Skip to content

Foray 382 · Purpose

What makes a project graph worth keeping?

End

In 2020 Lawrence kept an Oxford portfolio graph so that he could query a whole portfolio. His later architecture experiment added a way to show the result to someone else. The question now is whether a project reviewer can come back after months away and use one of those graphs in a meeting. Success means reaching the right records, recognising their limits and doing it within the ten minutes available.

Ways

  1. Start with the answer you expect. Record the question, the unit being counted and a source record you can check independently. A plausible picture can hide an unconnected architecture element or make one investment measure appear to belong to nine projects.
  2. Recover each representation on its own terms. Preserve the modelling choice that affects the answer. Keep a task distinct from its achieved state, a relationship distinct from its visual rendering and each federal export distinct from its neighbour.
  3. Run the same case twice. Compare counts, stable record identifiers and query results after reloading. A duplicated import can change the answer even when the drawing looks familiar. Record anything the historical material cannot establish.
  4. Time the return. Start the clock when opening the graph from a browser and stop when the checked result is in hand. Record the starting condition. Reopening a populated instance, resuming a paused one and rebuilding a missing one are three different trials. Try the timed scenario.

Means

The ArchiSurance CSVs supply elements, relationships and properties. Their typed links and isolated elements test whether Graphistry preserves what Neo4j can count. The maintained code is part of the first essai. Upstream ArchiSurance files · Architecture and Graphistry replay.

The February 2020 waste-removal scripts express one small job seven ways. The historical federal IT exports supply a larger case and an independently checked FDA question. Use the published research files for their stated purpose and keep the scope of the saved exports visible. Seven project graphs · Federal IT sources · FDA evidence check.

Neo4j Aura provides the browser workspace. The scripts and their source checks make the experiment recoverable after that workspace changes. Older Desktop, APOC and Graph Data Science settings can require a translation. Record the original operation, its replacement and the query that demonstrates equivalent behaviour. Connect to Aura.

A two-node portfolio example provides a quick test of a new connection. The original query and the later trial using the schema alone each return zero rows. The trial supplied with project names returns the proposed intervention. Future essais can revisit other archived graphs, including the Excel and Power BI connections in the earlier work. For each one, ask what decision the original graph supported and how much work it takes to reach that decision again. The human capital management query.