Originally published April 2017. Updated August 2026 to reflect changes in programmatic buying, audience data, privacy, inventory quality, measurement, and attribution.
Programmatic advertising has changed quite a bit since I originally wrote this article in 2017. The technology is better, the data ecosystem is different, privacy has become a much bigger part of the conversation, and programmatic has expanded well beyond traditional display advertising.
What hasn’t changed is the basic problem.
Programmatic can be complicated, and sometimes that complexity makes it really easy to sell.
A DSP can buy millions of impressions across thousands of websites, apps, streaming services, video platforms, and other digital properties. A campaign can generate a tremendous amount of data and produce a beautiful report. But none of that necessarily means you purchased good media.
As a buyer, you still need to understand what you are actually getting for your money.
So here is my updated Programmatic Buyer’s Guide to the Galaxy: ten questions I would ask before handing a programmatic vendor, platform, or agency my media budget.
1. What Are You Actually Trying to Buy Programmatically: Content, Audience, or Both?
One of the first questions that needs to be asked when buying programmatic is still the same question I asked almost ten years ago: What exactly are you trying to buy, and how?
Programmatic refers to the automated process used to buy and sell digital advertising. In 2017, that conversation was primarily about display advertising purchased through real-time auctions. Today, programmatic includes display, online video, connected TV, native advertising, audio, mobile, digital out-of-home, private marketplace inventory, programmatic guaranteed deals, and quite a bit more.
But programmatic describes how the media is purchased. It does not necessarily describe who you are trying to reach.
Are you targeting content? Are you targeting behavior? Are you using first-party customer data? Are you targeting a publisher’s audience? Are you using contextual signals? Are you buying modeled audiences? More likely, you are using some combination of all of them.
Contextual buying targets consumers based on what they are consuming at the time the ad is delivered. If someone is reading content about buying a new truck, there is a reasonable assumption that the person may have some interest in trucks.
Behavioral targeting attempts to identify the consumer based on previously observed behaviors and other signals.
First-party targeting starts with something your organization already knows about a customer or prospect and uses that information to create or find an audience.
None of these approaches is inherently better than the others. The important thing is knowing which one you are buying.
Don't accept “we're targeting auto intenders” as an answer.
Ask how they know those people are auto intenders.
2. Do You Know How Your Audience Is Actually Built?
This was important in 2017. It is probably even more important today.
One of the original examples I used in this article was an auto-buyer audience we created at Advance Digital. The audience carried a frequency of three and a recency of seven. In other words, we needed to observe three qualifying behaviors within seven days before we considered that person an active member of the audience.
The point wasn't that three and seven were magical numbers. The point was that we actually knew what went into the audience.
That's still what I want to know today.
If a platform tells me it has an audience of “small-business decision makers,” I want to know how those people were identified. Did the audience originate from publisher behavior? First-party customer data? Purchase data? Search behavior? Content consumption? A third-party data provider? Is it modeled from another audience?
How recently was the behavior observed? How many signals are required before someone qualifies? How much of the audience is deterministic versus modeled?
Audience names can sound fantastic on a PowerPoint slide. That doesn't make the data good.
Years ago I would have asked for a contributing behavior report from the vendor or DMP. Today, the terminology and technology may be different, but I would still ask for essentially the same thing: show me how this audience is constructed.
If the vendor can't explain it, that's a problem.
3. How Much of Your Budget Is Prospecting Versus Remessaging?
In the original version of this article I called this creative remessaging. Today, most marketers would probably call it retargeting or remarketing, but the concern is still valid.
Remarketing works.
Someone visits your site, looks at a product or service, leaves, and then sees your ads again. It can be an extremely effective way to build frequency and bring people back into the buying process.
It can also make a mediocre programmatic campaign look fantastic.
Retargeting existing website visitors is generally easier than finding and converting completely new customers. If a programmatic vendor combines prospecting and retargeting into one campaign report, the retargeting portion can make overall conversion performance look much stronger than the prospecting media actually was.
I have seen campaigns where a very large percentage of the impressions were being allocated to remessaging because it was cheaper, easier to deliver, and made the overall numbers look better.
Ask the question.
How much of my budget is being spent finding new customers, and how much is being spent talking to people who already know me?
Then ask to see the performance separately.
4. Where Is Your Inventory Actually Coming From?
This question has changed substantially since 2017.
Back then, I talked about buying through ad networks versus exchanges such as AppNexus, OpenX, and DoubleClick. Some of those names don't even exist in the same form anymore.
Today, I would ask about the entire supply path.
What DSP is being used to purchase the inventory? What SSPs or exchanges are involved? Are you buying through the open exchange? Are you using private marketplace deals? Are there direct publisher relationships? How many intermediaries exist between my advertising dollar and the publisher actually showing my ad?
This is where programmatic gets complicated very quickly.
Every technology provider sitting between the advertiser and publisher potentially adds cost. That doesn't necessarily mean the intermediary isn't providing value, but as a buyer I want to understand where my money is going.
