Before picking a niche, look up ten businesses in it and note who they pay for help now. Someone stuck with a mediocre provider already believes the problem matters.
It is very easy to build for an imaginary user who loves the clever technical bits as much as you do. Pendo found 80% of features in the average product are rarely or never used.
Lenient return policies raise purchases more than they raise returns, across a meta-analysis of 21 papers. A retainer's notice period works the same way, so month-to-month with 30 days' notice makes the yes cheaper.
When 103 intensive care units adopted a short set of routine steps for putting in central lines, bloodstream infections fell by up to 66%. None of the steps told a doctor how to treat anyone. That split is what I would copy for a productised service.
Before each customer call, write down what you expect them to say, then compare it with what they actually said. Once you're getting it all right, add a second person.
Say 12 of your 30 clients are local service businesses at 35% gross margin and the other 18 average 20%. Those 12 are 40% of the list and about 54% of gross profit. Your niche might already be sitting in the invoices, just not on the website.
Bain found 85% of B2B buyers purchase from their "day one" list, the vendors they already had in mind before they started searching. Most cold outreach reaches people after that list is written.
Amazon writes the press release before building the product. A small team can get away with one sentence, what a happy user would tell a friend, and then check every roadmap item against it.
A quick niche test is whether you can name fifty businesses that fit. "Fitness brands" is too vague to list. "Independent gyms with one location trying to open a second" is a list you could start tonight, and a named business is one you can find and write to.
I run a creator agency and I build backends. Having both meant I could build our team's tools without briefing a developer first. Your second skill is probably whatever you keep handing off.
From 1980 to 2012, maths-heavy US jobs with little social interaction shrank by 3.3 points as a share of the workforce. Jobs needing high levels of both maths and social skill saw especially strong growth in jobs and pay, in David Deming's research.
Every few months I reread old failure write-ups. When the same lesson has shown up three times, it goes on a checklist, so nobody has to learn it a fourth time.
On a team of about 100, the processes that hold up usually started as someone doing the job by hand, badly, many times. The ones designed on a whiteboard first tend to get rebuilt.
What are you avoiding this week because it might get a no? Early on, every no tells you where the real business is, so the task you keep dodging is usually the useful one.
Ask five people on your team to finish "last week, our customer was annoyed by ___" without comparing notes. Five different answers means five segments wearing one name.
Nearly everyone in a Richard Wiseman experiment noticed big dots flashing at the edge of a screen. Offer a large reward for watching the centre dot and over a third missed them. This month's target does the same to a founder.
"If I make a mistake on our team, it is not held against me."
Google used that line to measure psychological safety, the top factor in its study of 180 teams. Worth asking yours.