Hockey + Technology / Field note
Connecting Scouting, Video, and Data to a Coaching Workflow Staff Can Trust
Scouting notes, video, and event data become useful together when a coach can follow an observation back to its
evidence without reconstructing the entire analysis. The goal is not a universal dashboard. It is a repeatable
path from a specific question to a reviewable packet and an accountable decision. That path requires shared
definitions, reliable identifiers, and ownership of the handoffs between departments.
Design backward from the coaching meeting
Start with the meeting and its deadline. Ask who will use the material, what question they need to answer, and
how much context each clip must preserve. Decide whether the output is an opponent-preparation packet, a shared
definition review, or a follow-up list for analysts. Keep the scope to authorized senior professional hockey
material; a broad data collection exercise is not a prerequisite.
Define a packet that can be reviewed without navigating multiple systems: the question, an observation, supporting
and contradictory clips, relevant event context, uncertainty, and a proposed next check. Preserve the distinction
between what a scout observed, what a metric calculated, and what a coach decides. Combining those fields should
not turn a subjective label into an apparently objective fact.
Build the source contract before the interface
For each provider, document the source namespace and original IDs for games, events, video assets, and annotations.
Create a crosswalk rather than assuming matching names or timestamps mean matching records. Retain the original
identifier alongside any internal one. Record mapping confidence and unresolved matches, and allow for one event
to have multiple relevant clips. Quarantine ambiguous joins instead of choosing the nearest event silently.
Time alignment needs its own specification. A period and displayed game clock are not the same as video elapsed
time. Stoppages, replays, camera changes, and edited files can break a naive offset. Store period, clock direction,
video timecode, clip boundaries, and alignment version. Test examples across those transitions. A correctly
computed metric attached to the wrong play is a failed integration.
The contract should also specify tag definitions, units, coordinate orientation, author or responsible role,
capture time, revision history, correction policy, and permitted use. Check rights to store, transform, annotate,
share, and export the material. Protect the original record and version the derived interpretation. An annotation
change should not overwrite the evidence that supported a previous meeting packet.
Assign ownership where errors actually occur
The source owner handles licensing and upstream questions. An integration owner maintains mappings and ingestion
checks. A video owner verifies alignment and clip availability. A hockey subject-matter reviewer resolves
definitions, while the coaching workflow owner accepts the packet. Security staff approve access boundaries.
One person may cover several responsibilities, but none should remain implicit.
When tags disagree, retain the disagreement and use a documented adjudication process. Sample the same clips with
independent reviewers to discover where the vocabulary is unclear. Agreement alone does not prove that a tag is
useful; the meeting owner must confirm that the distinction helps answer the actual question. Otherwise the
integration simply automates inconsistent terminology.
Illustrative scenario: a forecheck preparation packet
Imagine an invented workflow in which a professional coaching group asks what supporting evidence exists for an
opponent forecheck tendency. This is not a client project or an assertion about a real team. The scout supplies
an observation and its scope. An analyst identifies eligible events and records exclusions. The video coordinator
checks clips before the group discusses the interpretation.
The packet would include examples that support the observation, examples that challenge it, and clips with
unresolved context. Each would carry source references and a clear tag version. If a critical support player is
outside the camera view, the packet would say the action cannot be established. The coach could request another
review or select a preparation question without the system inventing a tactical certainty.
Make acceptance criteria observable
Agree on the test set and tolerances before implementation. Include routine plays, clock transitions, corrected
events, missing video, duplicate imports, and users with different permissions. Every accepted test packet should
resolve its cited sources, play the intended sequence within the agreed alignment tolerance, and display the
current approved interpretation. Missing evidence should remain visible rather than disappear from the count.
Require repeatable output from the same permitted inputs and versions. Test that source corrections flag affected
packets for review, that restricted users cannot retrieve clips through exports or cached links, and that a named
owner receives an actionable failure notice. Measure preparation and review effort against the existing process.
Do not claim adoption from logins alone; ask whether staff can complete the meeting task and explain the evidence.
Integrate in a narrow lane, then earn expansion
The June 25, 2026 NHL-Presidio announcement describes data integration and cloud infrastructure supporting draft
and broadcast operations. The NHL Innovation Lab announcement describes a setting for testing technology. These
public examples show operational integration and testing in specific league initiatives, not a scouting system
available to consultants or proof of an individual club need.
Begin with a read-only packet alongside the established workflow. Keep the manual fallback, train the people who
own corrections, and review rejected packets before expanding to another use. The strongest handoff is one that
staff can inspect, correct, and maintain after the pilot team leaves.
Public sources and scope
The workflow above is an independent proposal, not a description of the systems in these announcements.
Discuss a scouting, video, and data integration review built around a real
coaching task, clear ownership, and testable acceptance criteria.