Compare comments across Instagram, TikTok, YouTube and Facebook in one sheet
The same video goes out on four networks and gets four different comment sections. Putting them side by side shows where the audience asks questions, where it only reacts and where it talks to each other. The exports share one set of columns, so the stacking is easy. Comparing fairly is the part that needs care.
Updated Nov 19, 2026 · 7 min read
Why the files stack without any mapping
Every export has the same thirteen columns in the same order, whichever network it came from: #, comment_id, username, user_id, comment, posted_at_utc, likes, is_reply, reply_to_comment_id, reply_to_username, mentions, hashtags, platform. The last one, column M, says instagram, tiktok, youtube or facebook on every row. That column is what lets four files share a sheet and still be told apart.
Export each version of the post from the any-platform export page. Per network, the pages for Instagram, TikTok, YouTube and Facebook list which links are accepted.
Stack the exports
Give each export its own tab, named Instagram, TikTok, YouTube and Facebook, with the header in row 1. In Google Sheets, File, Import each workbook and keep its Comments sheet, or paste with Copy for Sheets. Then add a tab named All and put this in A1.
Google Sheets:
=VSTACK(Instagram!A1:M1, FILTER(Instagram!A2:M, Instagram!B2:B<>""), FILTER(TikTok!A2:M, TikTok!B2:B<>""), FILTER(YouTube!A2:M, YouTube!B2:B<>""), FILTER(Facebook!A2:M, Facebook!B2:B<>""))
Excel (Microsoft 365):
=VSTACK(Instagram!A1:M1, FILTER(Instagram!A2:M20000, Instagram!B2:B20000<>""), FILTER(TikTok!A2:M20000, TikTok!B2:B20000<>""), FILTER(YouTube!A2:M20000, YouTube!B2:B20000<>""), FILTER(Facebook!A2:M20000, Facebook!B2:B20000<>""))
The FILTER around each tab drops the empty rows below the data. In Excel, format column F of the All tab as a date and time, since a spilled formula does not carry formats over. In older Excel, copy the rows of each sheet under one another by hand, or use Data, Get Data, Combine Queries, Append.
Comparing several posts per network works the same way. Add a post column to each tab before stacking and widen the ranges to column N.
A trap in Google Sheets: QUERY and mixed columns
QUERY gives each column one type. When a column holds numbers on some rows and text on others, it keeps the majority type and returns the rest as blanks, with no warning. In a stacked sheet that can happen in two columns:
user_id: Instagram IDs are digits and can arrive as numbers, while YouTube channel IDs start withUCand are text.comment_id: digits on Instagram, TikTok and Facebook, letters on YouTube.
The workbook stores both as text on every network, so a sheet built from workbooks is safe. If you pasted or used a live link, either keep ID columns out of your QUERY formulas (count M instead of B, as the examples below do), or use COUNTIFS and FILTER, which do not have this behaviour. More on ID types is in why spreadsheets round long comment IDs.
What is comparable and what is not
| Thing | Comparable? | Why |
|---|---|---|
| Top-level comments | Yes | Every top-level comment is read on every network. |
| Replies | With care | Instagram and YouTube exports read every reply. Free TikTok and Facebook exports read reply threads up to a limit, and the Read me sheet says how many were left unread. |
| Timestamps | Yes | All four are in UTC in the same format. |
| People, within a network | Yes | user_id is stable per network. |
| People, across networks | No | Nothing links one person's Instagram handle to their YouTube channel. On Facebook the username is a name plus four characters. |
| Likes on comments | Within a network | Liking a comment is common on YouTube and TikTok and rarer elsewhere. Compare shares, not raw likes. |
| Mentions | Not for Facebook | Facebook tags are names, not @handles, so the column is mostly empty there. |
| Comment length | With care | TikTok caps a comment at 150 characters. The other three allow thousands. |
| The count each network displays | No | Each one counts its own hidden and filtered comments differently. |
The practical rule: compare top-level comments, and compare rates (per comment, per person), not totals. A network where the post reached ten times as many people will win every raw count and tell you nothing.
