Programmatic Advertising: What Most Marketers Get Wrong

Programmatic digital advertising is the automated buying and selling of digital ad inventory, using real-time data and algorithmic decision-making to serve ads to specific audiences across websites, apps, and connected platforms. At its core, it replaces manual insertion orders with auction-based technology that matches advertisers to available impressions in milliseconds. Most marketers understand the mechanics well enough. Fewer understand where it actually creates value, and where it quietly destroys it.

That gap between understanding the system and using it well is where most programmatic budgets go wrong.

What Programmatic Actually Does (and What It Does Not)

Programmatic advertising automates the process of buying digital media through three core components: a demand-side platform (DSP), which advertisers use to set targeting and bidding parameters; a supply-side platform (SSP), which publishers use to make inventory available; and an ad exchange, which connects the two in real-time auctions. When a user loads a webpage, an auction fires, bids are evaluated, and a winning ad is served, all within roughly 100 milliseconds.

What it does well: it scales audience targeting across a fragmented media landscape without requiring individual publisher relationships. You can reach a defined audience segment across thousands of sites with a single campaign setup. That is genuinely useful, particularly for advertisers who need reach without the overhead of direct buys.

What it does not do: it does not guarantee that the impression was seen, that the environment was brand-appropriate, or that the person who saw the ad was actually in-market. The automation handles the transaction. The quality of the outcome depends entirely on the inputs you bring to it.

I spent several years managing large programmatic budgets across retail, financial services, and travel clients. The single most consistent finding was that the advertisers who got the most from programmatic were not the ones with the most sophisticated DSP setups. They were the ones who were most disciplined about what they were actually trying to achieve before they touched a platform.

The Efficiency Trap

Programmatic’s core commercial pitch is efficiency. Lower CPMs than direct buys, automated optimisation, real-time bidding that theoretically finds the best price for each impression. And on a pure cost-per-impression basis, that pitch is largely true.

The trap is treating efficiency as the goal rather than the mechanism. When media planning becomes a race to the lowest CPM, the result is predictable: high volumes of cheap inventory in low-quality environments, served to audiences who were loosely matched by third-party data of uncertain provenance, with viewability rates that would embarrass anyone who looked at them honestly.

I have sat in enough media reviews to know that CPM is one of the most-reported and least-meaningful metrics in a programmatic dashboard. What matters is whether the impression had a reasonable chance of influencing behaviour. A £0.50 CPM against an audience that bears no relationship to your actual customer is not efficiency. It is just cheap waste at scale.

The more useful framing is cost per quality impression: viewable, brand-safe, served to a verified audience segment with genuine relevance to the product. That number is always higher than the headline CPM. It is also the only one worth optimising against.

Programmatic fits into a broader go-to-market architecture, not as a standalone performance play. If you are thinking about how paid media connects to your overall growth strategy, the Go-To-Market and Growth Strategy hub covers the commercial logic behind how these channels should be sequenced and prioritised.

Brand Safety Is Not Optional

Brand safety became a mainstream conversation in digital advertising around 2017, when several large advertisers pulled spend from YouTube after their ads appeared alongside extremist content. The underlying problem was not new. It had been a structural feature of programmatic for years. The scale of the platforms just made it impossible to ignore.

The programmatic ecosystem, by design, connects advertisers to inventory at speed and scale. That means your ad can appear on a credible news site, a mid-tier lifestyle blog, a made-for-advertising (MFA) site designed to generate impressions rather than serve readers, or something considerably worse, depending entirely on how tightly you have configured your brand safety controls.

Default settings on most DSPs are not conservative. They are set to maximise reach, because that is what the platform is optimised for. Advertisers who do not actively configure inclusion lists, exclusion lists, content category blocks, and viewability thresholds are effectively running open-web campaigns with minimal guardrails.

The practical steps are not complicated: use a verified third-party brand safety solution alongside your DSP’s native controls, build a publisher inclusion list based on your own quality criteria rather than relying on pre-built contextual segments, set a minimum viewability threshold and stick to it, and audit placement reports regularly rather than trusting automated optimisation to handle it. None of this is difficult. It just requires treating brand safety as a planning input rather than an afterthought.

The Audience Data Problem

Programmatic’s targeting proposition rests on audience data: the ability to find and reach specific people based on their behaviour, interests, demographics, or intent signals. In theory, this is a significant advantage over traditional media buying. In practice, the quality of third-party audience data in the open programmatic ecosystem varies enormously, and much of it is less reliable than it appears.

