blockers, but it’s rigid and costly. Ad placements live outside your control, targeting is coarse, and reacting to a live moment (a goal, an overrun, a rights restriction) means waiting for a system you don’t own to catch up.

Neither problem is really about the mechanics of ad insertion. Both are about where the decision is made. If you want the fuller picture of how CTV and OTT ad targeting works end to end, we’ve covered the mechanics of VOD and OTT advertising separately. The short version: the closer that decision sits to your video stream at source, the more of it you control.

Source-side vs CDN-side: where the ad decision belongs

You already hold the most valuable point in the chain: the broadcast signal at the head-end, before your video stream fans out to CDNs, apps and devices. That’s where the ad decision can sit too.

What SCTE-35, ESAM and POIS actually do

Three standards do the work here. SCTE-35 marks the ad breaks, the points in the video stream where an ad can go. ESAM (Event Signalling and Management) manages those ad placements dynamically and makes sure each one executes correctly. POIS (Placement Opportunity Information Service) holds the intelligence: where the ad breaks are, and structured data about each, from duration to timing to conditions. We’ve broken down how POIS and ESAM move ad control to the source in more depth.

The distinction that matters: a CDN-side ad stitcher inserts ad content into the manifest. A head-end system decides what should go there, for whom, in real time, then lets the downstream ad stitching carry it out. The ad stitcher is mechanical. Decisioning is where the revenue and the control live. Our own POIS head-end platform puts that decision at the signal, so you can insert, replace or remove ads in seconds without depending on distribution-level systems.

Signalling tells you where an ad can go; deciding what belongs there is the other half of source-side decisioning. Our BSG MetaDAI analyses EPG metadata and maps each programme to standard IAB categories (Sports, Travel, Food & Drink), so a campaign is matched to programme context from the start, then refined with regional or audience rules. Sports fans see one ad during a live match, families another during a kids’ show, each decided at the signal rather than guessed downstream.

Regional targeting and blackout from one master feed

Head-end control also solves two things downstream systems handle badly. Regional ad targeting: serve targeted ads for a local business to an urban audience and a national brand to everyone else, from a single master feed, rather than maintaining parallel video workflows. And blackout, which means enforcing rights restrictions on live content, becomes a signalling decision at source, through SCTE-224, not a scramble in the app. Swap an ad after a goal, drop a premium mid-roll or post-roll into an overrun, black out a fixture in one territory: all decided where you already control the stream.

SSAI, SGAI and the measurement problem you can’t ignore

Moving the decision to the source doesn’t settle how the ad reaches the screen, and here you should be honest about the trade-offs between the main dynamic ad insertion methods.

How ad stitching works in server-side ad insertion

Server-side ad insertion (SSAI) stitches ads directly into the video stream on the server, so content and ads reach the device as one continuous single stream, the most deployed form of dynamic ad insertion. An ad stitcher splices in the selected video ads at each ad break, whether pre-roll, mid-roll or post-roll. SSAI technology beats ad blockers and holds a consistent viewer experience across devices, with no separate ad request to fail, which is why it’s the default for CTV, FAST and most live streaming.

Its cost is measurement. Because the ad is baked into the single stream, the client-side signals advertisers rely on to verify ad delivery get weaker. That’s not a small concern. In IAB research, 43% of CTV buyers said they weren’t confident where their ads actually ran. Weak measurement isn’t a technicality — it depresses what buyers will pay per ad impression.

Server-guided ad insertion and client-side ad tracking

This is where server-guided ad insertion (SGAI) comes in, the newest dynamic ad insertion approach. The server decides which ad plays and when; the player fetches the ad creative and handles playback, with client-side ad tracking restoring the measurement SSAI erodes. You keep the ad-block resistance and the shared, cacheable video content, and you open the door to interactive ads and richer ad formats that pure server-side ad insertion can’t carry, all without disrupting the viewer experience. The signalling behind SGAI is now standardised, through HLS Interstitials and MPEG-DASH Events, and the IAB Tech Lab’s December 2025 CTV ad format guidelines{target=”_blank” rel=”noopener”} gave the formats it enables, like pause and squeezeback ads, their first cross-platform specifications. Adoption is accelerating on the back of that, though SSAI technology remains the more mature default heading into 2026.

Here’s the point that ties back to the head-end: source-side decisioning doesn’t compete with server-side ad insertion or SGAI. It sits above both. The decision about what ad belongs in an ad break can happen at your signal whether that ad is stitched server-side or guided to the player for client-side ad tracking. Get the decision layer right and you can change ad delivery strategy underneath it without rebuilding.

What this looks like in practice

The theory is tidy. Live broadcast is not. So here’s how source-side ad control plays out in a real deployment.

We built Poland’s first real-time, signal-based ad replacement system for CANAL+, a major broadcaster and pay-TV provider. It runs on our POIS application: it takes the incoming video stream with its original metadata, manages the SCTE-35 markers dynamically, and outputs modified markers ready for real-time ad insertion, so ads can be substituted live, without relying on pre-scheduled ad breaks. The platform went live in December 2024 and reached full operational integration by February 2025, built in Java with a DevOps toolchain around it. You can read the full CANAL+ ad replacement case study for the detail.

The pattern that works, in our experience: don’t rebuild the video workflow in one move. Start with a pilot on a few channels or a single live event, prove the signalling end to end (marker normalisation, correction, generation at the encoder), then integrate with your encoders, ad servers and monitoring, and scale from there. It’s the same approach we take maintaining SCTE-35 marker management through ESAM for ITV, where the daily reality is keeping live signals standards-compliant under continuous monitoring, not a one-off install.

What both have in common: the hard part is rarely the protocol. It’s the seams between the encoder team emitting cues, the platform team running the signalling, and the ad-ops team owning the ad server exchange. Source-side ad insertion works when those seams are engineered deliberately.

The revenue case, and what source-side control won’t fix

The commercial argument is direct. Targeting quality now drives what buyers pay: in IAB’s 2026 survey, 49% of buyers named targeting capabilities their top criterion for TV and video spend, up ten points and overtaking content quality. Better decisioning at source, whether regional, contextual or real-time, is what makes targeted ads possible at scale, and it maps straight to CPMs and the value of every ad impression.

We’d rather be straight about the limits, though, because overpromising helps no one.

Source-side control won’t create advertiser demand you don’t have; it helps you capture demand more fully, not conjure it. It won’t replace client-side ad tracking; if verification matters to your buyers, you’ll still want SGAI or an OMID-based approach alongside it. And it isn’t plug-and-play: it needs integration with your encoders and ad servers, and it earns its keep on live and linear-style inventory more than on a simple post-roll or pre-roll on VOD. If your monetisation is entirely post-roll and pre-roll on demand, the case is weaker, and you should know that before you invest.

What it does do, reliably, is put the decision that determines your ad revenue back where you already control the video stream. For most broadcasters chasing the CTV shift, that’s the difference between setting the terms and accepting someone else’s.

The bottom line

The CTV money has arrived, and it’s still arriving. What decides how much of it you keep isn’t whether you run ads – it’s where you decide them. Push ad insertion downstream and you accept coarser targeting, weaker control and a share handed to systems you don’t own. Keep the decision at the head-end, and real-time, regional, targeted ads sit where you already hold the video stream, ready to be delivered by server-side ad insertion, SGAI, or whatever comes next.

If you’re mapping out how source-side ad control fits your broadcast chain, from a first pilot to full integration, see how our POIS head-end solution works. We’ve built and run it in live broadcast environments, so it shows exactly where source-side decisioning earns its keep, and where it doesn’t.

FAQ