There are also much better transparency tools today. Ads.txt helps publishers identify companies authorized to sell their inventory. Sellers.json helps buyers understand the companies involved in selling that inventory. Supply-chain information can provide even more visibility into the path an impression takes before it gets to the advertiser.
You don't need to become an ad-tech engineer.
But your vendor should be able to explain the supply chain in plain English.
If they can't tell you where your ads are running or how the inventory is being purchased, I would start asking more questions.
5. Are You Buying Cheap Impressions or Good Impressions?
Don't be fooled by cheap CPMs.
I wrote that in 2017, and I still believe it.
A $2 CPM isn't automatically better than an $8 CPM. It just means you bought 1,000 impressions for less money.
The question is what those impressions were worth.
Programmatic inventory is priced based on demand, supply, audience, format, placement, quality, geography, viewability, and dozens of other variables. If your only objective is finding the lowest CPM possible, the technology can probably accomplish that.
I'm not sure you will like what you bought.
There is an enormous amount of inventory available on the open internet. Some of it comes from excellent publishers with highly engaged audiences. Some of it comes from sites built primarily to generate advertising impressions. Some placements have great visibility. Others technically load an advertisement that almost nobody will ever notice.
That is why CPM needs context.
I would rather pay more for an impression that reaches the correct consumer in a quality environment than pay almost nothing to serve an advertisement nobody sees.
Don't ask your programmatic partner simply what CPM they can get.
Ask what happens to inventory quality as the CPM changes.
6. Can You Measure Viewability, Attention, and Inventory Quality?
Viewability was one of the biggest programmatic conversations when I originally wrote this article. It still matters, but the conversation has become considerably more sophisticated.
Viewability basically answers whether an advertisement had the opportunity to be seen.
That's important. If I'm paying for an impression, I would prefer that somebody actually have the opportunity to see it.
But viewability isn't the same thing as attention.
An ad can technically be viewable while the consumer pays absolutely no attention to it. That's why the industry is increasingly looking at attention measurement alongside traditional metrics such as viewability, reach, frequency, engagement, and conversion.
I don't think attention should replace every other metric. I do think it gives buyers another signal for understanding the quality of what they purchased.
At a minimum, ask whether your vendor can provide viewability reporting. Ask whether the measurement comes from a recognized independent measurement provider. Ask for the information by domain, placement, creative, and format when possible.
Then look at what happens.
If one portion of the inventory has a terrible viewability rate, poor engagement, questionable domains, and no conversions, why are you continuing to buy it?
Programmatic advertising creates an enormous amount of data.
Use it.
7. What Is Your Strategy for Fraud, MFA Inventory, and Brand Safety?
In 2017, I wrote about fraudulent clicks and recommended asking vendors for domain reports. I still recommend looking at domain-level reporting, but fraud prevention has come a long way since then.
Today I would make the question broader.
What is your strategy for invalid traffic? What technology are you using to identify fraud? How are you handling made-for-advertising sites? How are you protecting the brand from inappropriate content? Are you validating authorized sellers? Can I see where my impressions were delivered?
Sometimes questionable traffic is surprisingly easy to identify.
If a random website you've never heard of suddenly generates a click-through rate five times higher than everything else in the campaign, don't immediately celebrate your fantastic performance.
Look at it.
A domain report remains one of the easiest ways for a marketer to get a better understanding of what is happening behind the campaign averages.
There are also tools today that weren't widely available when I originally wrote this article. Ads.txt, sellers.json, supply-chain data, third-party verification, invalid-traffic detection, brand-safety controls, inclusion lists, and exclusion lists all give buyers more control.
Use them.
Programmatic shouldn't mean handing someone your money and hoping their algorithm figures everything out.
8. What Is Your Attribution Strategy?
Consumers are fickle, and they tend to forget.
I wrote almost that exact sentence in the original article, and this is one section where the underlying problem hasn't changed very much.
In fact, attribution may be harder now.
A consumer might see a programmatic display ad, Google the company later, watch a YouTube video, read a Reddit thread, ask an AI platform about the product, visit the website directly, leave, see another ad, and finally convert through branded search.
Who gets credit?
If your answer is simply “the last thing the consumer clicked,” you probably don't have the full picture.
Programmatic display and video often influence consumers before they are ready to convert. That means judging every campaign purely by last-click conversions can make upper- and mid-funnel advertising look less valuable than it really is.
Attribution analysis still has a place, but I would also look at incrementality testing, conversion lift, geographic testing, brand lift, assisted conversions, and when budgets become large enough, marketing mix modeling.
The question I care about is not just, “Did someone click my ad and buy?”
I want to know whether the advertising changed consumer behavior.
That's a harder question.
It's also the one marketers should actually be trying to answer.
9. What Happens to the Audience Data Generated by the Campaign?
This is another area that has changed dramatically.
In 2017, I talked heavily about DMPs, cookie pools, clicker audiences, and the ability to collect consumers who viewed or clicked advertising into audiences that could be targeted again.
That technology helped create some incredibly sophisticated audience strategies.
It also helped create a privacy problem the industry is still working through.
