Riley Tanaka

bike in was nice. more wind than i wanted once i got out of Holly, not enough to be dramatic, just that steady crosswind where every light makes you notice your shirt sticking a little. downtown looked weirdly washed out. got to 2nd & Brazos basically on time.

Salud's pretty mellow rn. got there right between the stroller wave and the line, so it's just forks, the espresso machine, one guy reading something on his iPad like he's at jury duty. tacos are solid. coffee's too hot still.

just did a tiny watercolor from the windowsill shadows. mostly blue-gray and that one ochre that's somehow always too much. paper dried with a hard line right through the middle and i left it. apartment's so still i could hear the rinse water slosh when i set the jar down.

starting the postmortem doc for EXP-2023-02-pricing-page now. rough shape is hypothesis -> design -> instrumentation -> result -> confounders. hypothesis is basically: clearer packaging + less visual clutter on pricing would lift started checkouts / visitor and paid conversion, mostly self-serve. design section needs the exact PostHog exposure rule for `pricing_page_2023q2_rev` and the 50% split from Mar 7 to rollback. instrumentation, i want spelled out enough that nobody does the annoying "maybe tracking broke" thing later -- pageview, flag exposure, signup start, checkout start, paid conversion, plus the segment cuts Daniela asked for in the CSV. result is simple, down 1.8 pts, done. and yeah i'm putting the Mar 2 signup-form layout change in confounders since it overlapped the window enough to muddy clean read-through.

I sounded mostly fine, just tired. Marcus was apparently kind of a pill about the pricing-page thing even though I did exactly what you're supposed to do there -- saw conversion off, killed the PostHog flag, move on. I'm calmer about it than he is. weirdly the onboarding test is finally starting to show something, so not all bad.

at Salud now. window's open and every table has that half-weekend look, people not in a rush yet. did the tacos, gonna do a long lap after. no laptop. just coffee and Kibo under the chair for once not trying to inspect everyone's bag.

I seem ok. little fried maybe, but in the normal way. I called the pricing-page rollback pretty clean, no drama, and now the onboarding one is finally showing some lift so I'm a bit more like myself. not asking for anything, just clocking it

I checked the dashboard this morning and EXP-2023-02-onboarding-trigger finally crossed. week-1 retention is up 2.9 pts in treatment, p=0.05, n=220 on that arm. pretty clean read, especially after the `pricing_page_2023q2_rev` mess. I'm still not turning it on yet though -- want one more cohort week before rollout, just to see it hold and not do the fake-first-significance thing. I think that's right. basically making it a standing rule now: first sig doesn't ship, wait a stabilization cohort.

onboarding cut still looks right. not victory-lapping it yet but the Mar 13+ starts are holding the same shape in PostHog and Owen's warehouse pull, which is usually where fake lift dies. pricing rev was a clean miss and got yanked; this one keeps surviving more cuts. my read seems good.

Marcus DM'd after seeing the onboarding-trigger early cut. I told him the same thing I've been saying internally: looks real, material enough to matter, but I'm waiting one stabilization cohort before calling it. He was fine with that. No extra drama in the thread, which honestly helps.

hot out. did a lap after and took Congress the long way so I wouldn't end up back at 2nd and Brazos too fast. whole downtown has that 3pm glare right now, everybody squinting, scooters everywhere. bought a topo chico and kept walking.

bike in was fine. headwind in little bursts, enough to make every block feel longer than it was. got to 2nd & Brazos with half the rollback postmortem already rearranged in my head and then immediately forgot one of the good lines locking up.

postmortem went out to Marcus Vail, Priya Devarajan, Daniela Schultz, Tomas Vega. main thing in there is the pricing rev never really got a clean read -- Mar 2 signup-form layout change was live at the same time, so EXP-2023-02-pricing-page couldn't isolate the pricing variable. I put it pretty plainly, no name-pointing, just design review note -> concurrent change -> confounded result. attached the full flag/segment cut again and the PostHog exposure logic for `pricing_page_2023q2_rev`. Marcus is gonna read between the lines anyway lol

put one hard rule in the experiment notes: if pricing is under test, no other layout/UI changes on that page, full stop -- attribution got muddy because of the Mar 2 signup-form change sitting in the same window as EXP-2023-02-pricing-page / `pricing_page_2023q2_rev`. if somebody asks later why some cleanup can't ship during a pricing run, that's the reason. content ok, copy ok, but don't move boxes around while we're trying to read conversion.

postmortem's in. Marcus wrote back "thanks for the doc" which... ok, i'll take it. still a little static there but lower. Sam's closing tonight so it's just me, leftovers, and not opening Slack again.

today is mostly keeping both lanes clean. pricing-page rev stays off -- PostHog flag `pricing_page_2023q2_rev` should still be 0% and I want the rollback numbers tightened before Marcus turns this into a whole thing. then onboarding-trigger A/B. checking the next cohort cut before I say lift out loud. Owen for the query if he's free, otherwise I'll just pull it myself.

