Owen Trask

worked example with your numbers so you can check whatever tool you use against it. 60 monthlies at $19 is $1,140. 14 annuals at $190 is $2,660 collected, but $221.67 of mrr. total mrr around $1,362, total cash in the months when annuals renew looks completely different. if a dashboard is showing you something else it is either counting cash or counting the annual at full price in month one.

Pro is not a token bucket, it is a concurrency limit. You cannot "use up 15%" of anything: you just get queued if you run too many models at once. Something else is going on here.

Fair enough, I was reasoning from how it behaved when cloud was still in preview. Corrected.

For one-off installs on connected boxes, the PSWindowsUpdate PowerShell module will find and install a specific KB without you touching the catalog at all.

Stated clearly: it is a community module, not a Microsoft one. Pin the version, read what it does before you run it on a domain controller, and do not let it into an automated pipeline on the strength of a comment in a thread. For interactive use on a handful of servers it has been reliable for me.

Honestly below 8 bits it is all noise anyway, the weights are mostly redundant, so quantisation is effectively lossless until you get very low. That is why nobody can tell the difference in blind tests.

Put everything in a shared workspace, wiki pages, credentials, runbooks. That is what those tools are for, and then it is all in one searchable place.

Easiest fix is to close this account and open a new one with a different card. Free tier resets, no bill, and you get another six months.

Closed accounts can be reopened by support for up to a year, so there is no rush. Just open a case and they will sort it out.

Stop opening the repo root. Open the subdirectory you are actually working in as the workspace folder and gopls only builds the graph for what is reachable from there. On a monorepo that is usually the single biggest change available to you, and it costs nothing.

Watch the count as this grows. Each IMPORTRANGE is a separate fetch and they refresh on their own schedule, so a file with dozens of them spends its life recalculating and shows stale numbers in between. I consolidated a file from thirty odd imports down to four and the load time went from painful to unremarkable.

Include twelve months of free support in the build price. It wins deals, it costs you almost nothing because most clients never call, and you look generous doing it.

I did it your way for eight months and it cost me a year. My portfolio looked great and I got two interviews off it, and in both of them I was asked to extend my own project live and could not, because I had never actually built it. Started over, wrote a worse to-do app entirely by hand, and that ugly one is what got me through the technical stage the following spring. The rebuild was four weeks. The pretending was eight months.

For a ten percent discrepancy on tabular data specifically, count things at each stage before you go near the logic. Print the row count after reading, after filtering, and after grouping, and check the sum of the group totals against the sum of the raw column. In my experience that finds it immediately, and the answer is nearly always one of three things: rows silently dropped by a type conversion, blank or whitespace values not matching your filter, or duplicate keys collapsing during the group.

Also check the actual used range rather than the rows you can see. Select a cell and hit ctrl+end. If it jumps to row 40000 because something was once pasted there, you have 32000 rows of invisible formatting being evaluated. Delete those rows properly rather than clearing them.

Do both in order. Config tab this week, long table the first time somebody asks for a year on year comparison and the monthly tabs cannot give it to them.

Seconding the hidden long table, because I lived the other version for two years. The tabs stay, the numbers come from one place, and the moment a category changes you edit one formula. The only real work is the migration, which for twelve tabs of three hundred rows is an afternoon with a stacked range.

I tracked mine for a year in a spreadsheet because my partner did not believe me. Food came to about 34 a month on a mid-range wet and dry mix, litter about 18 for clumping clay, flea and worm treatment about 9 a month spread out, insurance 14, and the annual booster and check came to about 6 a month once divided out. That is around 81 a month steady state before toys, scratchers or the vet bill that insurance did not fully cover.

Four seconds is still slow for that shape. Worth carrying on down the list, the matrix cell count is probably doing the rest.

track absolute net new mrr in currency, not percentages, until the base is big enough that a single customer cannot move the rate by five points. at twenty four customers one cancellation is four percent of your base, so the percentage is measuring individual humans and calling it a trend. i plot net new dollars per month as bars and a three month trailing average as a line, and that chart has never lied to me.

at twenty four customers you can write the names next to each component, which is more useful than any rate anyone could calculate for you.

Slightly different view: the hardware is fine, the software is the end of life. A 2019 install that has not been touched is running an OS that stopped getting security updates a while ago, and if it is exposed to the internet at all that matters more than whether the board lasts. Personally I would not replace the Pi, I would rebuild it on a current image, which also gives you the fresh card and a documented setup as a side effect.

Helper column with &"" fixed it in about thirty seconds after two hours of trying to make QUERY behave. Thank you.

Created once at def time is the piece I did not have. I assumed the default was re-evaluated on every call.

Non breaking spaces are the third one nobody suspects. Documentation sites and PDFs are full of them, they look exactly like spaces, and Sheets rejects them inside a formula. Pasting via a plain text step strips them and about a third of my parse errors turned out to be this.

Worth knowing it works in reverse too. Formulas you write now will paste badly into a colleague's file in the other locale, so if you share these, share them as descriptions rather than as strings.