The Data Is Not a Byproduct

by Nadia OkonkwoIndustry Provocateur, HMND

Summary

The industry keeps pretending data is something produced on the side, then cleaned up later if anyone has the patience. That habit is now the bottleneck, because multi omics integration, federated data, and semantic governance all point to the same truth: the data itself is the operating asset, and most organizations still cannot move a target, biomarker, and outcome through their stack without human translation and political damage.

Body

There is a familiar lie in this industry. A team runs a study, a platform team wires a pipeline, a standards group files another glossary, and everyone acts as if the hard part is extracting insight from data that arrived fully formed. It did not. The hard part is that the data was never treated as the core object of the business. It was treated as exhaust.

That is why adoption is painful. Multi omics integration does not fail because people lack enthusiasm. It fails because each layer of the stack speaks a different dialect of the same biological reality. One group names the target one way, another encodes the biomarker differently, and outcomes live somewhere else entirely, flattened into codes that were never built to carry context. By the time anyone tries to connect them, the meaning has already been negotiated by hand. That is not an analytical workflow. That is an administrative rescue mission.

Federated data was supposed to reduce friction. In practice, it often exposes how little shared structure exists in the first place. You can move computation closer to the data, but if the semantics are unstable, you have only distributed confusion. The data stays in place. The disagreement travels faster.

Semantic governance is where the industry starts to get honest, or should. Not because it sounds elegant, but because it forces a brutal question: what exactly are we claiming this record means, across systems, time, and teams? Without that discipline, every new tool just adds another opinionated layer between the biological signal and the decision. The stack gets taller. The truth gets thinner.

That is what failure looks like. Teams keep multiplying tools because tools feel like progress. They are easy to buy, easy to demo, easy to present. A shared ontology is harder. It is less visible, less theatrical, and far more important. Without it, every dataset is rich and every decision surface is dumb. The tables are full. The dashboards are busy. The organization still cannot answer a basic question without a meeting.

Operators feel this every day. They see abundance everywhere and usability nowhere. A sample has genomic depth, transcriptomic depth, clinical history, longitudinal follow up, and still no clean path from observation to action. Someone has to translate the same concept three times for three systems. Someone has to explain why the biomarker that matters in one context disappears in another. Someone has to defend why an outcome is not computable, only narratable. That is not a technical inconvenience. It is a structural insult.

The real frustration is not lack of data. It is that the data is valuable, but the organization has built systems that cannot recognize value without human intervention. So the experts become interpreters, the interpreters become bottlenecks, and the bottlenecks become political objects. Then everyone wonders why transformation is slow.

It is slow because most companies are still running biological complexity through social compromise and calling it interoperability.

The delusion is that better models will fix bad structure. They will not. If the ontology is weak, if the semantics drift, if the federated rules are inconsistent, if the integration work lives in spreadsheets and hallway arguments, then the company is not data rich. It is coordination poor.

That is the part nobody wants to say out loud.

If you have lived this from the engineering side or the R and D side, the annoyance is probably familiar in the same exact way. The data is there, the value is there, and yet the stack still makes simple questions feel expensive. Worth comparing notes on where the real friction sits, and what finally made the work legible instead of merely accumulable.