Why dependencies disappear between teams

A dependency is any output that one team must deliver before another team can start or finish its own work. Most dependencies are agreed in a meeting or a message and then tracked from memory, which works until two teams remember different dates. The coordinator usually learns a dependency is late only when the downstream team misses its own commitment. A register moves those promises into one list that both teams can see before the date is at risk.

The register is not a project plan, and it should not duplicate every task in the portfolio. It records only the handoffs that cross a team boundary and carry a date another team is waiting on. Keep it deliberately small so people trust the entries and keep them current. A register that copies the whole backlog becomes a second, stale plan within a month.

What belongs in the register

A useful entry starts with four facts: what is being handed over, who is waiting for it, who must deliver it, and the date it is needed. Add one line describing what the receiving team will do once the handoff lands, because that context helps a reviewer judge urgency. Link the entry to the request or ticket that contains the detail rather than pasting the whole history. The register should be readable in a single screen scan.

Record the person who owns delivery, not only the team name. A team without a named owner is a queue, and a queue cannot answer a question about whether the work has started. Add a fallback name for holidays and handoffs so the entry does not become ownerless at the worst moment. This one field prevents most of the chasing that fills a coordinator's week.

Set one status vocabulary

Use a short status set that every team interprets the same way. A workable set is not started, in progress, blocked, delivered, and accepted, with a separate flag for a changed date. Resist adding a status for every nuance, because a long list invites people to choose the label that sounds best. The coordinator should be able to read the status column and know which entries need attention today.

Define each status in one sentence and give an example of the evidence required to move into it. In progress should mean work has actually begun, not that someone intends to start it this week. Delivered means the output exists and the receiving team can use it, while accepted means the receiving team has confirmed that it works. Without those distinctions, status updates become reassurance rather than information.

Record the needed-by date, not the promised date

The needed-by date belongs to the receiving team, because it reflects when their work must begin. The promised date belongs to the delivering team and may move for good reasons. Keep both fields and show the gap between them, since a growing gap is an early warning that the handoff is drifting. A register with only one date hides the negotiation that matters most.

When the two dates differ, write one sentence explaining the change and who agreed to it. That note lets a later reviewer see whether the new date was accepted or simply announced. It also stops the same conversation from restarting every week. A date change without a note is indistinguishable from an oversight. Put the note in the entry itself, not in a separate thread that will be lost.

Review the register twice a week

Review the register on a fixed twice-weekly rhythm rather than waiting for an escalation. Read every entry that is due within the next ten working days and confirm its status with the named owner. Spend the meeting on entries where the date or the scope has moved, not on entries that are progressing normally. Thirty minutes is usually enough when the list is short and the owners are named.

Send the updated register to both teams immediately after the review, including the entries that did not change. A stable, predictable update reduces the number of side questions and makes the next review faster. If a team repeatedly misses the review, raise that as a coordination problem rather than an individual failing. The register depends on attendance more than on tooling.

Escalate only the dependencies that move

Escalate a dependency when the date is at risk and the two teams cannot agree on a new one. Bring the entry, the evidence of progress, and the downstream commitment that would break if the handoff slips. Ask the decision maker for a choice between moving the date, reducing the scope, or accepting the downstream impact. A clear question produces a decision, while a general warning produces another meeting.

Do not escalate every red status, because an escalation channel that fires constantly stops being read. Keep a threshold, such as a missed date or a gap larger than five working days, and apply it consistently. The priority dispute rulebook is the right place to settle competing claims once the dependency is genuinely contested. The register identifies the conflict; it does not resolve it.

Keep the register small enough to trust

Prune the register at the end of each month by removing entries that were delivered and accepted. Archive them with their dates rather than deleting them, because the history shows how long handoffs usually take. Look for one dependency that appears every cycle and ask whether it should become a standing commitment. A recurring entry is a sign that the process, not the date, needs attention.

Train each new coordinator on the fields and the review rhythm in their first week, using two real entries as examples. Ask them to write one entry themselves and review it before they run a review meeting. The weekly blocked-work review is a natural place to reuse the same register. Consistency across coordinators is what makes the list credible to the teams.

Sources