You cannot fix a comment section you have never measured. Most brands judge theirs by feeling — it seems busy, it seems mostly positive, someone replies when they can. Then a customer mentions the scam link sitting under yesterday’s post, and the feeling turns out to have been wrong.
A comment audit replaces the feeling with six numbers. It takes about an hour, needs no tools beyond a spreadsheet, and it works whether you have 200 followers or 200,000.
This is the exact process I use before recommending anything to a client. Do it once and you will know precisely what your comment section is costing you.
Key Takeaways
- A comment audit means sampling a fixed set of recent posts and counting six specific things — not scrolling until you feel reassured.
- The six numbers: spam rate, unanswered question rate, first-response time, negative share, off-hours share, and competitor mentions.
- Sample 10 posts over the last 30 days, including your best-performing one and any paid post. Skip the quiet ones and you will audit a fantasy.
- Unanswered questions are the number that maps most directly to lost money — each one is a person who raised their hand and got ignored.
- Count comments in every language your audience actually uses. An English-only review misses most of what is happening in a Bangladeshi comment section.
- Repeat the audit monthly. A single snapshot tells you where you stand; the trend tells you whether anything you changed worked.
What is a comment section audit?
A comment section audit is a structured review of a fixed sample of your recent social posts, in which you count specific categories of comment and measure how quickly each was handled. It converts a vague impression into numbers you can compare month over month.
It is deliberately boring. That is the point — boring is repeatable.
Why you cannot judge your comment section by feel
Three things make impressions unreliable here, and all three work in the same direction: they make things look better than they are.
You see the top comments, not all of them. Facebook and Instagram both rank comments, and ranking favours engagement. The two-word spam reply from a fake account sits far below the fold where nobody scrolls.
You check when you are working. Your comment section does not keep office hours. If you look at 10 a.m., you are looking at a section that has already been tidied by whoever was awake.
You are not the target of the harm. A competitor’s “we offer this cheaper, inbox us” reads to you as cheeky. To a first-time buyer comparing options, it reads as useful information.
What you need before you start
Nothing exotic. A spreadsheet, access to your Page, and one uninterrupted hour.
Set up seven columns: post link, total comments, spam, questions, unanswered questions, negative, and off-hours. One row per post.
Do not automate this on the first pass. Reading real comments with your own eyes is where the insight lives, and you only need to do it manually once to know what to look for afterwards.
Step 1: Pick your sample properly
Take 10 posts from the last 30 days. Not your ten favourites — a real mix.
- Your single best-performing post. Highest reach means the widest audience saw whatever was in the comments.
- Every paid or boosted post in the window. You paid to send strangers there; those comments matter most.
- Two product or price posts. These attract buying questions and competitor poaching.
- One post that went wrong — a complaint thread, a delay announcement, anything with heat in it.
- The rest at random, to keep yourself honest.
Thirty days is the right window. Go shorter and one unusual week distorts everything; go much longer and you are measuring a version of your business that no longer exists.
The six numbers that matter
Count these for each post, then total them across the sample.
1. Spam rate
The share of comments that are promotional, bot-generated, scam links, irrelevant tags, or fake-account replies.
Count anything a reasonable customer would recognise as junk. If you are unsure whether something counts, it counts — the reader will not give it the benefit of the doubt either.
Different platforms attract different junk, and it helps to know which spam types show up under brand posts before you start labelling.
2. Unanswered question rate
Of the comments that asked a genuine question, what share never received a reply?
This is the number I look at first. Every unanswered question is a person who raised their hand in public and was ignored in public.
Count “price?”, “delivery koto din?”, “available?” and every variation. They are not small talk — they are buying signals.
3. First-response time
For questions that did get answered, how long did it take? Record the median, not the average.
One reply that took three days will drag an average into nonsense. The median tells you what a typical customer actually experienced.
4. Negative comment share
The share of comments expressing a complaint, frustration, or public criticism.
A number above zero is normal and healthy. What matters is whether each one got a reply, and whether the same complaint keeps recurring — that second pattern is a product problem wearing a moderation costume.
