In the summer of 2013 the company I worked for bought a tag management system, and I barely noticed. I was at Adobe, on the Experience Manager side of the house, deep in what mobile was doing to the market. The Satellite acquisition landed for many of us on the product team like a memo: hey, we have this now. My read then, and still, is that it was a buy-versus-build call. Adobe Analytics and Adobe Target each loaded their features onto customer sites their own way, and the fastest path to one approach was to acquire it. Satellite became Dynamic Tag Management, and I went back to thinking about apps.
That is my confession, and it is also my thesis. The tag manager reshaped who controls the digital storefront more than almost any technology of its decade, and the people building the storefronts, me included, were not paying attention. That is what makes a Trojan horse work. It does not look like a siege engine. It looks like a gift.
The problem it genuinely solved
To be fair to the horse, the walls had it coming. By the late 2000s every analytics pixel, ad beacon, affiliate tracker, and testing tool required its own snippet of JavaScript pasted into the site. Every new vendor meant a ticket into the engineering queue, a release cycle, and a marketer waiting weeks to launch a campaign that lived or died on timing. Multiply by dozens of tools and you get the chaos the whole category was invented to end.
The fix was elegant: put one container tag in the page, once, and manage everything else from a web console. TagMan, founded in 2007, was one of the earliest companies built around that idea. The wave broke in 2011. Tealium, founded by WebSideStory alumni from the analytics story I told in the measurement history, launched Tealium iQ on July 26, 2011, and its own press release is the artifact I would put in a museum: the “industry’s first self-service tag management solution,” letting marketers manage tags “without having to burden IT resources.” A consultancy called Search Discovery publicly introduced Satellite the same year, and Qubit shipped Opentag free that August. Then in October 2012 Google shipped Google Tag Manager at no cost, and the category went from niche to default in a single announcement.
The consolidation that followed tells you who won. Adobe bought Satellite in 2013. Ensighten, the enterprise player founded in 2009, absorbed TagMan in 2014. BrightTag renamed itself Signal in 2014 and pivoted toward customer data, an early tell that tags were becoming a data-plumbing business. Adobe bought its tag manager; Google built its own; every platform that took the storefront seriously ended up with one. Read that Tealium release again, though, because the revolution was never really about tags. “Without having to burden IT resources” is the whole ballgame. The product was a deployment pipeline for marketing, and nobody guarding the production gate read it that way.
Marketers have wanted the storefront for 130 years
This is not a story about JavaScript. It is a story about who controls the storefront, and it is much older than the web.
In 1897 L. Frank Baum, three years before The Wonderful Wizard of Oz, launched The Show Window, the first American trade magazine devoted to dressing store windows. He organized the National Association of Window Trimmers the next year and served as its secretary, and in 1900 published the field’s first real manual, The Art of Decorating Dry Goods Windows and Interiors. His pitch to shopkeepers was blunt: everything turns on “the value of your window display as an attraction and salesman.” The window was not decoration on the building. It was a selling machine, and it deserved craft, budget, and motion: lifelike wax figures instead of inert draped dummies, electric light, mechanical illusions that stopped crowds on the sidewalk. The trade would later name the idea the “silent salesman.” Baum himself quit the magazine in October 1900 when Oz hit and never looked back, which I have always found perfect: the man professionalized the storefront and then left to build the most famous illusion in American culture.
I love this history for the same reason I loved the measurement history. The pattern repeats. Merchandisers and marketers always want the window; the people who own the building always treat it as a wall with a hole in it. When the money poured into digital in the 2010s and a generation of fit-for-purpose vendors sprang up to serve marketers, chronicled year over year in Scott Brinker’s ever-expanding martech supergraphic, the marketers finally got their tools. The tag manager was how the tools got into the store without a ticket to the building manager.
Who actually owned it
Here is the part the vendor histories leave out. In every company I have worked for and nearly every retailer I have ever sold to, the tag manager is owned by marketing. Not shared. Owned. Gartner saw the money moving early; a Gartner analyst made the famous 2012 prediction that by 2017 chief marketing officers would outspend chief information officers on technology, and I remember absorbing that at Adobe as an interesting market fact. It took me years to connect it to the mechanism. A marketing team with a tag manager and an agency partner that employs a couple of front-end developers is a software delivery organization. It ships JavaScript to production on the highest-traffic pages the company owns, on its own schedule, through a console the engineering team does not review and often cannot see. The consoles themselves are not the culprit; enterprise tag managers have shipped approval workflows, permissions, and rollback for a decade. Engineering was simply never in the approval chain, so the controls sat unused.
