How Compare works in Interview Signal: candidates side by side

Compare in Interview Signal puts one role's scored candidates in a single table: scorecard areas down the side, candidates across the top, best fit first, with each candidate's score in every area and the first quote behind it. Your own scores are marked, each column links to that candidate's review screen, and "Copy as text" turns the table into plain text for an email. Everything is read from the interview files on your computer. Compare arrived with the new home screen in 0.9.8 and is current in version 0.9.10 as of October 2026.

How do I open Compare?

1. On the home screen, click a role in the "Roles" panel. The right side shows that role's candidates, best fit first.
2. Press "Compare" in the header of that list. The button shows how many candidates it will include, such as "Compare 4".
3. If the button is grayed out, hover over it: it needs two scored candidates. An interview with no fit score cannot be compared.

When a role has more than six scored candidates, Compare takes the six with the highest fit. Past six columns a table stops being readable.

What does the Compare table show?

• The header reads "<role> — compared", with "← Back" to return to the role and "Copy as text" on the right.
• Each column is a candidate: the name, the fit score out of 10 in the same colors as the home screen (green at 8 or above, amber from 5, red below 5), when they were interviewed, "sent" if the submittal went out, and an "Open" button that goes to their review screen.
• Each row is a scorecard area. If an area's weight is more than 1, the row says so, for example "weight 3".
• Each cell shows the score in color, a "you" badge if the score is one you set yourself, the candidate's first quote for that area, and your note if you wrote one. A dash means that area was not scored for that candidate.

The corner of the table says "Best first. Quotes are the candidate's own words." They are: every quote shown has already passed the scorecard's check against the transcript, so a quote in Compare is something the candidate said word for word.

Which scores does Compare use?

The same ones the review screen shows. Where you changed an area's score with "+" during or after the interview, your score is used and marked "you"; elsewhere it is the model's score. The fit score at the top of each column is the one saved with the interview.

How should I read the table?

• Start with the heavy rows. A one-point gap on an area weighted 3 moves the fit more than a two-point gap on an area weighted 1, so read the weighted rows first.
• Look under the number. A model score above 5 always has a quote under it, because an area with no verified quote is capped at 5. A score of 5 or less with no quote means no verified evidence was found for that area; "Copy next round" on that candidate's review screen lists it for the next interview.
• Watch the "you" badges. A score you set yourself is not held to the quote rule, so check that your note says why.
• Treat dashes as gaps, not zeros. An unscored area is left out of that candidate's fit rather than counted against them.

What happens if the role's scorecard changed between interviews?

Each interview keeps the kit it was run with, so Compare reads each candidate's own scorecard. Areas are matched by name: rows follow the order of the newest interview's scorecard (the role's current one), and areas that only older kits had are added after it. If you renamed an area halfway through a search, it shows as two rows rather than silently merging two different things, and each candidate has a score in only one of them.

How do I share the comparison?

Press "Copy as text". The clipboard gets a heading, "<role> — candidates compared", and a table with one row per area, a column per candidate and a final "Fit" row, for example:

The numbers above are an example. The copied table uses Markdown's pipe format, so it reads as a table in plain text and displays as one wherever Markdown is supported. It carries scores and fit only; the quotes stay in the app and in each submittal. For the written case for one candidate, use the client submittal; for a method to decide between finalists, see how to compare candidates after interviews.

Is anything uploaded to make the comparison?

No. Compare reads the interview folders on your computer; nothing is fetched and nothing is sent. Interview Signal's servers never hold transcripts, quotes or candidate names to compare in the first place, as the security page explains.

Limits

• Six candidates at most, chosen by highest fit; you cannot pick which ones.
• Only scored candidates, and only within one role.
• Only interviews run on this computer. A colleague's interviews for the same role are on their machine.
• In 0.9.10 Compare opens from the home screen, not from an interview's review screen.
• The text copy has no quotes.