By MD Fozla Rabbi, SEO Specialist
Every Monday I used to open Meta Business Suite and stare at a follower graph. Up a bit. Down a bit. Then I'd close the tab and forget what I saw.
That's not a report. That's a mood.
So I wrote a small Python script instead. It reads the CSV you export from Facebook Page Insights and prints a plain weekly summary: new follows, unfollows, net growth, and the best day of the week. No pandas. No API keys. Just the standard library.
I built it for a few small Indian pages I help with. A sweet shop in Pune. A coaching centre in Jaipur. A guy who sells handmade lamps. They don't want dashboards. They want one paragraph on WhatsApp every Monday.
Let's build it.
What the export looks like
In Meta Business Suite go to Insights, pick your date range and export the followers data as CSV. Column names shift a little between accounts and languages. Mine looked roughly like this:
Date,Follows,Unfollows,Page followers 2026-11-01,14,3,2210 2026-11-02,9,1,2218 2026-11-03,22,4,2236 Yours might say "New follows" or "Lifetime Total Followers". That's fine. We'll match columns loosely instead of hard-coding names.
Step 1: read the CSV without guessing column names
import csv from datetime import date, datetime from collections import defaultdict ALIASES = { "date": ["date", "day"], "follows": ["follows", "new follows", "new followers", "page follows"], "unfollows": ["unfollows", "page unfollows", "unfollowers"], "total": ["page followers", "lifetime total followers", "total followers"], } def find_col(header, key): lowered = {h.strip().lower(): h for h in header} for alias in ALIASES[key]: if alias in lowered: return lowered[alias] return None def to_int(value): value = (value or "").replace(",", "").strip() return int(float(value)) if value else 0 def load_rows(path): with open(path, newline="", encoding="utf-8-sig") as f: reader = csv.DictReader(f) cols = {k: find_col(reader.fieldnames, k) for k in ALIASES} if not cols["date"]: raise ValueError("No date column found") rows = [] for r in reader: raw = r[cols["date"]].strip() if not raw: continue day = datetime.strptime(raw[:10], "%Y-%m-%d").date() rows.append({ "day": day, "follows": to_int(r.get(cols["follows"])) if cols["follows"] else 0, "unfollows": to_int(r.get(cols["unfollows"])) if cols["unfollows"] else 0, "total": to_int(r.get(cols["total"])) if cols["total"] else None, }) return sorted(rows, key=lambda x: x["day"]) Two small things matter here. The utf-8-sig encoding strips the invisible BOM that Excel loves to add. And to_int handles Indian-style numbers with commas, like 1,25,000, because people often open the file in Excel and save it again. It happens more than you'd think.
Step 2: group days into weeks
Indian clients think in weeks that start on Monday. Python's isocalendar() does exactly that.
def by_week(rows): weeks = defaultdict(list) for r in rows: year, week, _ = r["day"].isocalendar() weeks[(year, week)].append(r) return dict(sorted(weeks.items())) Each key is a (year, week) pair. So week 47 of 2026 never gets mixed up with week 47 of 2025. Small detail. Saves a headache in January.
Step 3: summarise one week
WEEKDAYS = ["Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun"] def summarise(week_rows): follows = sum(r["follows"] for r in week_rows) unfollows = sum(r["unfollows"] for r in week_rows) net = follows - unfollows best = max(week_rows, key=lambda r: r["follows"] - r["unfollows"]) churn = (unfollows / follows * 100) if follows else 0.0 return { "start": week_rows[0]["day"], "end": week_rows[-1]["day"], "follows": follows, "unfollows": unfollows, "net": net, "churn": round(churn, 1), "best_day": WEEKDAYS[best["day"].weekday()], "end_total": week_rows[-1]["total"], } The number I care about most is churn. It's unfollows as a share of new follows. If a page gains 100 followers and loses 40 the same week, the follower graph still goes up. But something's off. Maybe the posts got repetitive. Maybe a giveaway pulled in people who never cared.
