This is a use case, not a framework. A two-van aircon installation business, a free Facebook comment download, a spreadsheet, and a month of booked jobs that came from comments nobody had answered. If you want the mechanics of the export itself, the free Facebook comment download guide covers them; this post is about what the file is worth in revenue.
The job that sat in a comment thread for three weeks
The installer found us because of one comment. A public post in a suburban Facebook group — someone asking for a recommendation for a mini-split install in a rented condo — had fifty-one comments. Seven were tagged friends. Four were installers dropping a phone number. One, thirty comments down, was the original poster replying: "still looking, the two who messaged me never came back with a price."
That comment was nineteen days old. The job was still open. Nobody had scrolled far enough to read it, because Facebook does not show you the bottom of a fifty-comment thread unless you insist, and it certainly does not sort it by who is ready to buy.
That is the entire premise of this use case. In any local market, demand is already typed out in public comment threads, and the reason it goes unserved is not competition — it is that reading comments in the Facebook UI is a terrible job, so almost nobody does it properly. A downloaded file is not a terrible job. It is a sortable list.
What a buying comment looks like on Facebook
Before touching a spreadsheet, it helps to know what you are scanning for. From the installer's own thirty-day sheet, the comments that turned into paid work were almost always one of four kinds:
- The asker. "How much for a 1.5hp including install?" A question about price, spec or availability. Highest-value row in the file, because they are mid-decision and they will buy from whoever answers first and clearly.
- The unserved. "Messaged you last week, no reply." Someone a competitor already failed. The easiest conversion in the file and the one most businesses never see, because it is usually buried deep in a thread.
- The location-stater. "Do you cover Cheras?" A qualified lead with the qualifying question already asked. Yes or no closes it.
- The complainer about someone else. "Used a cheap guy, gas leaked in two months." Not ready today, but they have just described their own next purchase, and they are cheap to stay in touch with via the thread.
And the rows that look like leads but are not: tagged friend names with no text, "PM sent" with no question, sticker-only comments, and the comment that is really a competitor's advert. The sheet exists to separate these four from those, in one pass, without scrolling.
The three files this runs on
Each week the installer pulls three exports, which takes about ten minutes total. Paste each post permalink into the Facebook comment exporter and take the Excel file.
File one: their own last post. Comments on your own posts are not just praise — they contain the questions you failed to answer in the caption, and the people who asked publicly and never got a reply. The reply_to_id column tells you exactly which comments the Page answered and which it ignored, which is an uncomfortable and useful report.
File two: a competitor's most active post. Public Page posts export the same way yours do. The target is the unserved rows: questions the competitor left hanging and complaints about their turnaround. You are not stealing customers, you are answering a question a stranger asked in public that nobody answered.
File three: one public group thread. This is the highest-intent source for local services, and it is why Facebook still matters for lead generation in a way it does not for brand marketing. Search the group for the recommendation posts, take the one with the most comments, export it. Private groups are out of scope — if you need a login to read it, it is not public data.
The five-minute triage
The sheet has one job: get from a few hundred rows to a short list you can act on before lunch. In the exported file:
- Sort by
created_at, newest first. Freshness beats volume for service work. - Filter the text column for a question mark. That single filter isolates most of the askers and most of the location-staters, and it typically cuts a 300-row file to about 25 rows.
- Search the text column for a handful of local words — your service areas, your product words, "how much", "price", "still looking", "no reply". Each search adds rows the question-mark filter missed.
- Add two columns by hand:
typeandreplied. Type is one of the four kinds above. Replied is whether you have answered in the thread yet. That is the whole CRM. Anything more elaborate will be abandoned by week three. - Sort by likes within the askers. A question with fourteen likes is fourteen other people waiting for the same answer, which makes your public reply worth far more than a private one.
The public reply does the selling
The counterintuitive part, and the part the installer got wrong first: answer in the thread, in public, with a real number, before you message anyone.
A public reply with a specific price range, a specific timeline and no sales language does three things a DM cannot. It answers the fourteen people who liked the question and never typed anything. It is visible to the group for as long as the post lives, which is where the installer's second and third jobs came from — people who found the thread weeks later. And it makes the private message that follows a continuation of a conversation rather than an interruption, which is the difference between a reply rate you can build on and one you cannot.