Third-party data segments are built from a range of sources, including browsing behaviour, app usage, purchase signals, and modelled inferences. The further you get from a direct, verified signal (someone who actually bought your product, or submitted a form on your site), the more the data is an approximation. Segments labelled “in-market for financial products” or “interested in home improvement” may be based on one or two loosely related page views from weeks ago. The label sounds precise. The underlying signal is not.

The deprecation of third-party cookies, which has been delayed repeatedly but is still progressing in various forms across browsers, is forcing a long-overdue reckoning with this. Advertisers who built their programmatic targeting almost entirely on third-party cookie-based segments are now working out what their audience strategy actually looks like when that infrastructure is no longer available.

First-party data is the obvious answer, and it is the right one. CRM audiences, site visitor segments, email lists matched to programmatic identifiers, and contextual targeting based on content relevance rather than user tracking are all more durable and, in most cases, more accurate than third-party segments. The challenge is that building first-party data infrastructure requires investment and organisational commitment that many advertisers have deferred in favour of the easier option of buying audience segments from a data marketplace.

Early in my career, I built a lot of things myself because budget was not available to buy them. That experience taught me that understanding how something works at a fundamental level changes how you use it. Marketers who understand where their audience data actually comes from make very different decisions about which segments to trust and which to treat with scepticism.

Measurement and Attribution: Where the Numbers Lie

Programmatic advertising has a measurement problem that the industry has been slow to address honestly. The default attribution model in most programmatic campaigns is last-touch or last-click, which assigns conversion credit to the final ad interaction before a purchase or lead event. For display advertising, this creates a structural distortion.

Display impressions, by nature, are not typically the last touchpoint before conversion. They operate earlier in the consideration cycle, building awareness and familiarity that may influence a decision made days or weeks later via a different channel. Last-touch attribution almost always credits that conversion to paid search or direct, not to the display impression that contributed to it. This makes display look less effective than it is, and it makes paid search look more effective than it is.

The inverse problem also exists. View-through attribution, which credits a display impression if a user converts within a defined window after seeing (but not clicking) an ad, tends to overstate display’s contribution. Attribution windows of 30 days or more can pick up conversions that had nothing to do with the impression. Both measurement approaches are technically valid. Neither is accurate.

When I was running agency-side teams managing significant display budgets, we pushed clients toward incrementality testing: holding out a matched audience segment from programmatic exposure and comparing conversion rates between the exposed and unexposed groups. It is not a perfect methodology, but it is considerably more honest than view-through attribution with a 30-day window. The results were often humbling. Some campaigns showed genuine lift. Others showed that we were mostly reaching people who would have converted anyway.

Honest measurement in programmatic requires accepting that you will not always like the answer. Platforms that sell programmatic inventory have a structural incentive to present attribution models that flatter their channel. Advertisers need to apply their own measurement frameworks rather than relying on the platform to tell them whether the platform is working. Forrester’s work on intelligent growth models is a useful reference point for thinking about how measurement frameworks connect to commercial outcomes rather than channel metrics.

Private Marketplaces and Programmatic Direct

The open programmatic exchange is not the only way to buy programmatically. Private marketplaces (PMPs) and programmatic direct deals offer a middle path between the efficiency of automated buying and the quality controls of direct publisher relationships.

In a PMP, a publisher makes a curated selection of inventory available to a defined group of advertisers at a negotiated floor price, still transacted programmatically but with far greater control over placement quality, audience verification, and brand environment. Programmatic direct goes further, securing a fixed volume of specific placements at an agreed price, with the transaction handled through programmatic pipes rather than manual insertion orders.

For advertisers where brand environment genuinely matters, PMPs and programmatic direct are worth the premium over open exchange. The CPMs are higher. The quality of the inventory, the accuracy of the audience, and the transparency of the placement are all substantially better. For brand campaigns, sponsorship-adjacent placements, or any category where appearing next to low-quality content carries reputational risk, the open exchange is not the right default.

The practical question is how to balance the portfolio. A reasonable approach for most advertisers is to use PMPs and programmatic direct for brand-sensitive or high-value placements, and to use open exchange for lower-funnel retargeting where placement quality matters less than audience precision and cost efficiency. Treating the entire programmatic budget as open exchange is a choice that prioritises CPM over almost everything else.

Where Programmatic Fits in a Go-To-Market Strategy

Programmatic advertising is a media execution capability, not a go-to-market strategy. This distinction matters more than it might seem. Organisations that treat programmatic as a strategy end up optimising the channel rather than the outcome. They get better at buying cheap impressions rather than better at growing their business.