Owen ran the second cohort cut on EXP-2023-02-onboarding-trigger. treatment is +3.4 pts week-1 retention on that cohort, n=200. pooled now +3.1 pts, p=0.04, n=420. looks real enough that i'm past the "maybe noise" phase. still checking the boring stuff before I ask for rollout. want sample-ratio mismatch, signup-source skew, and any weirdness from the pricing rollback window ruled out -- even if this test is scope-disjoint from `pricing_page_2023q2_rev`. PostHog assignment looks stable at first pass. leaning hold one beat longer for a guardrail read on activation completion + support touches, then ramp. Marcus is gonna want the win today lol but i'd rather not eat a fake lift because one cohort got lucky.

bike in was pretty uneventful. warmer earlier than i expected and the air by Holly had that dusty-pollen thing going on. hit downtown in about the usual 22 and there were like six people all doing the slow garage-ramp merge at 2nd & Brazos.

checked with Ines. email QA is basically through the pass list now -- links/rendering/personalization all cleared, just one Outlook spacing issue left on the variant template. she dropped notes in the thread and said send is still on the same timing if that fix lands today.

owen's weekly cut looked basically right. pricing-page rev is still a miss even after the rollback noise got stripped out -- no surprise there, that one never recovered. onboarding-trigger A/B is the opposite: holding the lift week over week, not huge but real. i think my read is correct to leave it running one more cohort before I touch the flag. less drama than Marcus wanted, more useful.

friday Marcus Tanaka call was fine. he's flying into Austin next weekend, Apr 1-29, first time down here this year. I've already got the little plan -- South Congress, Veracruz tacos, then dinner at Sam's. I sounded kind of relieved having something normal on the calendar

Marcus Tanaka's in Austin today. I did lunch at Veracruz on South Congress with him, then we walked around a while after. I feel way more relaxed than I have in weeks, honestly. needed that. Sam's joining us for dinner too -- kitchen-table seats at the restaurant.

day's done. Sam's still at the restaurant so it's just me and the dishwasher for a minute. apt's pretty still. i moved my laptop off the table and that somehow made it feel like the night started.

bike in was quiet. a little cooler at the start in Holly, then basically fine once i got moving. hit a bunch of green lights coming toward downtown for once. locked up at 2nd & Brazos and my left cuff has chain grease on it again.

Marcus Tanaka's here through today so I did brunch with him this morning before his flight out. good timing honestly. I've been pretty locked in all month and felt lighter after. then a little watercolor tonight, apparently -- very sunday.

this morning is pretty much the second-cohort confirmation cut on EXP-2023-02-onboarding-trigger. want Owen's pull, then i'm checking the boring stuff one more time so nobody can nitpick it later -- sample ratio, signup-source mix, any weird segment wobble. if the +2.9 on week-1 retention is still there and not drifting, gonna call the ship.

ok so EXP-2023-02-onboarding-trigger cleared on the second cohort too. pooled week-1 retention is +3.1 pts treatment vs control, p=0.04, n=420 across the two cohorts. I called it. still doing the boring checks with Owen's pull -- SRM, signup-source mix, any weird country/device skew -- but unless something dumb pops up we're drafting the ship plan with Ines now and putting full rollout at 100% on Apr 10. nice clean one after the pricing-page mess lol

I had my 1:1 with Priya and shared the second-cohort confirmation on the onboarding-trigger thing. Priya was basically good work, tell Marcus tomorrow. I'm still being careful about it -- said cautiously optimistic, want one more cohort post-ship. first time I've used 'optimistic' since feb, which... yeah. good.

Sam got home late and we were both pretty cooked. I just said, "looks like the email change works" and that was basically the whole victory lap lol. very me. quiet apartment, decent mood.

bike in was mostly just legs today. kind of gray out, not cold exactly. Congress had that month-end traffic thing where nobody seems fully awake but everybody's inching somewhere. got to 2nd & Brazos a little sweaty, chain sounded mildly annoyed the last few blocks.

today is basically set up the watch. onboarding-trigger fix goes 100% on Apr 10 so i just want the post-ship cohort curve clean and boring, esp mid-seg. gonna keep the cut tight and not call anything early -- if there's lift i want a full week before i say it out loud.

this morning is Ines + one more pass on the email template QA before the Apr 10 ship. want the boring stuff clean -- token fallbacks, locale weirdness, unsubscribe/footer on the old customer.io branches. also Priya kept the Q1 activation target at 60% into Q2 as-is in our 1:1, so i'm not re-cutting that deck number yet. if the sub-50 seat tier really ships in Q3 then we'll move it to 65% in the small-business-tier planning.

staring at the retention chart and doing the dumb squint test before i touch the cohort cuts. mid-seg still looks a little too fat bc 50-75 is sitting in there and behaving weird. dashboard also decided every shade of blue should be the same blue today, which is helpful lol

Salud's got that late-brunch lull rn. couple half-finished plates, nobody turning tables fast. did breakfast tacos and the coffee's fine, which honestly feels rare. sitting by the window just watching people drift past with dogs and bike helmets.

final QA pass looked clean. Ines walked me through the whole path in staging -- create workspace, hit onboarding step 2, wait out the delay, email fires once, correct variant, no duplicate on backfill/re-open, and the suppression rule still catches converted accounts. trigger logic finally not doing anything weird. we're still gonna click through a few ugly edge cases before Monday -- manual invite, owner swap, timezone drift, the usual.

owen's weekly cohort cut still points the same way. onboarding-trigger lift is holding in the fresh cohorts, not huge but enough that i'd treat it as real for now. i'd still keep the guardrails on until the apr 10 full rollout -- mostly because the 50-75 seat band keeps muddying the read and marcus will overreact to one ugly chart. pricing-page rev still dead.

Ines did a quick pass on the email program and it's clean. nice to have one thing just be clean lol.

Marcus Tanaka called me for a bit this afternoon. short one. I mentioned next week's ship and he was doing the older-brother thing without making it a whole thing. I felt steadier after, which... good.

home. weirdly quiet. kibo did two laps like i was gone for a month, now he's fully passed out by the kitchen. curve's starting to look right on the pre-ship stuff. not calling it yet.

bike in was calm. overcast but bright enough, and the lake bit smelled kind of wet and metallic. a guy on cesar chavez had a coffee balanced in one hand the entire time somehow. rolled up to 2nd & Brazos right at the usual.