You want to start a business with no ideas, and the honest truth is that you don't need an idea first. You need a problem that real people already complain about, often enough and painfully enough that they'd pay to make it go away. Freelancers, small agency owners, shop owners and solo operators post these complaints every day in Reddit threads and in the job posts they write on Upwork. This playbook shows you how to collect those complaints with permission, sort them with a short Python script, pick one micro-pain point, and test it with a one-page Carrd site and a waitlist before you build anything.
Playbook at a glance
| Item | What it means for this playbook |
|---|---|
| Who pays | A narrow group that keeps describing the same costly, repetitive problem in public: for example freelancers who chase late invoices, or small store owners who rewrite product listings by hand. |
| What you sell | A small, specific fix for one pain point: a done-for-you service, a template pack, a simple tool, or a productised service with a fixed scope. |
| Business model | Start with paid pre-orders or a fixed-price service. Move to a subscription only after people buy twice. |
| Main cost drivers | Your time, a domain name, a landing page plan, email tool, payment processing fees, and any API access you need. |
| Skills needed | Basic Python (copy, paste and edit a script), clear writing, and the nerve to message strangers politely. |
| Regulatory touchpoints | Platform terms for every site you collect data from, privacy rules for the emails you collect, business registration and tax once money comes in. |
| Difficulty | Medium. The research and page are easy. The hard parts are staying inside platform rules and turning waitlist interest into actual payments. |
Evidence of demand
Here's the blunt version. The research behind this guide contains no market size, search trend or marketplace data for any specific pain point, so demand for whatever you pick is unproven.
That's not a flaw in the plan. It is the plan. Your job in the next section is to produce your own evidence: repeated complaints from real people, replies to your outreach, waitlist sign-ups, and finally payments.
Treat those signals in that order of strength. A complaint shows a problem exists. A sign-up shows mild interest. Only money shows willingness to pay.
What you need before you start
Skills
- Enough Python to install a package, set an environment variable and run a script from a terminal.
- Plain, clear writing for a landing page and short messages.
- Patience to read raw posts yourself (the script sorts; it doesn't decide).
- Basic spreadsheet skills.
Tools and accounts
- Python 3 installed on your laptop, plus a code editor.
- A Reddit account and, if you want automated collection, an approved route to Reddit's official Data API.
- An Upwork account if you plan to read job posts there, and time to read Upwork's current Terms of Use before collecting anything.
- A Carrd account for the landing page, and an email tool or form backend that stores sign-ups.
- A spreadsheet (Google Sheets or Excel) to track pain points and outreach.
- A payment link from a payment provider available in your country, for the pre-order test.
Starting budget categories
Don't spend before you validate. Plan for these categories and price them yourself on each vendor's current pricing page:
- Domain name (optional at first; a free subdomain works for testing).
- Landing page plan, depending on whether you need custom domains and form elements.
- Email or form tool.
- Payment processor fees, charged per transaction.
- A small ad test budget, only if organic outreach stalls.
Time commitment
Pick a fixed window and protect it. A sensible planning choice is two focused sessions a week for research and one for outreach, with a hard decision date written on your calendar. Without a deadline, research turns into procrastination.
Launch it step by step
Step 1: Get permission before you collect a single post
Most guides skip this. Don't.
Reddit's User Agreement prohibits scraping its services without Reddit's prior written consent, and it allows crawling only under the parameters in its robots.txt file. So a script that downloads Reddit pages and parses the HTML is off the table unless Reddit has given you written consent.
The permitted route is the official API. Reddit's Data API terms grant a revocable, non-exclusive, non-transferable, non-sublicensable licence to use the APIs only as those terms expressly permit. Read those terms in full, register an app from your Reddit account's app preferences page, and note any limits on commercial use, storage and rate of requests. If your intended use isn't clearly allowed, ask Reddit before going further.
For Upwork, the research behind this guide doesn't establish what Upwork currently allows. Open Upwork's own Terms of Use and any API terms, and look for the sections on automated access and data use. Until you've confirmed automated collection is allowed, collect Upwork job posts by hand: read them in your browser and paste the relevant text into a spreadsheet.
- Read the Reddit User Agreement and Data API terms end to end.
- Register a Reddit app and store the client ID and secret in environment variables, never in your code.
- Read Upwork's current terms and write down, in one sentence, what you're allowed to do.
- Decide on a data rule: store only post text, link and date, never usernames or personal details.
Check: You can state in writing which collection method you'll use on each platform and the exact clause that permits it. If you can't, stop here and use manual collection only.
Step 2: Collect raw complaints from the right places
Where you look decides what you find.