Nobody decided that on purpose. Marketing championed tools because attribution and personalization demanded them, and because the tools often paid for themselves in measurable revenue, which is why the horse kept getting fed. The digital team inherited whatever those tools did to the shopping experience. Engineering discovered, usually during an incident, that there was a second deploy pipeline in the building. Each group owned a piece; nobody owned the whole. That vacuum, not any single slow script, is the real cost of the Trojan horse.
The day I finally saw it
I said I was not paying attention at Adobe. I started paying attention at Gap.
I came to Gap Inc. in 2020 to build personalization at scale, and personalization is where every sin of the tag stack comes due, because it only works if the site can recognize a shopper and react before she loses interest. Instead I found what most large retailers of that era had: a storefront, oldnavy.com, where in my memory you could count to nine before the page settled. Engineering had recently shipped a fast new React storefront and was rightly proud of it. But the experience shoppers actually received arrived through the tag layer stacked on top: a personalization vendor still serving its content as HTML through a JavaScript tag, Tealium’s tag manager doing heavy lifting it was never designed for, and default content flashing on screen while the clever stuff loaded behind it. The build was fast. The experience was not. And the real-user data that would have told that story was not on any dashboard I was handed.
That was my conversion moment, in both senses. I got religion about server-side rendering, about sending an event once and letting vendors subscribe to it instead of loading a dozen beacons that each phone home separately, and about judging the storefront by what real users experience rather than what the framework benchmark says. The tag manager vendors reached the same conclusion; the server-side eventing products they sell today are the industry admitting that the browser was the wrong place for most of this work all along.
Years later at Yottaa, where the product’s job was making e-commerce storefronts fast and part of that job was sequencing when third-party tags load, I got to see the fleet-level version of what I had seen at one retailer. The Web Performance Index we built there aggregates 500 million real shopper sessions across more than 700 brands, and the picture is consistent: retailers run 40 or more third-party apps per page on average, and they account for nearly 60 percent of total load time. My Gap experience was not an outlier. It was the industry’s default configuration.
The reckoning that half arrived
Privacy law should have been the moment the walls got rebuilt. In Europe, the ePrivacy rules and the General Data Protection Regulation (GDPR) turned every tag into a legal question: this script collects data about a person, so on what basis does it fire? The California Consumer Privacy Act (CCPA) followed as the opt-out cousin. The tag manager sits in exactly the right place to enforce the answers, and for a while I expected consent management and tag management to merge into one governed gateway.
The plumbing did merge. Every major tag manager now carries consent state, and Google requires consent signals for its advertising features in Europe. But merged plumbing is not governance, and the audits prove it. Researchers crawling European sites found 49 percent installed profiling cookies before the user consented to anything, and a study of 22,949 sites running consent platforms caught sites registering a yes before the shopper had made any choice. That matches what my own eyes see: sites lighting up with vendor calls before I have touched the banner. The narrower audit I have never seen is the one this essay cares about, whether tag managers actually hold their payloads until consent, and I suspect the results would embarrass a lot of brands.
And the action has moved again. More and more of the data work now happens through server-to-server connections: partly for performance, and partly, frankly, to route around browser-level controls like ad blockers and tracking prevention. The consent signal can ride along on those connections. Whether it is honored is another matter, because no shopper, regulator, or researcher can watch what happens after the event leaves the browser. The browser was at least a glass box. The server is not. The horse is inside a different set of walls now.
Know who owns your window
The tag manager was never the villain. It solved a real coordination problem, and the people who built the category built good software. What it smuggled in was organizational: a second, unowned deployment pipeline running in the one place where milliseconds are money and where the customer is actually standing.
So if you run a storefront, digital or otherwise, the question is Baum’s, updated: who owns your window? Not who pays for the tools in it, and not who approved the container snippet years ago. Who is accountable for what a real shopper experiences when everything in that window fires at once? The companies I have worked for and sold to that can answer with one name, and give that name authority over marketing’s tags and engineering’s framework alike, are the ones whose windows still sell, usually because that same name is watching the real-user numbers every day. The ones that cannot answer are still unloading the horse.
This piece grew out of a chapter I drafted for The Commerce Chain, an eBook I co-authored during my time as CPO at Yottaa and co-own today. The history is expanded and corrected here; the opinions are mine.