Today, first-party data has become considerably more valuable. Privacy regulations, browser restrictions, platform changes, consent requirements, and the decline of traditional third-party tracking have changed what advertisers can and should do with consumer data.
The question, however, still matters.
What are you learning from the people who interact with the campaign?
Can campaign engagement be connected to your own first-party customer data in a privacy-compliant way? Can you build remarketing audiences? Can the platform create modeled audiences from high-value customers? Can publisher first-party data be used? Are clean-room technologies involved? How is consent being handled?
And probably most importantly: who owns the data?
If you're paying for the campaign, you should understand what happens to the data created by the campaign and whether that knowledge can be used to improve future marketing.
10. Can the Campaign Be Integrated With Your Website Analytics and Business Results?
Every dollar spent online should have some connection back to what happens on the website and, eventually, what happens to the business.
In 2017, my recommendation was basically: use campaign URLs and UTM parameters so you can see the traffic in your website analytics.
I still recommend doing that.
But website analytics should now be considered the starting point rather than the finish line.
Your programmatic campaign should have a consistent campaign naming structure. Click traffic should be identifiable in GA4 or whatever analytics platform you're using. Conversion events should be configured correctly. Important lead or transaction data should flow back into the advertising platform when technically and legally appropriate.
For lead-generation businesses, I also want to know what happened after the form was submitted.
Did the lead actually answer the phone?
Was it qualified?
Did it become an opportunity?
Did it turn into revenue?
A campaign that generated 100 leads isn't necessarily better than a campaign that generated 40. If only two of the 100 leads were qualified while 15 of the 40 turned into real sales opportunities, the campaign with fewer leads was considerably more valuable.
The closer you can connect programmatic advertising to actual business outcomes, the better decisions you can make.
Clicks are nice.
Revenue is better.
Programmatic Has Changed. Being a Smart Buyer Hasn't.
When I originally wrote this article in 2017, programmatic advertising was already complex. Almost ten years later, it has become significantly more powerful and significantly more complicated.
We're buying across more devices and formats. First-party data has become more important. Privacy requirements are stronger. Connected TV has become a major programmatic channel. Supply-chain transparency has improved. Viewability is now being supplemented by attention measurement. Attribution is being combined with incrementality and marketing mix modeling. AI and machine learning are making more campaign decisions automatically.
All of that technology is useful.
But technology doesn't eliminate the need for someone to ask good questions.
If anything, automation makes those questions more important.
Who are we actually targeting?
Where are the ads actually running?
How much are we really paying?
How much of the budget is reaching new customers?
Is the inventory any good?
Are people actually seeing the ads?
Can we trust the traffic?
Did the campaign actually change consumer behavior?
And, ultimately, did it help the business?
If your programmatic partner can answer those questions clearly, you're probably in pretty good shape.
If the answer to every question is some version of “the algorithm handles that,” I'd keep asking questions.
Because no matter how sophisticated programmatic advertising becomes, you're still the one paying for it.
Frequently Asked Questions About Programmatic Advertising
What is programmatic advertising?
Programmatic advertising is the automated buying and selling of digital advertising inventory using technology and data. It can include display, video, connected TV, native advertising, audio, mobile, and other digital media formats.
Is programmatic advertising the same as display advertising?
No. Display advertising is one type of media that can be purchased programmatically. Programmatic refers to the method used to buy advertising, while display refers to the advertising format.
How does programmatic audience targeting work?
Programmatic campaigns can use a combination of first-party data, contextual targeting, publisher data, modeled audiences, behavioral signals, geographic information, and other targeting methods. Advertisers should understand how an audience is created rather than relying only on the audience name provided by a vendor.
What is the difference between programmatic prospecting and retargeting?
Prospecting is designed to reach new potential customers who have not previously interacted with your brand. Retargeting focuses on people who have already visited your website, viewed a product, or taken another measurable action. The two should usually be measured separately because retargeting often converts more easily.
How can advertisers avoid low-quality programmatic inventory?
Advertisers should review domain and placement reporting, viewability, invalid traffic, brand safety, supply-chain transparency, and conversion performance. Tools such as ads.txt, sellers.json, verification platforms, inclusion lists, and exclusion lists can also help improve inventory quality.
What is viewability in programmatic advertising?
Viewability measures whether an advertisement had an opportunity to be seen by a user. It does not necessarily mean the user paid attention to the ad, which is why some advertisers now evaluate attention metrics alongside traditional viewability reporting.
How should programmatic advertising be measured?
Programmatic campaigns should be measured against the actual business objective. Depending on the campaign, that may include conversions, qualified leads, revenue, return on ad spend, assisted conversions, incrementality, brand lift, or other outcomes. Clicks and impressions alone usually do not provide enough information.
Is programmatic advertising still effective in 2026?
Yes, but the way it is used has changed. Privacy requirements, first-party data, connected TV, supply-chain transparency, AI-driven optimization, and measurement changes have all reshaped programmatic advertising. The technology remains powerful, but advertisers still need to understand what they are buying and how success is being measured.
