Domains Under Management is the default health metric for the domain industry. It is simple, comparable, and universally understood. It gives boards a growth number, analysts a ranking, and registries a headline they can defend in a quarterly review.

DUMs earned that position for good reason. It measures something real: the installed base of a namespace. How many domains exist in the zone right now. Whether that number is growing, shrinking, or holding steady. For market sizing, competitive benchmarking, and investor communication, DUMs works.

But in conversations with registry operators over the past several months, a question has come up repeatedly that DUMs does not answer: what does that installed base actually consist of?

What DUMs Measures Well

It is worth being specific about this, because the argument is not that DUMs is a flawed metric. It is that DUMs is a narrow one.

DUMs tells you the size of the namespace. It tells you the direction of growth. It tells you how you compare to peers on raw volume. It provides a clean denominator for revenue-per-domain calculations. And it is available to anyone with access to public reporting or a zone file.

For these purposes, DUMs is excellent. No structural analysis, no classification model, and no compositional lens will replace the basic need to know how many domains are registered in a namespace. That number matters, and it should continue to be tracked.

Where DUMs Reaches Its Limits

The limitation is not in what DUMs measures. It is in what it compresses.

A namespace with four million registrations might contain a healthy mix of business-operated domains, brand-protection registrations, investor holdings, and some structural noise. Another namespace with a similar count might be dominated by promotional-price acquisitions with low renewal probability, a substantial speculative layer, and a thin operator base. DUMs treats both identically.

In recent structural work across multiple nTLD namespaces, I observed that namespaces with broadly comparable registration volumes could have operator-to-investor balances that varied by nearly fourfold. Their noise profiles differed not just in volume but in composition. The largest segment in every namespace was a structurally neutral middle that could represent either genuine adoption or inventory waiting to churn. None of these differences appeared in the registration count.

This is not a theoretical concern. When a registry evaluates whether a promotional campaign improved the namespace, DUMs will show whether registrations increased. It will not show whether the new registrations are structurally similar to existing adoption-grade inventory or whether they added primarily to the speculative and noise layers. When a board asks whether a TLD is healthy, DUMs provides a size answer. It does not provide a composition answer.

Why This Matters Operationally

In conversations with registry professionals over the past several months—including practitioners responsible for premium inventory and namespace strategy—a version of the same observation has come up repeatedly: registration volume can be inflated by aggressive pricing, and that inflation often comes at the cost of namespace quality. This is not a controversial claim. Most experienced operators understand it intuitively.

What is less commonly discussed is that the industry does not yet have a widely adopted way to measure that quality. DUMs remain the primary health metric in most registry dashboards, investor reports, and industry analyses. Renewal rates add a second dimension but still operate at the aggregate level. Neither metric distinguishes between a namespace that is growing through durable business adoption and one that is growing through promotional volume with high structural churn risk.

The consequence is that strategic decisions about pricing, registrar incentives, premium inventory, and market positioning are often made against a metric that describes the size of the namespace but not its character.

What a Composition Layer Might Look Like

The argument here is not that registries should replace DUMs. It is that they might benefit from an additional measurement dimension that sits alongside it.

Structural composition analysis offers one version of what that might look like. By classifying registered inventory into categories based on observable zone file characteristics, it becomes possible to track not just how many domains are in the zone, but what types of domains they are and how that mix is changing over time.

The specific categories matter less than the principle. Whether a registry uses a four-layer model, a three-tier approach, or an entirely different taxonomy, the underlying question is the same: what is the namespace actually made of, and is that composition moving in the direction the registry wants?

For example, a registry running a promotional campaign could compare the structural profile of domains registered during the campaign to the profile of the existing namespace. If the campaign-acquired inventory is structurally similar to the adoption-grade layer, it is likely adding durable value. If it is structurally similar to the noise or speculative layers, the registration volume increase may not survive the first renewal cycle.

A composition layer would not need to be complex. Even tracking two or three basic ratios over time—the proportion of domains that appear adoption-grade, the proportion that appear speculative, and the proportion that are structural noise—would add a dimension that DUMs alone cannot provide.

What We Don’t Know Yet

Honest disclosure: the structural composition work published so far, including my own, has demonstrated that namespaces have measurably different compositions. It has not yet demonstrated that tracking composition improves registry decision-making.

That is an important gap. Observing that a namespace is 55% neutral-middle inventory is analytically interesting. Whether knowing that number changes a pricing decision, a registrar incentive structure, or a renewal campaign is an operational question that has not been answered with evidence.

The hypothesis is plausible: if you can see that a promotional campaign added primarily to the noise layer rather than the operator-grade layer, you might make different decisions about the next campaign. But plausible is not proven, and the gap between structural observation and operational impact is where the real work remains.

It is also possible that some registries already track composition-like metrics internally that are not publicly visible. The argument here is not that this thinking is absent from the industry—it is that it has not yet become a standard, shared measurement practice.

The Question Worth Asking

DUMs will remain the industry’s headline metric, and it should. It measures something essential. But the question worth asking is whether it should remain the only structural metric most registries track.

If two namespaces can report similar DUMs while having materially different compositions—different operator-to-investor balances, different noise profiles, different proportions of durable versus transient registrations—then DUMs is describing the size of the namespace without describing its character. Both dimensions matter. Only one is currently measured as standard practice.

The industry measures namespace size with precision. Whether it is ready to measure namespace character with the same discipline is an open question—but one that the data increasingly suggests is worth answering.