The rules, and when I'd drop them

Launch to the people who have the problem, never to the people who follow you.

I'd drop it the day a launch to strangers brings me 3 paying customers in its first week. It hasn't happened yet.

A store is a shelf, not a shop assistant. Nobody walks you to the customer.

I'd drop it when a store listing on its own brings me a paying customer I did not send there, who stays for 3 months.

If you cannot say who will hear about it on day 2, you have a post, not a launch.

I'd drop it when a single post in a room where I am a stranger brings a customer who stays for 3 months.

A problem you had at a job you have left is a memory, and memory sands off the hard parts. Check it with somebody who has it today.

I'd drop it when a remembered problem passes the other 6 filters and sells in its first fortnight.

Write the kill date while you are hopeful. The person on day 60 cannot be trusted with it.

I'd drop it for a product that is flat but meets the test for running itself: no unplanned work for 8 weeks, under 40 minutes a week, costs covered. That one can stay, without me.

If one company's terms are your whole product, you rent it. Budget for the day the rent goes up.

I'd drop it when a product built on one platform's goodwill survives 2 changes of its terms and still grows.

Warm to cold. The people who know you go first, while the thing is new.

I'd drop it when one of my cold-first launches outsells one of my warm-first ones. None has.

Spreading on its own is something to catch, never something to wait for.

I'd drop it when a second product of mine spreads on its own. In 9 years, one has.

Being in the room is not the same as being known in it. Count the people who would notice if you stopped posting.

I'd drop it when a room where I am only a member buys from me on the first ask.

Free users prove that people like free things. Charge from the first customer.

I'd drop it when I have a way of turning free users into paying ones that works without me. I never have.

Decide what you will charge before you build, or you never will.

I'd drop it when a product I left unpriced finds its price later without losing the people using it.

Ask early whether anybody in the room has ever paid for a piece of software.

I'd drop it when a room with no history of buying software buys mine.

Free for good causes is still a product with a support queue. Decide who pays for that queue.

I'd drop it when a free product costs me under 20 minutes of support a month for a whole year.

One yes is one person's problem. A product needs the same problem in 3 people who do not know each other.

I'd drop it when a product with a single paying customer reaches 3 without me changing it.

If you cannot defend a decision to a customer, you did not make it.

I'd drop it when a product assembled from a model's suggestions beats one I decided line by line.

Price for the heaviest sensible customer, because that is who arrives when it works.

I'd drop it when a flat price on per-use costs survives 6 months of its keenest customers.

Match the shape of the payment to the shape of the value.

I'd drop it when a monthly price on twice-yearly value keeps its customers past month 3.

Never sell a feature that the platform you plug into could ship on a Tuesday.

I'd drop it when a platform leaves a gap like that open for 2 years.

Name 10 people before you build. If you cannot, you are building for a word.

I'd drop it when a product aimed at a category finds its group by itself.

A feeling is not a market. Find the people who share a task, not a mood.

I'd drop it when a product sold to a mood keeps its customers for a quarter.

In month 2, talk to the people who paid before you change anything.

I'd drop it when a second-month rewrite moves the line before a conversation does.

If it is empty until it is crowded, it is not a product for one person.

I'd drop it when I already have the crowd, before I build.

A social product is a crowd you have to bring with you. If you cannot bring it, do not build it.

I'd drop it the same as the marketplace: when the crowd exists before the product.

If you stop caring about a working product, sell it to somebody who does. Closing it is your boredom on their bill.

I'd drop it when finding a buyer for a product that small would cost more of my time than closing it, and the customers have somewhere good to go.

The peer excuse lets you build on 2 yeses. Check you are actually a peer.

I'd drop it when a product built on the peer excuse reaches 3 customers I did not already know.

Nothing free should call a model, and every key that does needs a hard cap you've watched fire.

I'd drop it when every provider I use ships a hard cap that is on by default for every new key. None of mine does.

Do the sum on 30 real inputs before you build, and keep the heaviest sensible customer's model fees under a tenth of their price.

I'd drop it if a product of mine ran at a fifth for a full year with no surprises in the bill. None has.

Read what each customer costs every week, and open the account before you act: keen gets a better price, broken gets a fix.

I'd drop it if my 3 heaviest customers on a metered product stayed under a quarter of the bill for 6 months running. They never have.

Put an allowance on anything with a meter, set at the heaviest sensible month, with a charge above it that pays.

I'd drop it if a flat price with no limit on a metered product of mine survived 6 months of its keenest customers. The only time I tried, it didn't survive 3.

Offer no free tier on anything with a meter; show saved examples, run a sample by hand, or promise the money back instead.

I'd drop it if a free tier on a model product of mine paid for its own model bill out of upgrades within 3 months. I have never run one long enough to find out, and I don't intend to.

Raise the price for new customers first, tell the existing ones in writing with the reason, and never cut the product quietly instead.

I'd drop it if a price rise for everybody on a product of mine ever kept most of them. The only one I tried kept a handful.