Pick three to five subreddits where your likely buyers talk shop, such as communities for freelancers, ecommerce sellers, bookkeepers or small agency owners. Pick one or two Upwork job categories where clients describe recurring tasks they want done. Then collect for a fixed window.
The script below uses PRAW, the Python wrapper for Reddit's official API. It searches chosen subreddits for complaint phrases and saves the text to a CSV. Install it with pip install praw, then set your credentials as environment variables.
export REDDIT_CLIENT_ID="[YOUR_CLIENT_ID]"
export REDDIT_CLIENT_SECRET="[YOUR_CLIENT_SECRET]"
export REDDIT_USER_AGENT="pain-research/0.1 by [YOUR_REDDIT_USERNAME]"
# collect_reddit.py
# Uses Reddit's official API via PRAW. Only run this if Reddit's
# Data API terms permit your use case.
import csv
import os
import time
import praw
SUBREDDITS = ["[SUBREDDIT_1]", "[SUBREDDIT_2]", "[SUBREDDIT_3]"]
PHRASES = [
"is there a tool",
"i hate doing",
"takes forever",
"how do you handle",
"anyone else struggle",
"would pay for",
"wasting hours",
"sick of",
]
LIMIT_PER_QUERY = 50
reddit = praw.Reddit(
client_id=os.environ["REDDIT_CLIENT_ID"],
client_secret=os.environ["REDDIT_CLIENT_SECRET"],
user_agent=os.environ["REDDIT_USER_AGENT"],
)
seen = set()
with open("reddit_posts.csv", "w", newline="", encoding="utf-8") as f:
writer = csv.writer(f)
writer.writerow(["source", "subreddit", "created_utc", "title", "body", "url", "phrase"])
for sub in SUBREDDITS:
for phrase in PHRASES:
query = f'"{phrase}"'
for post in reddit.subreddit(sub).search(query, sort="new", limit=LIMIT_PER_QUERY):
if post.id in seen:
continue
seen.add(post.id)
writer.writerow([
"reddit",
sub,
int(post.created_utc),
post.title.replace("\n", " "),
(post.selftext or "").replace("\n", " "),
"https://www.reddit.com" + post.permalink,
phrase,
])
time.sleep(2) # be gentle; respect the API's rate rules
print(f"Saved {len(seen)} posts to reddit_posts.csv")
Notice what the script doesn't store: usernames. You don't need them, and leaving them out keeps you clear of a whole class of privacy headaches.
For Upwork, build a matching file by hand called upwork_posts.csv with the columns source, category, date, title, body, url. Paste the job title and the description of each post that describes a recurring task or frustration.
Check: You have a few hundred Reddit rows and a few dozen Upwork rows, all collected by a permitted method. If a subreddit returns almost nothing, swap it out rather than adding more phrases.
Step 3: Score and cluster the pain points
Now the script does the boring part. It merges both files, looks for pain and money signals, and counts which topics repeat.
# score_pains.py
import csv
import re
from collections import Counter, defaultdict
PAIN_WORDS = ["hate", "frustrat", "annoying", "takes forever", "manual",
"every week", "every month", "struggle", "nightmare", "tedious",
"waste", "sick of", "keeps happening"]
MONEY_WORDS = ["pay", "paid", "budget", "hire", "hiring", "cost", "invoice",
"client", "customer", "revenue", "fixed price", "hourly"]
TOPICS = {
"[TOPIC_A]": ["[keyword1]", "[keyword2]", "[keyword3]"],
"[TOPIC_B]": ["[keyword4]", "[keyword5]"],
"[TOPIC_C]": ["[keyword6]", "[keyword7]", "[keyword8]"],
}
def load(path):
try:
with open(path, encoding="utf-8") as f:
return list(csv.DictReader(f))
except FileNotFoundError:
return []
rows = load("reddit_posts.csv") + load("upwork_posts.csv")
scores = []
topic_hits = Counter()
topic_examples = defaultdict(list)
for r in rows:
text = (r.get("title", "") + " " + r.get("body", "")).lower()
pain = sum(1 for w in PAIN_WORDS if w in text)
money = sum(1 for w in MONEY_WORDS if w in text)
score = pain * 2 + money
matched = [t for t, kws in TOPICS.items() if any(k in text for k in kws)]
for t in matched:
topic_hits[t] += 1
if len(topic_examples[t]) < 5:
topic_examples[t].append(r.get("url", ""))
scores.append((score, r.get("source"), r.get("title", "")[:90], r.get("url", ""), ";".join(matched)))
scores.sort(reverse=True)
with open("scored_posts.csv", "w", newline="", encoding="utf-8") as f:
w = csv.writer(f)
w.writerow(["score", "source", "title", "url", "topics"])
w.writerows(scores)
print("Topic frequency:")
for t, n in topic_hits.most_common():
print(f" {t}: {n} posts")
for link in topic_examples[t]:
print(f" {link}")
Run it, then open scored_posts.csv and read the top 50 posts yourself. Fill in the TOPICS dictionary as you notice recurring themes, and re-run. Two or three passes are usually enough.