Churn under 20% is usually healthy for small local pages. Above 35% and I start asking questions.
Step 4: compare with last week
A number alone means little. "Net +48" sounds great until you learn last week was +120.
def pct_change(now, before): if before == 0: return None return round((now - before) / abs(before) * 100, 1) def weekly_report(path): weeks = list(by_week(load_rows(path)).values()) summaries = [summarise(w) for w in weeks] lines = [] for prev, cur in zip([None] + summaries[:-1], summaries): change = pct_change(cur["net"], prev["net"]) if prev else None trend = f"{change:+}% vs last week" if change is not None else "first week" lines.append( f"{cur['start']:%d %b} to {cur['end']:%d %b}: " f"+{cur['follows']} / -{cur['unfollows']} " f"(net {cur['net']:+}, churn {cur['churn']}%), " f"best day {cur['best_day']}, {trend}" ) return "\n".join(lines) if __name__ == "__main__": import sys print(weekly_report(sys.argv[1])) Run it like this:
python fb_weekly.py followers_export.csv And you get something like:
27 Oct to 02 Nov: +61 / -14 (net +47, churn 23.0%), best day Sun, first week 03 Nov to 09 Nov: +88 / -19 (net +69, churn 21.6%), best day Fri, +46.8% vs last week 10 Nov to 16 Nov: +54 / -27 (net +27, churn 50.0%), best day Wed, -60.9% vs last week Look at that third week. Churn jumped to 50%. When I checked, the sweet shop had run a "tag 3 friends" contest. Lots of follows. Lots of quick exits too. The script didn't fix anything. But it made the problem impossible to miss.
Step 5: make it WhatsApp-friendly
My clients read this on their phones. So the last step just trims the latest week into one friendly message.
def whatsapp_message(path, page_name): weeks = list(by_week(load_rows(path)).values()) cur = summarise(weeks[-1]) msg = ( f"*{page_name} weekly update*\n" f"New followers: {cur['follows']}\n" f"Unfollows: {cur['unfollows']}\n" f"Net growth: {cur['net']:+}\n" f"Best day: {cur['best_day']}" ) if cur["churn"] > 35: msg += "\nNote: lots of unfollows this week. Let's check what we posted." return msg Copy. Paste. Done. Takes ten seconds on a Monday morning.
What the numbers usually tell you
After running this on a handful of Indian pages for a couple of months, a few patterns kept showing up.
Best day is rarely the day you post most. One coaching centre posted daily but grew most on Sundays. Turns out parents scroll on weekends. So we moved the big posts there.
Reels move follows, but not always good ones. A Reel spike often came with higher churn the following week. Fine for reach. Less fine if you want buyers.
Consistency beats bursts. Pages that posted three solid times a week had steadier net growth than pages that posted ten times one week and zero the next.
None of that is new advice really. But seeing it in your own numbers lands differently. If you want a full list of non-technical tactics to pair with this script, this guide on 15 free ways to grow Facebook followers in India covers page setup, posting times, Reels and Stories in more depth.
Ideas to extend it
A few things I've added for some clients. Maybe useful for you too:
- Save history. Append each week's summary to a
history.csvso you can chart months later. - Flag dead weeks. If net growth is zero or negative two weeks in a row, print a warning at the top.
- Run it on a schedule. A cron job on a cheap VPS can read the latest export from a shared folder every Monday at 9 AM IST.
- Handle several pages. Loop over a folder of CSVs and build one combined message for agencies that manage many pages.
I also keep a small notes page on social growth tools for Indian creators at smmyodha.com if you want more context on where follower numbers come from and what affects them.
Wrapping up
This is maybe 80 lines of Python. It won't impress anyone on GitHub. But it turned a vague Monday habit into a clear one-paragraph update that clients actually read.
And honestly, that's most of what reporting is. Not fancy charts. Just the right three numbers, every week, without fail.
If you try it on your own page export and the column names don't match, drop the header row in the comments. I'll help you add the alias.

