How to compare candidates across interviews without re-reading your notes

Comparing candidates across interviews means putting the same facts about each person side by side: stage, rate or salary, notice period, location, the concerns raised. This guide is for recruiters, hiring managers and freelancers who hire contractors for clients, running several rounds of interviews on Google Meet or Microsoft Teams. It covers why notes make that comparison slow, how to set up a table that does it instead, and how TrackHeros keeps the table current.

Why is it hard to compare candidates from interview notes?

Because notes are written per interview and a comparison runs across them. Five candidates through three stages is fifteen sets of notes, and the fact you need, the day rate the second candidate quoted, is in one of them.

So the comparison happens from memory, and memory has favourites. The most recent interview feels strongest, the most fluent candidate seems the most experienced, and a figure said once in a screening call is remembered rounder than it was. A table with the same columns for every candidate, each cell tied to the moment it was said, takes memory out of it.

Should a candidate tracker have one row per candidate or one per interview?

One per candidate. A row per interview is a log: useful for "what happened on Tuesday", poor for "who is cheapest", because each candidate's figures are spread over three rows that may disagree.

Trackers in TrackHeros come in both shapes. A new row for every meeting suits a standup log. The same rows, updated meeting after meeting, suits candidates: each row is named by the candidate's name, and when a later interview mentions the same person, their row is updated. The answer on one row per interview covers when the other shape is the right one, and building an interview tracker covers the columns.

What happens to a candidate's row at the next interview?

It is updated in place. A newer interview's value becomes the current one, the older value moves into the cell's history with the meeting, the speaker and the words, and a value said again is recorded as a confirmation rather than a change.

One candidate for a backend contractor role at a fintech client, across three rounds:

Column	Screening, Tue 6 Oct	Technical, Tue 13 Oct	Final, Tue 20 Oct
Stage	Screening	Technical	Final
Day rate	550 EUR	not mentioned	600 EUR
Notice	Four weeks	not mentioned	Free from Monday 2 November
Concerns	Little Kubernetes	Little Kubernetes	Removal waiting for review

After the final, the day rate reads 600 EUR, and its history shows 550 EUR from the screening with both quotes. In the final she walked the panel through a cluster she runs and an interviewer said that settled it, so TrackHeros proposed removing the Kubernetes concern. Removing an item from a list always waits for a person.

Two column rules are especially useful for candidates:

• Keep the first value on a column such as "Rate first quoted". If the candidate names a different figure later, the table shows a conflict for you to decide rather than quietly replacing it. Beside a "Current rate" column that follows the latest value, you see both the anchor and the move.
• Lists for strengths and concerns, so each interviewer's points add up across rounds and a concern disappears only when someone says it is resolved.

What if the same candidate appears under two names?

The tracker flags it rather than guessing. When a new row looks like an existing one, "Sam Okafor" and "Samuel Okafor", it is shown as a possible match, and you answer "same as" or "new row".

Rows are never merged automatically, because merging two different people is worse than a duplicate. When a new row may be an existing candidate, it waits for you to answer same as or new row, and same as keeps its history under the existing row. Describing the key column as "the full name of the person being interviewed, not the interviewers", as the Interview scorecard template does, prevents most duplicates in the first place.

How do you read the comparison in the table?

Open the tracker and read across. Search finds a candidate by name, and filters narrow the table to rows changed recently, rows changed several times, rows with conflicts, and rows with values waiting for review.

Every cell opens its history: each value, the interview it came from, who said it and when. A link opens the transcript at that moment and the recording jumps there, so "did she say 600 or 650?" takes one click rather than a search through notes. Each interview's meeting page also lists what that interview changed in each tracker, which is the quickest check straight after a call.

For a spreadsheet you can sort, filter by formula or send to a hiring manager, the current table and its full history download as CSV, which opens in Excel or Google Sheets. Pricing says which plans include export.

Can I ask which candidate asked for the highest rate?

Yes. The assistant chat reads trackers, so "who asked for the highest rate?" or "which candidates can start before December?" is answered from the table, with the interview each value came from.

The answer is only as complete as the table, so clear the review queue first: a value still waiting for review is not yet the cell's value, and the assistant answers from current values, marking any cell with a change waiting for review. Treat an empty cell as "not said in an interview", never as zero. The assistant reads trackers and does not change them. More on asking questions of a client's meetings is on ask your meetings.

How do you keep the comparison fair?

Ask every candidate the same core questions, so the same cells can fill. A tracker is told to fill only what the meeting actually says and to leave everything else empty, so a candidate you never asked about notice simply has no notice period. An uneven table usually means uneven interviews.

Keep judgement out of the columns too. Let the recommendation fill only when the panel states one, and write evidence rather than impressions in your feedback; how to write interview feedback and interview notes without losing the candidate cover both.

How does TrackHeros keep candidates for different clients apart?

TrackHeros is an AI meeting notetaker that connects every Google and Microsoft calendar you hold, recognises close to 100 spoken languages on its own, writes the notes in nine of them, and keeps every client's notes in a sealed workspace of its own. Every tracker belongs to one client's workspace, so candidates for Client A and Client B sit in separate tables.

A tracker can cover all your clients with one definition. TrackHeros then keeps a separate table for each client, and the Trackers page shows them together with a client label on every row. Two candidates with the same name at two clients are never merged. Whether your plan includes trackers across all clients is on pricing.

The tracker reads each interview after its notes are ready, not during the call. An interview held in Spanish or Hindi feeds the same columns as one in English, read through the same transcript as the notes, with text values written in the notes' language. The full list is on the languages page, and the recruiters page covers the rest of a recruiter's week.