Then score your top five topics in a spreadsheet with these columns, each rated 1 to 5 by you:
- Frequency: how often the problem shows up across different people.
- Pain: how angry or stuck they sound.
- Money nearby: whether they mention clients, hiring, budgets or lost revenue.
- Reachability: can you find and message these people again?
- Fit (your skills, your interest).
Check: One topic clearly beats the rest on frequency and money nearby, and the complaints come from many different people, not one loud thread. If nothing stands out, collect from new communities for one more window before choosing.
Step 4: Turn one pain point into a narrow offer
Vague offers don't sell. Narrow ones do.
Write the offer as one sentence: I help [WHO] [GET RESULT] without [PAIN THEY MENTIONED], by [HOW]. Use the exact words people used in the posts. If they said "chasing invoices", don't write "accounts receivable optimisation".
Then choose the smallest deliverable that solves it:
- A done-for-you service with a fixed scope and turnaround.
- A template or checklist pack.
- A small tool or automation, built only after people pay.
Start with option 1 or 2. Services let you learn the problem inside out before you write code. Set a price you'd be comfortable defending, and write down why it's cheaper than the buyer's current workaround (their time, a freelancer, or doing nothing).
Check: Read your one-sentence offer to someone who fits the buyer profile. If they can repeat it back in their own words and ask "how much?", move on. If they look confused, narrow it again.
Step 5: Build the Carrd landing page and waitlist
One page. One promise. One form.
- In Carrd, start from a blank or minimal template so nothing distracts from the copy.
- Add a headline built from the buyer's own words, a subheadline with the result, and three bullets on what they get.
- Add a form element with an email field and one short question, such as "What's the hardest part of [PAIN] for you right now?" That answer is research gold.
- Connect the form to where sign-ups should land (Carrd's form settings list the supported destinations). Confirm on Carrd's pricing page which plan includes forms and custom domains before you pay.
- Add a clear line about how you'll use the email address, and a link to a simple privacy note.
- Add a second call to action: "Want it sooner? Reserve a founding spot," linked to a payment link for a small refundable pre-order or deposit.
- Publish, submit the form yourself to test it, and confirm the entry arrives.
That second button matters most. A waitlist sign-up is an initial interest signal, not proof that anyone will pay. The pre-order link turns interest into a real test.
Check: The page loads on mobile, the form delivers entries, the payment link works end to end with a test purchase you refund, and a stranger can explain the offer after reading for ten seconds.
Step 6: Get the first sign-ups and paying customers
Don't spam the communities you researched. That's the fastest way to get banned and burn the very audience you need.
- Read each community's rules about self-promotion before posting anything.
- Where allowed, write a helpful post that answers the common question in full, and mention your page once at the end.
- Reply only to threads where your offer is directly relevant, and lead with an actual answer.
- Message people in your own network who fit the buyer profile, using the scripts below.
- If you have an Upwork freelancer profile and the offer fits, respond to relevant job posts through Upwork's normal proposal process, inside its rules.
- Email every sign-up within a day, ask one question about their situation, and offer a short call.
- On each call, describe the founding offer and ask directly for the pre-order or deposit.
Track every visitor source, sign-up, reply, call and payment in your spreadsheet. Set your go/no-go thresholds before you start, in writing, so you can't move the goalposts later.
Check: By your decision date, the number of paid pre-orders or deposits meets the threshold you wrote down. Sign-ups with zero payments is a no-go on this offer: rework the price, the audience or the promise, then test again.
Step 7: Deliver to the first customers by hand
Deliver manually. Every step you do by hand teaches you what to automate later.
- Write a delivery checklist before the first order and follow it every time.
- Log how long each task takes and where you got stuck.
- Ask each customer three things after delivery: what nearly stopped them buying, what they'd change, and who else has this problem.
- Request permission before quoting any feedback publicly. Never edit a quote to sound better.
Check: Customers get the promised result on time, and at least some ask for more or refer someone. If refunds or complaints cluster around one part of the offer, fix it before taking more orders.
Step 8: Systemise what works
Only now do you build.
- Turn your delivery checklist into a standard operating procedure document.
- Schedule your research scripts to run on a regular cadence, still within the API terms, so you spot new complaints early.
