An input is any piece of raw material that comes into Glare: a report, a deck, an interview transcript, an analytics export, a survey file or an email. It's where every other piece of evidence starts.
For example, the product team receives an analytics export showing that only 12% of new SecondBank cardholders visit the rewards tab in their first month. That export is an input. Nothing has been concluded yet; it's simply what came in, from where, and when.
Why it matters
When inputs aren't tracked, nobody can say where a claim came from. A number gets quoted in a meeting, repeated in a deck, and three weeks later no one knows whether it came from analytics, a survey, or someone's memory of a call.
Tracking inputs gives every situation, need, signal and decision a trail back to its source. Leaders can check the evidence behind a call in one step, and the team can see when a decision rests on a single email instead of real data.
Log what came in, where it came from and when, then decide whether it's worth reading closely before anyone draws conclusions from it. In Collecting, what comes in is gathered into a Source Map: what the team knows, how far to trust each source, and where the gaps are.
How we write one
Every input is written as one short line that names what kind of material it is, where it came from, and when:
A source type from a source, a date.
Analytics export from the card app dashboard, October 2026.
Keeping the line this short makes inputs easy to scan and compare. What the input says is captured in the nouns pulled from it, not in the input itself.
How it differs from similar ideas
-
A situation is an interpreted reading of what's happening. An input is the raw material that reading is built from.
-
An observation is something seen or heard during testing. An input is the file or message that contains it, such as the test transcript.
-
A signal is evidence that has been interpreted and is ready to move forward. An input hasn't been interpreted at all.
-
An audience is who the evidence comes from. An input records which audience it covers, but the input is the material, not the people.
A well-logged input
-
Names the type of material
-
Says where it came from
-
Has a date
-
Names the author or owner
-
Says which audience it covers
-
Lists what was extracted from it
What happens to it
-
Received. Someone uploads it, forwards it, or Ray picks it up from a connected tool.
-
Catalogued. Ray records its type, source, date, author and the audience it covers.
-
Processed. Ray reads it and pulls out situations, needs, assumptions, observations and other nouns.
-
Skipped. Ray judges it has nothing new or useful, logs why, and leaves it unread in detail.
-
Superseded or archived. A newer version replaces it, or it ages out, and it's kept so old nouns still trace back to it.
For Ray
This part tells Ray how to catalogue an input, such as a report, a deck, a transcript, an analytics export, a survey file or an email, and how to decide whether it's worth processing.
How to find one
-
Identify the type. Name what kind of material it is: report, deck, transcript, analytics export, survey responses, email, roadmap, retro notes or other.
-
Identify the source. Record the system, tool or person it came from, such as the analytics dashboard, a research repository or an inbox.
-
Find the date and author. Use the date the material was created, not the date it was uploaded. Record the author or owner.
-
Identify the audience it covers. Note whose voices or behavior the material reflects: Project Team, Stakeholders, Customers or Participants.
-
Check for duplicates and versions. If the same material, or a newer version of it, is already logged, link to it instead of logging it again.
-
Decide whether to process it. Process it if it contains new evidence, direction or questions. Skip it if it repeats what's already stored, and record the reason.
-
Write the line. Use the format: [Source type] from [source], [date].
-
Fill the record, score it, link it and store it using the rules below.