If you can write the right answer as a rule before you see the input, it belongs in ordinary code, and it moves there before any other saving.

I'd drop it if a model ever became cheaper and more exact than a line of code at arithmetic. I do not expect to see it.

Test the cheap model on 50 real inputs, read every disagreement, and route: cheap first, the strong one only when the check fails.

I'd drop it if the strong model's cost fell to near the cheap one's. Then I would use it for everything and stop routing.

Queue the work, check it, store it, then tell the customer; call the model live only when somebody is genuinely waiting.

I'd drop it if a customer of mine left because a result took minutes rather than seconds. None has ever said so.

Cap the attempts on every item, never retry a bad input, retry a wrong answer once at most, and read the pile every week.

I'd drop it if the pile stayed empty for 6 months running. Then I'd read it monthly.

Set a daily ceiling from the shape of the month, and at the ceiling stop, hold the work and tell yourself.

I'd drop it if my provider ever let me set a daily cap per key that held the work rather than failing it. Then the ceiling could live there.

Right, then cheap, then fast: anything about money, dates, a customer's data or words in your name goes through a check first.

I'd drop it never for money or a customer's data. For the rest, if a feature went a year without the check catching anything, I would look at it again.

Give the model labour you can check, and keep the judgement: the price, who it's for, what you say when it goes wrong, and what to switch off.

I'd drop it if a product of mine assembled from a model's decisions ever outsold one I decided line by line. The one I tried closed in 5 months.

Keep one door in the code, 3 months of stored inputs and a second option you know, and change model only after rerunning your own past work.

I'd drop it if a provider ever guaranteed a model's price and existence for longer than my products tend to live. None has.

Read the same pages every week in the same order, and write down what you find before you fix it.

I'd drop it if a product of mine went a year of Sundays with nothing new on any page. Then I would make it fortnightly, as I did for the landlord tool.

Start no second product until the newest paying one has settled into a named slot it fits.

I'd drop it if a product of mine launched into another's first 2 months ever outlived one launched after the settle. None has.

Judge each product by what it leaves behind, the list, the loop, the room and the autopsy, as well as by whether it lived.

I'd drop it if I met somebody whose first-year product kept paying for 5 years, from nothing earlier. I have met people who say so. On a closer look, there was always something earlier.

Every product gets a named slot, a new one needs a free slot, and 3 skipped slots in a row means it wins its slot back or goes.

I'd drop it if I found a product that paid for years with no slot at all. The landlord tool comes closest, at 40 minutes a fortnight.

Read cancellation replies in one sitting for the sentence that goes past the question; the same paragraph from 3 strangers is a product.

I'd drop it if a product found this way ever did worse than one I thought up from nothing. Of my 5 living products, none came from nothing.

Close a product properly, with the reason, the date, the refund and a tested export, then add one line at the end for anyone curious.

I'd drop it if the line at the end ever cost me goodwill I could see. Nobody has ever complained about it.

When a product dies, spend one evening harvesting the people, the parts, the rooms and the autopsy.

I'd drop it if one of my living products had started from nothing. None did.

If anybody paying you is asking for something you haven't built, build that before you start anything new.

I'd drop it if a feature asked for by 3 paying customers ever went unused. It has not happened to me yet.

Run 6 checks before buying anybody's product: the inbox, the price, the stack, the dependencies, the customers and the slot.

I'd drop it if I bought a product that failed 2 checks and it outlived one I built. I have not bought one to find out.

Before closing a working product, look for somebody who cares about its customers more than you do, and hand it over properly.

I'd drop it if handing over a small product ever cost me more time than closing it. I have not tried, which is the point of this chapter.

Count a product's cost in evenings first and money second; close the ones you open out of duty, and leave the slot empty until something earns it.

I'd drop it if a product I had kept out of duty for a year ever started growing on its own. The driving instructors' tool had 3 years to try.

Build every product on the same machinery, so a new one only adds the part that is new.

I'd drop it if a product genuinely could not be built on the shared pieces and was worth more than the extra slot it would take. None has been yet.

Give the small products one shared weekly page: what changed, what's waiting, and a line for every fault that once reached a customer.

I'd drop it if a small product went a whole year with nothing new on the page. Then it could move to a monthly look.

Write one page per product while you're well: where it lives, what runs on its own, what needs a person, the command that stops it, and what to tell its customers.

I'd drop it if a product of mine ever needed less than one page to run without me. None has.

Once a year, put every product on one page with its slot, its year in a line, and one word: keep, grow, hand over or close.

I'd drop it if 3 Januaries in a row produced no change at all. Then it could become every other year. So far every one has changed something.

Give each product its own costs and takings, split shared costs by a rule you can say in one line, and read them before you write anything on the January page.

I'd drop it if I ever had only one product again. Then one column is enough.

Give every idea one page (who said it, their words, where it came from, the filters, the date), and build nothing from the folder until a slot is free and somebody has asked twice.

I'd drop it if an idea I built on the day I had it ever outlived one that waited in the folder. None has yet.