DSP Advertising: What Agencies Won’t Tell You About Programmatic Buying

A demand-side platform (DSP) is software that lets advertisers buy digital ad inventory across multiple ad exchanges and supply sources through a single interface, using real-time bidding to serve ads to specific audiences at scale. It sits at the operational core of programmatic advertising, giving buyers control over targeting, pacing, frequency, and bid strategy without negotiating directly with individual publishers.

That’s the textbook answer. What it leaves out is the part where most advertisers are either paying too much for access they could get cheaper elsewhere, or handing the keys to a managed service they barely understand and calling it a strategy.

What Does a DSP Actually Do?

At the mechanical level, a DSP connects to ad exchanges and supply-side platforms (SSPs), evaluates available impressions against your targeting parameters, and places bids in real time, typically within 100 milliseconds of a page loading. The whole process happens before the page finishes rendering.

What makes a DSP more than just a bidding tool is the layer of control it gives you over who sees your ads, when, how often, on what types of inventory, and at what price. You can set audience segments based on demographics, intent signals, browsing behaviour, CRM data, or lookalike modelling. You can apply brand safety filters, block specific domains, cap frequency by user, and optimise toward conversion events rather than just impressions or clicks.

The major DSPs in the market include The Trade Desk, Google’s Display and Video 360 (DV360), Amazon DSP, Xandr (now part of Microsoft), and a handful of specialist platforms for specific verticals or formats like connected TV. Each has different inventory access, data partnerships, and pricing structures. The right choice depends on your media mix, your audience, and whether you have the internal capability to run the platform effectively.

I’ve worked with most of the major platforms across different client categories over the years. The honest observation is that the platform itself rarely determines campaign performance. The strategy, the audience logic, and the creative do. A sophisticated DSP running a poorly constructed campaign will outspend its way to mediocre results just as efficiently as a simpler one.

How DSPs Fit Into the Programmatic Ecosystem

DSPs don’t operate in isolation. They sit inside a broader programmatic supply chain that involves several moving parts, and understanding where each piece fits matters if you’re going to make good buying decisions.

Publishers make their inventory available through SSPs, which connect to ad exchanges. Ad exchanges aggregate supply from multiple SSPs and make it available to buyers. DSPs connect to those exchanges on the demand side, evaluating and bidding on impressions on behalf of advertisers. Data management platforms (DMPs) or customer data platforms (CDPs) can feed audience segments into the DSP to improve targeting precision.

The whole system runs on real-time bidding (RTB), though private marketplace deals (PMPs) and programmatic guaranteed deals sit outside the open auction and give buyers preferred access to premium inventory at negotiated terms. Most sophisticated programmatic strategies use a combination: open exchange buying for scale and efficiency, PMPs for quality and brand safety, and direct deals for high-value placements where you need guaranteed delivery.

DSP advertising fits naturally into the kind of go-to-market architecture I write about in the Go-To-Market and Growth Strategy hub, where media investment decisions need to connect directly to commercial outcomes rather than being treated as a separate operational function.

Self-Serve vs. Managed Service: The Commercial Reality

Most DSPs offer two access models. Self-serve means you operate the platform directly, setting up campaigns, managing bids, and pulling your own reports. Managed service means you hand that work to the DSP vendor or an agency that holds the seat, and they run it on your behalf, usually for a percentage of media spend or a flat fee on top of media costs.

The managed service model is sold as a way to access expertise and reduce operational burden. Sometimes it genuinely is. More often, it’s a margin layer that sits between your budget and your results, with limited transparency into what’s actually happening inside the platform.

When I was running an agency and we were scaling our programmatic capability, the pressure from certain DSP vendors to push clients toward managed service was noticeable. The economics for the vendor were obviously better. The economics for the client were not always better, particularly once you factored in the service fee, the lack of raw data access, and the fact that the optimisation decisions were being made by someone who wasn’t accountable for the client’s commercial results in the way we were.

If you have the internal capability or agency support to run self-serve, it’s almost always the better commercial choice. You get full data access, direct control over bidding logic, and no hidden margin layer. The barrier is operational: you need people who know how to use the platform, and you need enough volume to justify the investment in that capability.

The minimum spend thresholds for self-serve access vary by platform. The Trade Desk, for example, typically requires working through an agency or reseller partner. DV360 is accessible through Google’s ecosystem but has its own complexity. If you’re spending less than a few hundred thousand per year on programmatic, the managed service route may be the only practical option, but go in with clear eyes about what you’re paying for.

