Results Chain Extractor

The problem

Organizations write down why they expect a strategy to work: do this, which leads to that, as long as certain conditions hold. That reasoning is the most valuable thing in a plan or an evaluation, and it is usually buried in dozens of pages of prose.

Search can find the words in a report. It cannot find the reasoning. So nobody can easily check whether a plan’s logic holds together, or compare the logic of one project with another.

What we built

The Results Chain Extractor takes a report or plan and returns its chain of cause and effect as a diagram a person can review and correct.

• It reads the whole document, including diagrams and tables, not just the text.
• Every step is tied to a quote. The tool checks each quote against the document and flags any step it inferred or could not find support for.
• A person stays in charge. The result is a draft. The reviewer edits steps, adds missing ones, and marks each as reviewed.
• Terms are lined up. Steps, places and organizations are linked to shared reference lists, so two documents that use different words for the same thing can be compared.
• The result is reusable data. A reviewed chain exports in a standard format, ready to be combined with others.

It is built on the Results Chain Ontology, an open-source model of how strategies lead to results, published by Erik Anderson.

Where this applies

The first use is conservation planning, where the results chain is a standard tool. The same approach fits any organization whose reasoning is locked in documents: strategy papers, project reviews, evaluations and lessons-learned reports.

What comes next

Two additions are planned and not yet built: finding past projects that used the same logic, to warn of risks seen before, and totaling the assumptions across a whole portfolio to show where many projects depend on the same fragile condition.

The live tool needs an access code to read a new document. Its built-in example opens without one.