Cues to look for
-
Files and messages arriving through upload, a connected tool or a forwarded email
-
Headers, footers and metadata that show author, date and system of origin
-
Material that reports behavior, opinions, results or direction
-
Attachments inside emails or decks, which are often separate inputs
What doesn't count
<table xmlns="http://www.w3.org/1999/xhtml" style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p>If the material...</p></th><th colspan="1" rowspan="1"><p>Store it as</p></th></tr><tr><td colspan="1" rowspan="1"><p>Is a reading of what's happening, written by someone on the team</p></td><td colspan="1" rowspan="1"><p>Input, and extract the Situation from it</p></td></tr><tr><td colspan="1" rowspan="1"><p>Is a single quote or behavior pulled from a test</p></td><td colspan="1" rowspan="1"><p>Observation, linked to the input that contains it</p></td></tr><tr><td colspan="1" rowspan="1"><p>Is a duplicate of something already logged</p></td><td colspan="1" rowspan="1"><p>A new source on the existing input, not a new input</p></td></tr><tr><td colspan="1" rowspan="1"><p>Is a link with no content Ray can open</p></td><td colspan="1" rowspan="1"><p>Nothing yet; add it to the review digest to request access</p></td></tr><tr><td colspan="1" rowspan="1"><p>Is a summary of other inputs, such as a findings deck</p></td><td colspan="1" rowspan="1"><p>Input, linked to the inputs it summarizes</p></td></tr></tbody></table>
The record
<table xmlns="http://www.w3.org/1999/xhtml" style="min-width: 75px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p>Field</p></th><th colspan="1" rowspan="1"><p>What to fill in</p></th><th colspan="1" rowspan="1"><p>Values</p></th></tr><tr><td colspan="1" rowspan="1"><p>Statement</p></td><td colspan="1" rowspan="1"><p>The one-line description</p></td><td colspan="1" rowspan="1"><p>[Source type] from [source], [date].</p></td></tr><tr><td colspan="1" rowspan="1"><p>Type</p></td><td colspan="1" rowspan="1"><p>What kind of material it is</p></td><td colspan="1" rowspan="1"><p>Report, Deck, Transcript, Analytics export, Survey, Email, Roadmap, Retro notes, Other</p></td></tr><tr><td colspan="1" rowspan="1"><p>Source system</p></td><td colspan="1" rowspan="1"><p>Where it came from</p></td><td colspan="1" rowspan="1"><p>Tool, system, repository or inbox</p></td></tr><tr><td colspan="1" rowspan="1"><p>Date</p></td><td colspan="1" rowspan="1"><p>When the material was created</p></td><td colspan="1" rowspan="1"><p>Date</p></td></tr><tr><td colspan="1" rowspan="1"><p>Author</p></td><td colspan="1" rowspan="1"><p>Who wrote or owns it</p></td><td colspan="1" rowspan="1"><p>Name or role</p></td></tr><tr><td colspan="1" rowspan="1"><p>Audience covered</p></td><td colspan="1" rowspan="1"><p>Whose voices or behavior it reflects</p></td><td colspan="1" rowspan="1"><p>Link to Audience</p></td></tr><tr><td colspan="1" rowspan="1"><p>Extracted</p></td><td colspan="1" rowspan="1"><p>The nouns Ray pulled from it</p></td><td colspan="1" rowspan="1"><p>Links to Situation, User Need, Assumption, Observation and others</p></td></tr><tr><td colspan="1" rowspan="1"><p>Processing</p></td><td colspan="1" rowspan="1"><p>Whether Ray read it in full</p></td><td colspan="1" rowspan="1"><p>Processed, Skipped (with reason), Pending</p></td></tr><tr><td colspan="1" rowspan="1"><p>Version</p></td><td colspan="1" rowspan="1"><p>Earlier or later versions of the same material</p></td><td colspan="1" rowspan="1"><p>Link to Input, or blank</p></td></tr><tr><td colspan="1" rowspan="1"><p>Status</p></td><td colspan="1" rowspan="1"><p>Where it stands</p></td><td colspan="1" rowspan="1"><p>Received, Catalogued, Processed, Skipped, Superseded, Archived</p></td></tr><tr><td colspan="1" rowspan="1"><p>Confidence</p></td><td colspan="1" rowspan="1"><p>How trustworthy the source and its details are</p></td><td colspan="1" rowspan="1"><p>High, medium, low</p></td></tr></tbody></table>
Inputs are entry points, so nothing links into them. Every other noun links back to the input it came from.
Scoring confidence
Confidence measures how reliable the material and its details are, such as whether the source and date are known, not how interesting the content is.