Audience Targeting: Where DSP Performance Is Won or Lost

The targeting capability of a DSP is the main reason to use one rather than buying directly from individual publishers. You can reach specific audiences across thousands of sites and apps simultaneously, based on who they are and what they’re doing, not just which site they happen to be reading.

The targeting options available through most DSPs include demographic segments, interest and intent categories, contextual targeting based on page content, geographic and device targeting, retargeting based on your own site or app data, lookalike modelling based on your customer lists, and third-party data segments from data providers like Experian, Oracle, or Lotame.

The quality of third-party data segments varies enormously, and this is where a lot of programmatic spend goes to waste. A segment labelled “in-market for financial services” might be built from signals that are weeks old, loosely defined, or modelled from proxy behaviours that don’t actually correlate with purchase intent. Buying that segment at scale produces impressions, not results.

First-party data is consistently the most reliable targeting input. If you can bring your own CRM data, site visitor data, or conversion lists into the DSP and use those as the foundation for targeting and lookalike modelling, you’ll typically see better performance than relying on purchased segments. The deprecation of third-party cookies has made this more important, not less.

I’ve seen campaigns where switching from broad third-party segments to tighter first-party lookalike audiences cut the volume of impressions by 60% and improved cost per acquisition by more than that. The instinct to buy reach is understandable, but reach to the wrong audience is just expensive noise.

Bidding Strategy and Budget Allocation

DSPs give you multiple bidding options, and the right choice depends on your campaign objective and how much confidence you have in your conversion tracking.

CPM (cost per thousand impressions) bidding is the standard for brand awareness campaigns where you’re optimising for reach and frequency rather than direct response. CPC (cost per click) bidding optimises for traffic. CPA (cost per acquisition) or target ROAS bidding uses the platform’s machine learning to optimise toward conversion events, adjusting bids dynamically based on predicted conversion probability.

The algorithmic bidding options have improved significantly over the past five years. Platforms like The Trade Desk and DV360 have sophisticated optimisation engines that can genuinely improve performance over time if they’re given clean conversion signals and enough volume to learn from. The problem is that they need data to work, which means early in a campaign, before the algorithm has accumulated enough conversion events, you’re often better off with more manual control and conservative CPM or CPC targets.

Budget allocation across campaigns, audiences, and inventory types is where strategic thinking matters most. A common mistake is spreading budget too thin across too many targeting segments simultaneously, which means none of them accumulate enough data to optimise effectively. Concentrating budget on your highest-confidence audiences first, then expanding as you validate performance, produces better results than trying to cover everything at once.

For teams thinking about how DSP investment fits into a broader commercial growth framework, the BCG perspective on commercial transformation and go-to-market strategy is worth reading, particularly the emphasis on concentrating resources on the highest-value opportunities rather than spreading investment uniformly.

Measurement, Attribution, and the Reporting Problem

DSP reporting is one of the areas where I’d urge the most scepticism. The platform has a commercial interest in showing you positive results. View-through attribution windows, last-click models, and inflated reach metrics can all make a campaign look more effective than it is.

View-through attribution is particularly problematic. If your DSP claims a conversion every time someone saw an ad and then converted within a 30-day window, you’re almost certainly overcounting. Most of those conversions would have happened anyway. The only way to isolate the actual contribution of your DSP activity is through incrementality testing: running holdout groups who don’t see the ads and comparing their conversion rates to those who do.

Judging Effie entries over the years, I’ve seen this pattern repeatedly. Campaigns with impressive in-platform metrics that couldn’t demonstrate any incremental business impact when you looked at the underlying data. The platform reported success. The business hadn’t moved. Those two things can coexist, and often do.

Independent measurement is the answer. Use your own analytics platform to track actual conversions, not just the DSP’s reported conversions. Set up clean UTM parameters so you can attribute traffic accurately. Run incrementality tests where your budget allows. And treat the DSP’s attribution numbers as one signal among several, not as the definitive measure of performance.

Vidyard’s analysis of why go-to-market execution feels harder than it used to touches on a related point: the proliferation of channels and tools has made it harder, not easier, to get a clear read on what’s actually driving commercial outcomes. DSP advertising is a good example of that dynamic.

Brand Safety and Inventory Quality

The open programmatic ecosystem has a quality problem that hasn’t gone away. Ad fraud, made-for-advertising (MFA) sites, and brand safety incidents are real risks when you’re buying at scale across thousands of domains through an automated system.