The comparison table
On a tab named Summary, type the four platform values in A2 to A5 exactly as they appear in column M: instagram, tiktok, youtube, facebook. Then fill these across row 2 and copy them down. The Excel versions use fixed ranges; the first six work unchanged in Google Sheets too.
| Column | Measure | Formula in row 2 |
|---|---|---|
| B | Rows | =COUNTIF(All!M:M, A2) |
| C | Top-level comments | =COUNTIFS(All!M:M, A2, All!H:H, "no") |
| D | Reply share | =IF(B2=0, "", (B2-C2)/B2) |
| E | Question share | =SUMPRODUCT((All!M$2:M$20000=A2) * (All!H$2:H$20000="no") * ISNUMBER(SEARCH("~?", All!E$2:E$20000))) / C2 |
| F | Share that tags someone | =COUNTIFS(All!M:M, A2, All!H:H, "no", All!K:K, "?*") / C2 |
| G | Share from the brand | =COUNTIFS(All!M:M, A2, All!C:C, "fernhill.market*") / B2 |
| H | People | Sheets: =COUNTUNIQUE(FILTER(All!D2:D, All!M2:M=A2))Excel: =ROWS(UNIQUE(FILTER(All!D2:D20000, All!M2:M20000=A2))) |
| I | Median length of a top-level comment | Sheets: =MEDIAN(FILTER(LEN(All!E2:E), All!M2:M=A2, All!H2:H="no"))Excel: =MEDIAN(IF((All!M2:M20000=A2) * (All!H2:H20000="no"), LEN(All!E2:E20000))) |
| J | Hours until half the comments were in | Sheets: =(MEDIAN(FILTER(All!F2:F, All!M2:M=A2)) - MIN(FILTER(All!F2:F, All!M2:M=A2))) * 24Excel: =(MEDIAN(IF(All!M2:M20000=A2, All!F2:F20000)) - MINIFS(All!F2:F20000, All!M2:M20000, A2)) * 24 |
The trailing * in column G lets the same handle match the Facebook form with its four extra characters. Column J needs posted_at_utc as a real date; test with =ISNUMBER(All!F2). Format D to G as percentages.
For a quick cross-check of columns B and C in Google Sheets, one formula builds the split by itself:
=QUERY(All!A:M, "select M, count(M) where M is not null group by M pivot H", 1)
Reading the table: a worked example
Fernhill Market posts one recipe video everywhere. The summary comes out like this:
| platform | Top-level | Reply share | Question share | Tags someone | Median length | Hours to half |
|---|---|---|---|---|---|---|
| 412 | 18% | 6% | 41% | 24 | 5 | |
| tiktok | 280 | 9% | 4% | 22% | 17 | 31 |
| youtube | 96 | 34% | 21% | 1% | 88 | 52 |
| 143 | 27% | 15% | 0% | 46 | 9 |
Three readings. Instagram is where people tag friends: reach through the audience. YouTube has the fewest comments and by far the most questions and the longest ones: those viewers are cooking the recipe. TikTok's conversation took 31 hours to reach its halfway point where Instagram's took 5, because the video kept being shown to new people for days. That last number decides when to export: a TikTok read on day one misses most of the section.
The TikTok reply share of 9% deserves a check before it is quoted. If that export's Read me sheet has a Replies not read line, the true share is higher, and the fair statement is "at least 9%".
Go further
- What each audience talks about. Run the word frequency formula once per network by filtering on column M first.
- Tone per network. Add the labels from sentiment analysis in Google Sheets and count them with COUNTIFS on the label and
platform. - Open questions per network. The columns from find questions and buying intent work on the
Alltab if you paste it as values first. - A chart for other people. The
Alltab is the right shape for a Looker Studio dashboard, withplatformas a filter.
Before any of it, confirm each export is whole: the five checks in how to check an export is complete take a minute per file, and one short file skews every share in the table.
FAQ
Can I tell whether the same person commented on two networks?
No. Each network has its own IDs and handles, and nothing in public comments links them. Count people per network and do not add the four numbers together.
Do all four exports really have the same columns?
Yes, the same thirteen in the same order. Some are emptier on some networks: mentions on Facebook, for example, because Facebook tags are names and not @handles.
Why compare top-level comments instead of all rows?
Every top-level comment is read on every network, while free TikTok and Facebook exports read reply threads only up to a limit. Top-level counts are the like-for-like number.
When should I export so the comparison is fair?
At the same age for each post, for example seven days after publishing. Networks deliver comments at very different speeds, so exporting everything on the same calendar day favours the fast ones.