5. Off-hours share
What share of all comments arrived outside your working hours — evenings, nights, weekends, holidays?
This is the number that surprises people most. In Bangladesh, a great deal of social activity happens after 9 p.m., which is precisely when nobody is watching the Page.
6. Competitor and poaching mentions
Comments where someone else — a competitor, a reseller, an opportunist — is redirecting your audience elsewhere.
Count “inbox me”, “we have it cheaper”, and rival page tags. On a boosted post, this is a direct transfer of your ad budget to somebody else.
How to count without any tools
Open each post, expand every comment thread, and read from the bottom up.
Bottom-up matters. Ranking pushes the worst content down, so reading in reverse shows you what a scrolling customer eventually reaches.
Click “View more comments” until it disappears, and expand every “1 reply” link — nested replies are where poaching usually hides.
Tally in your spreadsheet as you go, one row per post. Resist the urge to fix anything yet; you are measuring, and stopping to reply will wreck both your timing and your count.
What the numbers actually mean
These thresholds are my working rules of thumb from moderating Bangladeshi brand pages — practical judgement, not published research. Treat them as a starting point and adjust to your own industry.
- Spam above 10% of comments — your section is visibly unmanaged to a new visitor.
- Any unanswered question rate above 20% — you are losing money you already paid to acquire.
- Median first response beyond 4 hours — most buying intent has cooled before you arrive.
- Off-hours share above 40% — office-hours moderation cannot cover your audience, no matter how diligent your team is.
- Competitor mentions on any paid post — fix this first; it is the most expensive leak on the list.
If several of these land in the warning zone at once, the problem is capacity rather than effort. That is usually the point at which the signs your brand needs AI moderation stop being theoretical.
The language blind spot that ruins most audits
Here is where audits done in English quietly fail.
In a Bangladeshi comment section, a single thread will mix Bengali script, romanised Bangla, English, and emoji — often inside one sentence. “Price koto?” is a buying question. “Vai eta scam” is a reputational emergency.
If you only counted the English comments, your spam rate and your unanswered-question rate are both understated, possibly badly.
Count in every language your audience uses. This is also the reason automated tools trained only on English perform poorly here, which I have written about in more depth in moderating Bengali and mixed-language comments.
What to fix first
You will finish the audit with a list longer than your week. Fix in this order.
- Paid posts. Every leak here costs you twice — once for the ad, once for the lost sale.
- Unanswered questions. Go back and answer them, even the old ones. People do return.
- Visible spam and scam links. These damage trust with readers who never comment at all.
- Off-hours coverage. Not by asking staff to work nights — by changing how the gap is covered.
- Recurring complaints. If the same issue appears repeatedly, moderation is treating a symptom.
Before you change anything, write down what you are committing to. A short comment moderation policy settles what gets hidden, what gets answered, and what gets escalated — so decisions stop depending on who happens to be online.
A worked example
Here is what a typical first audit looks like for a mid-sized retail Page.
Ten posts, 1,240 comments. Spam came in at 14%, mostly repeated promo replies from a handful of fake accounts. Questions: 186, of which 61 were never answered — a 33% miss rate.
Median first response was 6 hours 40 minutes. Off-hours share was 52%.
The costly detail was narrower than any of those figures. On the two boosted posts alone there were nine “inbox me, cheaper” replies from the same competitor, sitting under ads the brand had paid to promote.
Nobody on that team was lazy. They were covering a 24-hour comment section with an 8-hour team, which is arithmetic rather than attitude.
When to stop doing this by hand
The manual audit is the right first step every time. Doing it monthly forever is not.
Once you know your numbers, the question becomes whether the gap is worth closing with people or with automation — and that is a straightforward cost comparison rather than a matter of taste. I have broken down both sides of it in the real cost and ROI of comment moderation.
What you should not do is skip the audit and buy a tool. Without your baseline six numbers, you will have no way of proving whether the tool helped.
Your one-hour checklist
- Build a spreadsheet with seven columns.
- Pick 10 posts from the last 30 days, including every paid post and your best performer.
- Read each thread bottom-up, fully expanded, in every language.