Most DSPs have built-in brand safety controls, and third-party verification providers like IAS (Integral Ad Science) and DoubleVerify integrate with the major platforms to provide additional filtering. These tools can block ads from appearing on sites with specific categories of content, flag suspicious traffic patterns, and verify that ads were actually viewable by a human.

The practical approach is to use inclusion lists (whitelists of approved domains) rather than relying solely on exclusion lists for high-value campaigns. Open exchange buying on an inclusion list is slower to scale than open buying across all available inventory, but the quality difference is significant. For brand awareness campaigns where you’re paying CPM regardless of performance, inventory quality directly affects the value of every impression you buy.

Private marketplace deals offer the best combination of programmatic efficiency and inventory quality. You’re buying from named publishers at agreed terms, with the targeting flexibility of programmatic. The CPMs are higher than open exchange, but the quality and viewability rates typically justify the premium for brand-sensitive categories.

DSP Advertising as Part of a Growth Architecture

The mistake I see most often with DSP advertising isn’t tactical. It’s structural. Teams treat the DSP as a standalone channel with its own KPIs, its own budget, and its own logic, disconnected from the broader commercial strategy it’s supposed to serve.

Programmatic display and video are primarily demand generation tools. They work best when they’re building awareness and intent upstream of channels that capture demand more efficiently, like paid search. If you’re running DSP campaigns to reach audiences who’ve never heard of your brand, and then not investing in the search and retargeting infrastructure to capture the demand you’re creating, you’re leaving a significant portion of the value on the table.

The growth loop concept is relevant here. DSP advertising can feed the top of a loop that drives trial, retention, and referral, but only if the downstream experience and conversion infrastructure are built to receive that traffic effectively. A programmatic campaign driving cold audiences to a weak landing page with poor conversion rate optimisation is an expensive way to generate bounce rates.

The early days of paid search taught me something that still applies to DSP buying. When I was at lastminute.com running paid search for a music festival campaign, the thing that made it work wasn’t the sophistication of the platform. It was the clarity of the offer, the relevance of the audience, and the fact that the landing page made it genuinely easy to convert. The same principles apply to programmatic. The platform is the pipe. What flows through it is the strategy.

If you’re thinking about where DSP advertising sits in a wider commercial growth framework, the Go-To-Market and Growth Strategy hub covers the strategic architecture that makes individual channel investments like this one actually deliver against business objectives rather than just media metrics.

For teams exploring growth mechanics more broadly, the SEMrush breakdown of growth hacking examples across different business models is a useful reference for understanding how paid media fits alongside organic and product-led growth approaches.

When DSP Advertising Makes Commercial Sense

DSP advertising isn’t the right tool for every situation. It makes most sense when you have a few specific conditions in place.

You need meaningful scale. If your total digital advertising budget is small, the operational overhead of running a DSP, whether self-serve or managed, may not be justified compared to simpler alternatives like social advertising platforms, which offer many of the same targeting capabilities with lower complexity and lower minimum investment.

You need to reach audiences that aren’t well served by social platforms. If your target audience isn’t concentrated on Facebook, Instagram, or LinkedIn, programmatic display and video can reach them across the broader web, connected TV, digital audio, and mobile apps. The reach of the open programmatic ecosystem is genuinely vast.

You have first-party data worth using. If you have a substantial customer database, meaningful site traffic, or strong conversion data, a DSP gives you the infrastructure to use that data for retargeting, lookalike modelling, and suppression at scale. That’s where programmatic genuinely outperforms simpler buying methods.

You have the measurement infrastructure to evaluate it honestly. If you can’t run incrementality tests or at minimum separate your DSP-attributed conversions from your independently measured conversions, you won’t be able to tell whether the channel is actually working. Investing in a DSP without the measurement capability to evaluate it is a way of spending money with confidence you haven’t earned.

BCG’s work on brand strategy and go-to-market alignment makes the point that media investment decisions need to be grounded in commercial logic, not channel familiarity. DSP advertising is worth investing in when the commercial case is clear, not because programmatic has become the default expectation for sophisticated advertisers.

About the Author

Keith Lacy is a marketing strategist and former agency CEO with 20+ years of experience across agency leadership, performance marketing, and commercial strategy. He writes The Marketing Juice to cut through the noise and share what works.

Frequently Asked Questions