The reply that worked was three lines: the price range for the spec they described, the caveat that changes it, and "happy to quote properly if you send the room size". No brochure, no link, no "DM me for details" — a comment that withholds the answer to force a DM reads as a trap, and on a local group it gets called out.
From reply to message, without tripping Meta's rules
Only after a public reply, and only to the askers and the unserved, does a private message make sense. Meta's Messenger Platform policy is the document to read before you scale any outreach: unsolicited messaging is restricted, and the practical enforcement — restricted Pages, throttled inboxes — lands on whoever sends the most identical messages. The installer's rule was two sentences long: message only people who asked a question, and reference the question. Never paste the same message twice.
Everything else stays in the thread. The location-staters get a public yes or no. The complainers get an acknowledgement and nothing else. The file is deleted once the week's list is worked, which is also the sane answer to the privacy question: comment rows contain names, and under the GDPR that makes them personal data for EU and UK commenters. Research and reply, do not accumulate. Collecting public data is itself lawful — the Ninth Circuit's decision in hiQ v. LinkedIn settled that under the CFAA — but "lawful to read" and "mine to keep forever" are different claims.
Thirty days, anonymised
The installer's numbers, shared with permission and rounded. Twelve exports over four weeks, about 2,400 comment rows. After triage, 96 rows worth acting on: 41 askers, 12 unserved, 29 location-staters, 14 complainers.
Every asker and unserved row got a public reply. Thirty-one of those 53 people responded in the thread or in the inbox. Nine became site visits, seven became paid installs, and two of the seven came from the same nineteen-day-old group thread that started all of this — the original poster, and a second person who read the reply a fortnight later and asked whether the price still stood.
The comparison that mattered to them was not against ads in general but against their own boosted posts: the same four weeks of boosting produced more messages and fewer jobs, because a boosted post reaches people who were not asking for anything. A comment thread is the opposite — everyone in it already raised their hand.
How we know this works
We operate the exporter, so we see the shape of Facebook comment data across a lot of accounts, and three things we observe directly shaped the workflow above.
Facebook threads have the longest useful life of any platform we export. Instagram and TikTok comment sections are effectively finished within 72 hours; Facebook group posts keep collecting comments for weeks, and a large share of the files pulled on our Facebook endpoint are re-pulls of the same permalink days later. That is why the installer's second export of the same thread was worth doing, and it is why we do not cache a comment count for long.
The row count will not match the badge, and that gap is moderation rather than a failed export. On Pages with an active blocked-word list it is routine to see the badge run well ahead of the public list, which is also a warning for competitor research: the complaints you most want to read may be the ones that were hidden.
Group comments carry materially more explicit purchase intent than Page comments — more prices asked, more locations named, more "still looking". It is the one platform-specific finding we would stake a recommendation on, and it is why file three exists in this workflow at all.
The caps and column names quoted here are the live product values. Anything about Meta's own rules links to Meta's documentation rather than our summary of it.
What they tried first and abandoned
Two failures worth copying the lessons from. The first was DM-first: forty messages in a week to everyone who commented on a recommendation post, including the tagged friends. Four replies, one of them hostile, and an inbox warning. Public-reply-first fixed the reply rate more than any change to the message wording ever did.
The second was scale: exporting every post in four local groups and building a 4,000-row master sheet, which took an afternoon and was never opened again. Three files a week that get worked beat forty files that get stored. The comment file is a worklist with a shelf life, not a database.
Run it this week
Pick your own most recent post, one competitor post, and one public group thread asking for a recommendation in your service area. Export all three, run the question-mark filter, and reply in public to every question with an actual number. That is a complete first cycle and it costs nothing — the first 100 comments of any post are free, no signup; only bigger threads need a one-time payment, never a subscription, and you can see what that costs on pricing.
Download Facebook comments free →
Related reading: exporting Facebook group comments, Facebook Page comment export for brand monitoring, turning cold comment leads into hot ones on TikTok.