- Tally spam, questions, unanswered questions, negatives, off-hours and competitor mentions.
- Work out your six totals and your median response time.
- Compare against the thresholds above and mark your warning zones.
- Answer every unanswered question you found — today.
- Diarise the same audit for next month and compare.
The first audit is uncomfortable. That discomfort is the most useful thing you will get from it, because it is specific — and specific problems are the kind you can actually fix.
What to do about competitor comments specifically
Of the six numbers, this is the one people handle worst — usually because they overthink it.
The instinct is either to argue publicly or to delete and hope nobody noticed. Both are mistakes, and for the same reason: they hand the other party more attention than they earned.
Hiding is almost always the right tool here. A hidden comment stays visible to its author and their friends, so it does not trigger the “they deleted my comment” escalation — but nobody else browsing your post will see it.
Deleting is for scams, doxxing and abuse. Banning is for the repeat offender who keeps returning under the same account.
The distinction matters more than it sounds, and getting it wrong is how a minor nuisance becomes a public argument. I have set out when each action is appropriate in hide, delete or ban on Facebook comments.
One thing worth doing during the audit itself: note the account names. If three of your nine poaching comments came from one account, you do not have a moderation problem so much as a single persistent individual — and that is a five-minute fix.
How the audit changes for Instagram and WhatsApp
The same six numbers work across channels. What changes is where the risk concentrates.
Instagram shifts the weight toward spam and impersonation. Comment sections attract bot replies at a higher rate, and fake accounts using your brand name in their handle are far more common than on Facebook.
Your off-hours number also tends to look worse on Instagram, because usage skews later into the evening.
WhatsApp inverts the exercise entirely. There is no public comment section, so spam and poaching largely vanish — but response time becomes the dominant metric, because a customer messaging privately expects a reply far sooner than one commenting publicly.
If you are active on more than one platform, audit them separately and never average the results together. A healthy Facebook page and a neglected Instagram account will produce a comfortable-looking combined number that describes neither.
What your audit tells you about a brewing crisis
The six numbers are a monthly health check. Read as a trend, they are also an early-warning system.
Two patterns are worth watching for specifically.
A negative share that climbs while volume stays flat. This usually means an operational problem — late deliveries, a faulty batch, a billing change — surfacing in public before it reaches your support inbox.
A sudden spike in comment volume on an older post. Traffic returning to a post you published weeks ago is rarely organic. It generally means someone has shared it somewhere with commentary attached.
Both are far easier to handle at the point they appear in your numbers than at the point they appear in your mentions. What to do when one does escalate is a separate discipline, which I cover in comment crisis management when a post blows up.
Frequently Asked Questions
How long does a comment section audit take?
About an hour for 10 posts, assuming a few hundred comments in total. Larger pages with thousands of comments per post should sample within each post rather than reading everything — take the first 100 comments and the last 100.
How many posts should I include in the sample?
Ten posts across the last 30 days is the sweet spot. Fewer than that and one unusual thread skews your numbers; many more and the exercise becomes too tedious to repeat monthly, which defeats the purpose.
Which of the six numbers matters most?
The unanswered question rate. It maps most directly to lost revenue, because every entry is a person who publicly asked to be sold to and was ignored. Competitor mentions on paid posts come a close second.
Should I count comments in Bengali and romanised Bangla separately?
Count them together in your totals, but note the split. If most of your comments are not in English, any tool or agency you evaluate later must be tested specifically on mixed-language threads rather than on English samples.
What is a normal spam rate for a Facebook page?
It varies hugely by industry and by how much you advertise, so treat any single benchmark with suspicion. As a practical rule, once spam passes roughly 10% of visible comments, new visitors start reading your page as unmanaged.
Do I need a tool to run this audit?
No, and you should not use one for the first pass. Reading real comments manually teaches you what your particular audience does, which is the knowledge you need to judge whether any tool is actually working later.
Know your numbers before you buy anything
Run the audit. Bring your six numbers. If they show a gap that people alone cannot close, that is the conversation worth having — and it will be grounded in your own data rather than a sales pitch.