Notes from building SiteOps, written plainly
23notes so far
- You are probably under-emailing your listThe draft has been ready for nine days and you still have not sent it, because you do not want to annoy anyone. Here is the arithmetic on how often to email, and the signals that actually mean you are sending too much.
- You can see their ads, not their resultsFourteen tabs of competitor research, and not one number in any of them. A live ad tells you it exists, never what it costs or returns. Here is what competitor research is actually good for, and what to use instead.
- Too small to A/B testEvery growth article says to split your traffic and test. At a few hundred visitors a week that test cannot resolve. Here is the arithmetic, and the three things to do instead.
- One channel until it worksBeing on five channels at once feels like sensible hedging. In practice it means none of them ever gets a real attempt. Here is the arithmetic for running one properly, and how to know when to add the second.
- Your inbox is the content calendarMost content calendars start as empty boxes you then have to fill. The questions people already ask you are a better source: real demand, real phrasing, and a first draft you have written once already.
- Two kinds of zeroA row reads zero for six weeks, so you cut the channel. But a real zero and an unrecorded one print the same character, and only one of them is a result. Here is how to tell them apart before you decide.
- Ask them how they found youYour analytics says the visit was direct. The person says a friend sent them a screenshot. One optional question on the signup form gets you the half of the story no link tag was ever able to record.
- The draft that waits for youAutomation nobody reviews has to be cautious, because everything it makes goes straight to real people. Automation that cannot reach anyone without your yes is free to be wrong, and that changes how much you let it try.
- The welcome email everyone opensThe message that goes out the minute somebody signs up reaches a reader who chose you seconds ago. That is the most attention you will ever have, and most small products spend it on nothing at all.
- One goal, one numberEvery campaign reports a wall of numbers, all of them true and none of them a verdict. Watch five and one is always up, so you can never be wrong — and never learn. Pick the one number before you launch.
- Marketing on thirty minutes a dayThirty minutes a day is two and a half hours a week, and most of it disappears into looking rather than making. A weekly routine that survives a bad month gives each day one job and cuts the rest.
- When your ad and your page disagreeSomeone taps your ad and lands on your page. For about a second they are checking one thing only: am I in the right place? If the ad and the page answer that differently, the click is spent and the visit is over.
- Reading a dip without panickingTraffic drops eighteen percent and you have a theory within a minute. The theory is probably wrong. Here is how to tell an ordinary wobble from a real break before you rewrite a page that was fine.
- UTM tags you can still read in six monthsOpen a report from a campaign you ran six months ago and count how many rows are the same campaign spelled differently. Link tags fail quietly, and the fix is a naming scheme small enough to keep by hand.
- When to kill a campaign, and when it just needs a weekTwo hundred dollars in, two signups out. Is that a losing campaign or one that has not said anything yet? Most kill decisions get made on a number too small to mean anything in either direction. Here is how to tell them apart.
- Ten thousand views and nobody cameA short video collects ten thousand views and the site records ninety visits that day. Neither number is lying to you. They measure three different things, and only one of them ever decides whether a channel is worth your time.
- What actually moves your search positionsYour average position can fall in a week where you did everything right, and rise in a week where you did nothing at all. Here is what genuinely moves rankings month to month, and what to ignore all week long.
- The unsubscribe you should be glad aboutHow often should you email your list? Everyone wants a number back, and the number is not what decides whether a list survives. What burns a list is sending when you have nothing for the reader.
- What the first $100 of ad spend can tell youA hundred dollars buys roughly sixty-seven clicks. That is enough to price the auction and judge whether your ad earns attention, and nowhere near enough to judge whether your product converts.
- Every campaign launches pausedAn engine that can spend money automatically is one bug from spending it wrong, continuously, while you sleep. So every ad campaign SiteOps builds goes live paused, spending nothing until a human flips the switch.
- The claim linter: catching numbers your AI made upAn AI copywriter reaches for confident specifics — cut your workload in half, join thousands of teams — and invents the proof. A claim linter reads each draft, flags any assertion the system cannot back, and refuses to ship it.
- When to skip the send: a scheduler that can say noA scheduled email that fires into a broken landing page is worse than no email. We made 'skip' a first-class outcome, not a failure — here is the guard list a scheduler runs before it lets a send go out.
- One ops room instead of nine dashboardsGA4, Amplitude, Search Console, Bing, a backlink tool, a payment dashboard, YouTube Studio, Instagram insights, two ad managers. None of them answer the only morning question that matters: is this product OK today? The fix is not another tool that replaces them. It is a room that reads them.