- Replace the most repetitive manual delivery step with a template, form or small script.
- Move the waitlist page into a proper sales page with the real price and a checkout.
- Decide whether to raise prices, add a recurring plan, or test the next pain point from your list.
Check: You can deliver an order by following the written procedure without improvising, and your contribution margin in the worksheet below is positive after fees. If it isn't, change price or scope before scaling.
Unit economics worksheet
Fill this in with your own numbers. No benchmarks are given because none are verified for this kind of offer, so your figures from the pre-order test are the only ones that count.
| Line | Variable | Formula |
|---|---|---|
| Price per order | P | Your listed price |
| Your hours per order | H | From your delivery log |
| Value of your hour | R | What you'd accept per hour elsewhere |
| Direct cost per order | C | Tools, contractors and materials used for this one order |
| Payment fees per order | F | P multiplied by your processor's percentage, plus any fixed fee from its pricing page |
| Acquisition cost per customer | A | Total ad or outreach spend divided by number of paying customers |
| Contribution margin per order | M | M = P minus C minus F minus A |
| Margin after your time | M2 | M2 = M minus (H multiplied by R) |
| Monthly fixed costs | X | Landing page plan, email tool, domain, software subscriptions |
| Break-even orders per month | B | B = X divided by M |
| Waitlist to paid conversion | W | Paid pre-orders divided by waitlist sign-ups, from your own test |
If M is negative, no amount of volume fixes it. Raise P, cut C, or find a cheaper acquisition channel first.
Scripts and templates
Research reply (only where community rules allow)
Hi [NAME OR "there"],
I saw your post about [SPECIFIC PROBLEM IN THEIR WORDS]. I'm researching
this exact problem for [WHO, e.g. freelance designers].
Quick question, no pitch: how do you handle it today, and what's the most
annoying part?
Thanks either way.
Network outreach message
Hi [NAME],
I'm testing a small service for [WHO] who struggle with [PAIN].
It would [RESULT] in [TURNAROUND] for a fixed price.
Does this sound like something you or someone you know deals with?
If yes, here's the page: [LANDING_PAGE_URL]
Happy to hear "not for me" too. That helps.
Waitlist follow-up email
Subject: Thanks for joining the [PRODUCT_NAME] list
Hi [FIRST_NAME],
Thanks for signing up. You mentioned: "[THEIR FORM ANSWER]".
I'm opening [NUMBER] founding spots at [FOUNDING_PRICE] before the
public launch. Founding customers get [BONUS] and a full refund if
[GUARANTEE CONDITION].
Reserve a spot here: [PAYMENT_LINK]
Or reply with one question and I'll answer personally.
[YOUR NAME]
Offer page outline
HEADLINE: [Result] for [who], without [pain in their words]
SUBHEAD: [How it works in one sentence]
PROBLEM: 2-3 lines quoting the pain the way buyers describe it
WHAT YOU GET:
- [Deliverable 1]
- [Deliverable 2]
- [Deliverable 3]
HOW IT WORKS: Step 1 [..] / Step 2 [..] / Step 3 [..]
WHO IT'S FOR / NOT FOR: [..] / [..]
FORM: Email + "What's the hardest part of [PAIN] for you?"
SECOND CTA: Reserve a founding spot [PAYMENT_LINK]
PRIVACY LINE: We only use your email to [PURPOSE]. Unsubscribe anytime.
Pricing section copy
FOUNDING OFFER: [FOUNDING_PRICE] (regular price later: [REGULAR_PRICE])
Includes:
- [Deliverable and scope]
- [Turnaround time]
- [Number of revisions]
Guarantee: [Clear refund condition]
Spots: [NUMBER] at this price
[BUTTON: Reserve my spot]
Legal and admin checklist
- Platform terms: Reddit's User Agreement bars scraping without Reddit's prior written consent. Use the API only as the Reddit Data API terms expressly permit. Check Upwork's current Terms of Use yourself before any automated collection.
- Registration and licences (US): The SBA says businesses may need registration, tax IDs, and federal, state, county, or city licences and permits depending on location and activity.
- Annual return (US): The IRS says self-employed individuals generally must file an annual income tax return and pay estimated taxes quarterly.
- Filing threshold (US): Per the IRS Self-Employed Individuals Tax Center, you generally must file a return if your net earnings from self-employment are $400 or more.
- Estimated tax (US): The IRS estimated taxes page says individuals, including sole proprietors, generally must make estimated tax payments if they expect to owe $1,000 or more when filing their return.
