Run the arithmetic on the toddler scenario, because it is the one everyone panics about. To reach attempt 13 you have to sit through 1 + 5 + 15 + 30 + 90 minutes and then 4, 12 and 24 hours of enforced waiting. That is more than a day and a half of deliberate, spaced-out attempts.
A child in a pram or a pocket touch cannot get there in one session, and repeated identical entries collapse into one failure anyway. The realistic victim is a person who genuinely forgot the PIN and keeps coming back to it over two days, which is exactly the person least able to afford a wipe.
Two or three a month with 30-hour attacks is roughly three days a month out of your life, and that's the framing that finally got me taken seriously: at work and in a consulting room.
When I said "I get bad migraines" nothing happened. When I said "I lose 30 to 36 hours, two to three times a month, which is about 10% of my waking year," my GP referred me and my manager stopped suggesting paracetamol. Same facts, countable version.
Grade blind, and grade the next day. Strip model names from the transcripts and diffs, give the files neutral IDs, shuffle them, then score.
My hit rate for guessing which model produced which diff, on models I was sure I could tell apart, was barely better than a coin flip once the names were gone. That result alone changed how much I trust my own impressions from a normal working day.
And put that eval in CI with the model name as config. Then the next price change is a one-line diff and a green or red run, instead of a fortnight of arguing about whether it feels worse.
Practical policy that ended this argument for me: one metric, one source, written down. Anything that touches money or a decision comes from a SQL query saved in the repo with a comment explaining the definition. Everything else can come from the tool.
It sounds bureaucratic for a one person company and it takes about an hour to set up, but it means you stop rediscovering the same discrepancy every quarter. Also exclude your own accounts and your test emails in the query itself rather than remembering to filter, because you will forget.
Founder-led, and it isn't close, because self-serve at 12 hours a week means four months building a funnel for traffic you don't have.
The thing people get wrong is treating "it doesn't scale" as a reason not to do it. At ten customers you don't need it to scale. You need to know why anyone says yes, and there is no way to learn that from a signup form. Six calls will teach you more than six weeks of analytics on 40 visitors.
12 hours is plenty for this. Eight calls a week is about five hours including scheduling and notes. The other seven go on building the two things those calls told you were missing.
The switch happens when you've done the same manual onboarding step for the fifth time and you're bored of it. That's the trigger to automate that one step. Not before.
The reason the posts contradict each other is that there are two layers and people only ever describe one of them.
Layer one is the general data protection rules, where direct marketing is explicitly acknowledged as something that can rest on legitimate interest rather than consent. That is the layer the 'it's fine' posts are talking about.
Layer two is the electronic communications rules, which are implemented separately by each country and specifically govern unsolicited commercial email. This is where it diverges. Germany is strict enough in practice that commercial email to a business generally needs prior consent. The UK, by contrast, exempts corporate subscribers, so B2B mail to a company address is permitted with identification and an opt-out. Several other member states have their own B2B carve-outs with their own wording.
So the honest answer is 'mostly yes, with real exceptions, and it depends where they are'. Practical baseline regardless: identify yourself and the company clearly, say why you are contacting that specific person, include a working opt-out in every message, honour deletion requests promptly, and record where each address came from.
It isn't dead, it's that a templated pitch to a stranger always converted terribly and now it's measurable. And $39/mo through cold ads is a great way to spend $600 discovering your CAC is $310.
96 opens and 3 replies means they read it and did not care. That's a positioning problem, not a deliverability problem, and it's the easier of the two to fix.
Your email tells them what your software does. They already have booking software: that's literally your qualifier. So the email lands as "here is another one of the thing you already have." Nobody books a call for that.
Replace the product line with a symptom line. Something like: "you're on X for booking, are you still moving no-shows into the calendar by hand, or did you find a way around that?" You're not selling, you're asking about a specific annoyance you know exists in that stack. It costs them nothing to answer and it's about them, so replies go up a lot.
Also: at $39/mo, 15 minutes of a clinic owner's time is worth more than your first three months of revenue. Don't ask for a call in email one. Ask a question. Book the call in email three.
Email the teacher and ask which working gets full marks. Two lines, no complaint in it, and you will get a clear answer plus goodwill. Most of these standoffs are two adults guessing at each other.
Keep a file with each error message and a one line note about what caused it. After thirty entries you have your own index and you stop searching for them.
I added google sign-in for a support reason, not a security one, and I'd do it again on that basis alone.
At about 700 users, password resets were 6 of every 10 support emails. Adding one social provider took a weekend and dropped that to almost nothing within two months, because the people who forget passwords are the same people who never had one with google. That's the actual argument: it removes a whole support category. It does not make your session handling safer, since you still issue a session afterwards exactly as you do now.
Strongly agree on placement. I wrote fourteen articles in a fortnight and deflected nothing, because people in trouble do not go looking for a knowledge base, they reply to the last email they got from you. Moved three of those articles into the actual error states and the ticket count moved that week.
I ran both approaches across two terms of the same kind of material. Term one: cards from the reading, huge deck, high review load, mediocre exam. Term two: cards only from problems I got wrong plus a small set of definitions I kept confusing, deck about a fifth the size, better exam and far less time. The small deck also stayed alive because it never felt like a debt.
From the shop side, the number breaks down as a bench fee to look at it, then labour by the hour or by a flat rate for common jobs, then parts at retail. A carb clean and service is a routine flat rate job everywhere I have worked. The jobs where I tell homeowners to buy new are anything involving valves, a scored bore, or a bent crank from hitting something solid, because the labour alone passes the price of a new machine.
0.4% free-to-paid with no sales touch is roughly normal, honestly. The 2-5% people quote usually comes with a real activation flow and an email sequence behind it. The conversion number alone isn't the emergency.
Five of eight churning in 60 days is the emergency. Email those five. Not a survey, an actual email from you with one question: "what were you trying to get done the week you signed up?" You'll get two or three replies and they'll tell you more than the whole funnel dashboard.
Completely expected, and the software is doing exactly what it should.
The scheduler aims for a target retention, often around 90 percent. That means one card in ten is supposed to fail at the moment it comes due. If nothing ever failed, your intervals would be too short and you would be spending your life reviewing things you already know. A card that survived to a nine month interval and then died has bought you eight months of not thinking about it.
What would worry me is a pattern: same card, dies at every long interval, twelve times running. That is a leech and the card is badly written. One-off blanks are the price of long intervals.
The content of Foundations is the smallest of any module you will take that year. What changes is that nobody cares whether you got the answer any more - they care whether every line follows from the one above it, and that is a skill you have never been asked for. Expect three weeks of feeling stupid and one week where it clicks.
On whether it's still alive: send one more, in a completely different shape. No follow-up energy, no circling back.
"Hi X, I'm assuming this isn't a priority right now and that's completely fine. I'll stop following up. If I've got that wrong, reply with a date and I'll work around it."
Break-up emails get replies at an embarrassing rate, 30-40% in my experience, and half of those are a clean no, which is also worth having because it closes the loop in your head and you stop spending attention on it.
A design partner who pays nothing is not a design partner, they're a very enthusiastic focus group. That's the lesson and it's an expensive one, sorry.
Enthusiasm on a weekly call is free. It costs an ops lead nothing to describe their dream feature to two people who will go and build it, and it makes the call pleasant for everyone. The signal you needed was never "they're engaged", it was "they signed something": a paid pilot, a letter of intent with a date, a budget line, anything with a name on it.
Next time, charge for the pilot even if it's $500, and make the commercial ask before you build. "If this does X by June 15, is there budget for six seats?" If the answer is "I'd have to check with finance": that's the same answer you got on May 2, just three months earlier and for free.