<table xmlns="http://www.w3.org/1999/xhtml" style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p>Level</p></th><th colspan="1" rowspan="1"><p>When to use it</p></th></tr><tr><td colspan="1" rowspan="1"><p>High</p></td><td colspan="1" rowspan="1"><p>Comes from a known system or author, with a clear date, and covers a named audience</p></td></tr><tr><td colspan="1" rowspan="1"><p>Medium</p></td><td colspan="1" rowspan="1"><p>Source is known, but the date, author or audience is missing or unclear</p></td></tr><tr><td colspan="1" rowspan="1"><p>Low</p></td><td colspan="1" rowspan="1"><p>Forwarded secondhand, undated, or of unknown origin</p></td></tr></tbody></table>
If any part of the record can't be filled from the source, leave it blank and lower confidence by one level. Every noun extracted from a low-confidence input starts no higher than medium.
Linking
-
Check for duplicates first. If the same file or message is already logged, add the new source location to it instead of creating another.
-
Link to the audience it covers.
-
Link to every noun extracted from it, recording the exact passage, slide or row each one came from.
-
Link versions. When a newer version arrives, link the two and mark the older one Superseded.
-
Link summaries to their sources. If the input summarizes other inputs, link to them so evidence isn't counted twice.
When to store and when to ask
-
Store automatically when confidence is high or medium.
-
Add to the review digest when confidence is low, when Ray can't open it, or when Ray skips something that came from a leader or a test.
When to flag
-
A decision rests on a single input
-
An input was processed but produced no nouns
-
A summary input is being counted alongside the inputs it summarizes
-
An input is older than 12 months and still supports an active decision
-
An input's audience is unknown
Worked example
Input: an analytics export attached to a message from the product analyst.
"Attached is the October pull from the card app dashboard. New cardholders, first 30 days: rewards tab visits 12%. Let me know if you want it split by acquisition channel."
What Ray finds:
-
"October pull from the card app dashboard" gives the type, source system and date.
-
"New cardholders, first 30 days" is the audience covered: Customers.
-
"Rewards tab visits 12%" is new evidence, so the input is worth processing.
-
Ray extracts a Situation and a Benchmark from it and links both back here.
-
"Want it split by acquisition channel" is an offer, not evidence, so nothing is extracted from it.
Stored record:
<table xmlns="http://www.w3.org/1999/xhtml" style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p>Field</p></th><th colspan="1" rowspan="1"><p>Value</p></th></tr><tr><td colspan="1" rowspan="1"><p>Statement</p></td><td colspan="1" rowspan="1"><p>Analytics export from the card app dashboard, October 2026.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Type</p></td><td colspan="1" rowspan="1"><p>Analytics export</p></td></tr><tr><td colspan="1" rowspan="1"><p>Source system</p></td><td colspan="1" rowspan="1"><p>Card app analytics dashboard</p></td></tr><tr><td colspan="1" rowspan="1"><p>Date</p></td><td colspan="1" rowspan="1"><p>October 2026</p></td></tr><tr><td colspan="1" rowspan="1"><p>Author</p></td><td colspan="1" rowspan="1"><p>Product analyst</p></td></tr><tr><td colspan="1" rowspan="1"><p>Audience covered</p></td><td colspan="1" rowspan="1"><p>Customers: new cardholders, first 30 days</p></td></tr><tr><td colspan="1" rowspan="1"><p>Extracted</p></td><td colspan="1" rowspan="1"><p>Situation (new cardholders rarely open the rewards tab); Benchmark (rewards tab visits 12%)</p></td></tr><tr><td colspan="1" rowspan="1"><p>Processing</p></td><td colspan="1" rowspan="1"><p>Processed</p></td></tr><tr><td colspan="1" rowspan="1"><p>Status</p></td><td colspan="1" rowspan="1"><p>Processed</p></td></tr><tr><td colspan="1" rowspan="1"><p>Confidence</p></td><td colspan="1" rowspan="1"><p>High: known system, named author, clear date and audience</p></td></tr></tbody></table>
Next step: stored automatically; the extracted Situation and Benchmark are linked back to this input.
Edge cases
-
One file, many inputs. A deck with an attached survey export holds two inputs. Log each one and link them.
-
Forwarded chains. An email that forwards a report is a pointer. Log the report as the input and note the email as where it arrived.
-
Live dashboards. A dashboard changes daily. Log each pull as its own dated input so old nouns keep their original numbers.
-
Nothing useful. If an input adds nothing new, mark it Skipped with a reason. Never delete it; a later reader may need to know it was seen.