- Outside the US: No rules for other countries are covered here. In India, for example, check the Income Tax Department and the GST portal, and your state or city authority for any trade licence, before taking payments.
- Privacy: You'll be collecting email addresses. Check the data protection law where you and your sign-ups live, and say clearly on the page how you'll use the emails.
- Refunds and pre-orders: Write your refund terms on the page before taking a deposit, and check your payment provider's rules on pre-orders.
Why this fails and how to avoid it
You scrape first and read the terms later
Warning sign: your script fetches web pages with requests and parses HTML. Fix: switch to the official API under its terms, or collect by hand. Your research isn't worth an account ban or a legal letter.
You confuse loud with common
Warning sign: your top topic comes mostly from one viral thread. Fix: count distinct posts and distinct communities, not total comments.
You pick a pain nobody pays to fix
Warning sign: complaints never mention clients, hiring, budgets or lost income. Fix: weight the "money nearby" score heavily and favour problems tied to how people earn.
You treat sign-ups as validation
Warning sign: the waitlist grows but nobody clicks the pre-order link. Fix: make the payment ask explicit on calls and in email. If people still won't pay, the offer or price is wrong.
You build the tool too early
Warning sign: you're writing app code before anyone has paid. Fix: deliver by hand first and automate only the steps your delivery log shows are repetitive.
You get banned from your best channel
Warning sign: moderators remove your posts or warn you. Fix: stop promoting, re-read the community rules, and contribute useful answers without links for a while.
You never decide
Warning sign: you keep "collecting a bit more data". Fix: the decision date and thresholds you wrote down in Step 6 are final. Go, rework or drop.
Hand this to ChatGPT or Claude
You can paste the brief below, or this page's link, into an AI assistant along with your city, budget and skills to get a plan tailored to you. Fill in the brackets first.
GOAL: Help me find one profitable micro-pain point and validate it
with paying customers before I build anything.
CUSTOMER: [Who I think the buyer is, e.g. freelancers, small shop owners]
My city/country: [CITY, COUNTRY]
My budget: [BUDGET]
My skills: [SKILLS]
Hours per week: [HOURS]
OFFER: Not chosen yet. Help me write a one-sentence offer from the
pain point I pick, using the buyer's own words.
VALIDATION STEPS:
1. Collect complaints only through permitted methods: Reddit's official
Data API within its terms, and manual collection from Upwork unless its
current terms allow automation.
2. Score posts for pain and money signals with a Python script; cluster
by topic; read the top posts manually.
3. Rate top topics on frequency, pain, money nearby, reachability, fit.
4. Build a one-page Carrd site with an email form, one research question
and a refundable pre-order link.
LAUNCH STEPS:
1. Outreach that follows each community's rules plus my own network.
2. Follow up every sign-up and ask for the pre-order on a short call.
3. Deliver manually with a checklist; log time per task.
4. Automate only proven repetitive steps.
NUMBERS TO FILL IN: price, hours per order, direct cost per order,
payment fees, acquisition cost, monthly fixed costs, contribution margin,
break-even orders, waitlist-to-paid conversion, my go/no-go threshold
and decision date.
CONSTRAINTS: No scraping without permission. No storing usernames or
personal data. Check tax, registration and privacy rules for my country.
Don't invent statistics; tell me where to verify anything you're unsure of.
Frequently Asked Questions
Can I really start a business with no ideas at all?
Yes, if you swap "find an idea" for "find a repeated, costly complaint". The research steps above give you a short list of real problems, and the landing page test tells you which one people will pay to solve.
Is it legal to scrape Reddit for business research?
Reddit's User Agreement prohibits scraping without its prior written consent. Use the official API within its Data API terms, or read and note posts by hand.
Can I scrape Upwork job posts?
This guide doesn't confirm what Upwork currently allows. Read Upwork's own Terms of Use before any automated collection, and default to manual collection until you're sure.
How many waitlist sign-ups mean the idea is validated?
None on their own. Sign-ups show interest; payments show demand. Set your own threshold of paid pre-orders before you launch the page, and judge the test against that.
Do I need to know Python well?
No. You need to install a package, set environment variables, edit lists of words and run two scripts. Reading the posts and choosing the pain point is the real skill.
Why Carrd instead of building a website?
A single page is all a test needs, and Carrd lets you publish one quickly without code. Check Carrd's pricing page for which plan includes forms and custom domains.
What if no pain point stands out?
Change where you look, not how hard you look. Try different communities or job categories for one more collection window, then decide.
Your first step
Today, open the Reddit Data API terms and Upwork's Terms of Use and write one sentence for each describing what you're allowed to collect and how. Then list five communities where your likely buyers complain about their work. That's your starting line.