Programmatic works best when it is positioned within a clear commercial framework: what problem are we solving, for which audience, at what stage of the purchase cycle, and how does this channel connect to the other touchpoints in the customer experience. Without that framework, programmatic campaigns tend to drift toward whatever the platform’s default optimisation algorithm rewards, which is usually click-through rate or view-through conversions, neither of which maps reliably to business outcomes.

The channel is genuinely useful for building reach efficiently at the top of the funnel, for retargeting site visitors with relevant messages, for suppressing current customers from acquisition campaigns, and for supporting product launches or seasonal campaigns where speed and scale matter. It is less useful as a primary demand-generation channel for products with long consideration cycles, complex purchase decisions, or audiences that require genuine content engagement rather than banner exposure.

One pattern I saw repeatedly when working with clients across retail and financial services was the tendency to add programmatic to a media plan because it was expected rather than because it solved a specific problem. The channel was included because “we should be doing programmatic”, not because there was a clear hypothesis about what it would achieve. That is a reasonable way to waste a significant portion of a media budget.

Vidyard’s analysis of why go-to-market execution feels harder than it used to touches on something relevant here: the proliferation of channels has made it easier to be busy and harder to be effective. Programmatic is one of the channels that benefits most from strategic clarity before execution, because the automation will happily spend your budget without it.

There is also a scaling dimension worth considering. BCG’s research on scaling up makes the point that operational rigour matters more as organisations grow, not less. The same logic applies to programmatic: the discipline required to run it well (audience governance, brand safety controls, measurement frameworks, creative testing) becomes more important as spend increases, not less. Organisations that do not build those disciplines early tend to find that scaling programmatic spend just scales the problems alongside it.

If you are mapping programmatic into a broader growth architecture, the thinking around channel sequencing, budget allocation, and commercial prioritisation is covered in more depth across the Go-To-Market and Growth Strategy section of The Marketing Juice.

Creative Is Still the Variable Most Advertisers Underinvest In

Programmatic conversations tend to focus on targeting, bidding strategies, and platform selection. Creative is treated as a production task rather than a strategic variable. This is a consistent mistake, and it is one of the clearest patterns I observed across years of reviewing campaign performance data.

The targeting can be precise, the placements brand-safe, the bidding strategy well-configured, and the campaign will still underperform if the creative does not do a job. Display creative in particular suffers from a tendency toward lowest-common-denominator production: a product image, a logo, a call-to-action button, and a background colour. It is technically an ad. It is rarely an effective one.

Dynamic creative optimisation (DCO) has made it easier to test creative variables at scale, serving different combinations of headline, image, and offer to different audience segments and optimising toward the best performer. Used well, it is genuinely useful. Used as a substitute for creative thinking, it just optimises among a set of mediocre options and finds the least bad one.

The advertisers who consistently get the most from programmatic treat creative as a first-order variable, not a production afterthought. They brief creative with audience context and channel context in mind, they test meaningfully different creative approaches rather than minor variations, and they use performance data to inform the next creative cycle rather than just to report on the last one. Later’s work on creator-led campaigns is a useful illustration of how audience-first creative thinking applies across paid channels, including programmatic environments.

The Walled Gardens Question

Any honest discussion of programmatic advertising has to address the walled gardens: Google, Meta, Amazon, and to a lesser extent, the major retail media networks. These platforms operate their own closed programmatic ecosystems with proprietary data, proprietary measurement, and limited external verification. They are also where the majority of digital advertising spend goes.

The tension for advertisers is real. The walled gardens offer genuine advantages: large, well-defined audiences, strong targeting signals based on actual user behaviour (rather than inferred third-party data), and measurement frameworks that, while imperfect and self-reported, are at least based on real ad delivery. The trade-off is limited transparency, restricted data portability, and a measurement environment where the platform is effectively marking its own homework.

The open programmatic web, by contrast, offers more transparency and more control, but also more complexity and more exposure to quality issues. The right balance depends on the advertiser’s category, audience, and commercial objectives. Most advertisers benefit from a portfolio approach: using walled garden programmatic for the audience scale and targeting precision it genuinely offers, while using the open web for reach extension, retargeting, and placements where contextual relevance or premium publisher environments add value.

Forrester’s analysis of go-to-market challenges in regulated categories is a useful reminder that channel strategy cannot be separated from audience and regulatory context. In categories where data privacy, consent management, and targeting restrictions apply, the walled garden versus open web decision has compliance dimensions as well as commercial ones.

I have seen advertisers swing too far in both directions: abandoning the open web entirely in favour of walled garden simplicity, and refusing to use Meta or Google programmatic on principle while running open exchange campaigns with minimal brand safety controls. Neither extreme serves the commercial objective. The question is always what the channel can actually do for this audience, at this stage, with this creative, measured against these outcomes.

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