ExchangeWire spoke with Matthew Beck, VP Partnerships & Strategy at Nano Interactive, about DSP- vs SSP-centric activation, QPS economics, and how agentic AI could reshape programmatic pipes...

With sell-side decisioning gaining momentum, what materially changes for buyers and publishers?

The most interesting shift is where, and how fast, decisions get made. As logic moves to the sell side, crucial work happens in the auction's initial ~10ms window, before bid requests are sent, and close to the impression and its richest signals. That proximity surfaces more context, reduces waste, and rewards publishers who structure their supply. It also enables a more even conversation between demand and supply about which signals matter, and when.

However, this doesn’t make DSPs obsolete. Buyers still want the breadth, workflow, governance, and measurement that DSPs provide. The balance shifts: more intelligence sits where the impression originates, while DSPs orchestrate across channels and markets. The task now is making those worlds speak cleanly to each other.

Why do QPS and infrastructure economics matter so much to data and decisioning quality?

Every query per second is a real compute decision. When partners send categorisation or evaluation requests, each call incurs a processing cost. In practice, that means most companies cap QPS and filter by format, region, ad type, ID/cookie or DealID to prevent spiralling monthly costs. The unintended consequence is obvious: if you cannot listen to every opportunity, you inevitably make some decisions with partial information.

Running models inside the SSP changes that. With containerised execution you’re less bound by your own QPS limits and can evaluate far more supply. Not for scale alone, but to uncover quality opportunities you’d otherwise miss. For instance, enriched requests would contain DealIDs and not be filtered out.

Practically speaking, what are the trade-offs between activating data via a DSP versus via an SSP?

Activating in a DSP exposes data to the widest set of SSPs. Traders stay within familiar platforms, and audience discovery is generally more mature. That’s helpful for adoption and incremental self-serve spend. On the other hand, many DSPs quietly down-weight bid requests when no user identifier is present. That makes sense for ID-centric tactics, but it can under-value a growing slice of inventory where consented IDs are scarce. 

A sell-side route can see all the signals an SSP has, whether an ID is present or not, which is attractive if you are trying to widen reach with intent and contextual signals rather than cookies. In short, DSP routes maximise supply breadth and familiar workflow, whilst SSP routes maximise raw signal visibility, especially on ID-light supply.

How should buyers think about ID-free supply and line-item design in 2026?

Avoid mixing ID-based and ID-free tactics in one line item. Any ID constraint will exclude ID-free supply, even when intent/contextual logic could work. Use parallel lines - one optimised for ID-present, one for ID-free - to preserve reach, comparability, and clean decisioning. Rather than being ideological, it is about providing each approach the environment in which it works best, then measuring outcomes with clarity. ID-free supply is not a consolation prize; handled properly, it is often less crowded and closer to content that signals live intent.

IAB’s "Containerisation Project" for OpenRTB has sparked interest. What, if anything, should we expect in the near term?

The promise is a standard way to execute decisioning inside the SSP, with richer request access and without punishing QPS costs. Standards bodies are now giving it proper air time, with a small number of SSPs (including some Nano partners) and other platforms working on Containerised oRTB standards in conjunction with IAB Tech Lab. That’s healthy, and it signals that sell-side decisioning will not be a niche tactic.

That said, we should be candid about timelines. Some partners are further along than others, and launches are staggered. Expect incremental progress as teams harden the containers, document interfaces, and expand who can participate. The goal should be making systems portable and auditable, not adding more one-off integrations.

Where is the appetite for innovation coming from on the buy side?

Large holding groups often carve out teams to trial new routes to ROI, and this moment fits that brief. If your narrative is that closer-to-publisher signals, executed faster, can deliver cleaner outcomes and fairer value exchange, that resonates with organisations tasked with testing alternatives to the default DSP-dominant pattern. Don’t rip out and replace just yet, but instead perform controlled experiments that respect existing guardrails, while proving where new decisioning belongs in a modern plan.

Critically, innovation has to meet planners where they work. If activating new signals forces buyers into brand-new UIs, or into fresh legal and operational overhead, enthusiasm will fade. The solutions that endure will be those that improve outcomes while slotting into the workflows traders already trust.

Data discovery still feels stronger in DSPs than SSPs. Does that slow sell-side adoption?

It has an influence. In many SSPs today, most data activation happens with buyers you have already briefed. In DSPs, by contrast, proper self-serve data discovery can generate spend from teams you have never met. That organic discovery matters if you care about scale without constant sales hand-holding. The gap is closing, but buyers will need both robust discovery, where they already plan, and reliable execution wherever the best signal sits.

For publishers, the opportunity is complementary: if better discovery in DSPs drives demand, and better execution in containers lifts eligibility and match rates, everyone benefits. The trick is ensuring consistent taxonomy and measurement so that buyers trust what they are selecting, regardless of where they select it.

Agentic AI and AdCP have generated headlines. Will they collapse the DSP/SSP split, or just change the interfaces?

The honest answer is that it’s too early to tell. AdCP proposes a common language for buying and selling agents to communicate directly - a provocative idea that triggers speculation about whether the traditional DSP/SSP roles remain essential. Some serious companies are leaning in, which makes the conversation worth having. Equally, there is a lot of punditry emerging. The sensible near-term stance is curiosity with healthy scepticism, paired with hands-on prototypes that prove specific and bounded tasks.

Agentic tools will arrive first inside companies, quietly automating workflows that humans already own. As those mature, standards like AdCP may help agents talk beyond company walls. Whether that replaces the pipes or simply makes them smarter is an empirical question. In the meantime, buyers and publishers should ask for clarity: what does the agent do, what data does it touch, and how is the outcome verified?

Finally, "AI with intent" is a phrase we’re starting to hear more often from Nano. What does that mean in practice?

It has a dual meaning. First, AI amplifies intent-centric approaches by parsing content (for instance, through an intent graph), learning from feedback, and proposing tactics based on what people are trying to do, not who they are. That’s useful when identifiers are patchy and privacy matters. 

Second, using AI "with intent" should be a standard, not a slogan. So be purpose-driven: how are you using AI to drive outcomes, why is it better than legacy approaches, and how are results evaluated? Hype comes and goes, but being able to show transparent lift and where AI meaningfully reduces waste will outlast the buzzwords.