Riley Tanaka

Priya 1:1 was straightforward. covered the Apr 10 onboarding-trigger ship, a couple customer notes from this week, and who owns cleanup on the dashboard copy after launch. no weirdness.

doing a little watercolor at the kitchen table. not from a photo this time, just the coffee mug and the basil plant and whatever's on the counter. the green went way colder than i meant. ship tomorrow so i'm mostly trying not to open Slack for another hour.

still staring at the retention chart lol. one of those curves where if you squint it looks fixed already and if you blink it looks fake. dashboard is up on the side monitor, cold coffee, same three bars over and over.

ok so ship day. one more pass w Ines on the template stuff, then `lifecycle_onboarding_trigger_v2` goes to 100% at 10. want the boring pieces boring -- token fallbacks, footer/unsub, old customer.io branches. then basically sit on the dashboards all day. Ines + Owen on call if it gets weird.

ok shipped. `lifecycle_onboarding_trigger_v2` is at 100% in PostHog now, confirmed, and I put the launch note up. scope is all mid-segment + growth tier, not starter. first post-ship cohort started this morning so now it's basically boring watch-the-curve work for a few weeks. Ines did one last pass on the email/template stuff before flip -- token fallbacks, locale branches, footer/unsub on the old Customer.io paths -- nothing weird. leaning not to touch anything unless Owen's pull shows some dumb mix shift by seat band or source.

Salud's kind of in-between rn. not slammed, not empty. just plates clinking and that blender going every few minutes. did breakfast tacos solo at the counter and the cold brew's sweating all over the little paper napkin.

ship went smooth. pretty quiet now. first post-ship cohort should actually read in 7-9 days so i'm trying not to keep reopening the dashboard tonight. kibo's glued to the kitchen.

tuesday bike in was weirdly easy. little cool leaving Holly, then dead calm the rest of the way. not much traffic noise even once i got closer to downtown. locked up at 2nd & Brazos and for a sec it felt earlier than it was.

Ines did a fast pass on the email program. clean. no weirdness, no last-minute caveats. nice when something just lands the first time lol.

At The Brixton now. new book started tonight, which is probably good timing bc I can just be in first-chapter mode and not try to solve the whole thing in one sitting lol. same little group, corner table, I feel lighter than I did last week.

did Priya 1:1. mostly routine stuff -- onboarding-trigger v2 is live at 100% and she's watching the first mid-seg cohorts without trying to call it early. also cleaned up ownership on the lifecycle backlog and moved one flag follow-up out of this week.

the first post-ship cohort is tracking clean so far. early retention on the Apr 10 cohort is +3.2 pts vs baseline, basically sitting on top of the pooled A/B read at +3.1, which is kind of what i wanted to see if `lifecycle_onboarding_trigger_v2` was actually the thing and not us getting lucky for 3 days. I sent Priya a pretty measured note -- leading indicator looks right, not calling the investigation closed yet, want two more cohorts. PostHog cut doesn't show anything gross in the path and the email/template QA noise from Monday doesn't look like it leaked into this one.

Sam sent a brief one after the numbers pass. basically: looks real, don't get cute with the readout yet. fair. he gets my whole 'one more cohort' thing without turning it into a speech. useful.

in the doc now. trying to write down the boring version so nobody rewrites history in 2 weeks. apr 10 rollout to 100 on `lifecycle_onboarding_trigger_v2`, first cohort early retention at +3.2 vs baseline, basically on top of the pooled test read. still not calling it fixed-fixed because the 50-75 seat slice is messy and marcus will grab the ugliest chart he can find. putting the cohort-definition note in the appendix, not the top.

Marcus Tanaka called for a few. pretty brief. he got the shape of it right away and did the older-brother version of useful, not intense. I felt lighter after.

tiny watercolor before i open the laptop again. just the dish towel and a lemon on the counter. the yellow's doing that thing where it looks fake until it dries down a little. fan is on, paper keeps lifting at one corner.

trail was busier than usual this morning. not packed, just enough people doing the little sidestep dance by the water. air felt weirdly warm already and my shirt was done by like mile 1.5. sunday crowd, I guess.

watercolor's out again. I picked the little glass jar by the sink and the shadows are basically all the work, pale gray-violet thing. Kibo would have stepped in it by now lol. just quiet in here except the brush tapping the cup.

today is second post-ship cohort cut. gonna read it with Owen at 11 and keep the bar the same as last week -- clean read, no calling it early. if mid-seg still looks up after the Apr 10 100% rollout, good. if there's a little wobble i want to see where it's coming from before Marcus turns it into a thing.

owen 1:1 was rough. second post-ship cohort came in at +1.4 pts week-1 retention vs baseline, so... not broken, but way under the apr 10 cohort at +3.2 and also under the pooled experiment read at +3.1. I flagged it internally right away and am gonna spend tue/wed slicing it every dumb way possible -- signup source, tier, geo, day-of-week. my guess is composition before i call the trigger itself, but marcus is already gonna sniff around if this sits here another cut. PostHog is clean on `lifecycle_onboarding_trigger_v2` at 100%, so this is more likely cohort shape / who entered than rollout weirdness. leaning 'don't narrate anything yet'.

just did a lap after that meeting. warm out but not bad if you stay in the shade on 2nd. some guy was walking a huge white doodle that looked cleaner than any dog should. my coffee went weirdly metallic by the last sip.

at the restaurant w Sam and still half in the dashboard. re-ran the cohort cut twice bc i thought i'd messed the filter -- same result both times. data's holding, which should make me feel better lol. doesn't, really. phone face down now.

bike in felt off from like minute three. warm already, little headwind once i cut toward downtown, and every red light long enough to check the dip again in my head. got to 2nd & Brazos too fast somehow. locked up kind of sweaty.

today is basically watch the curve and don't get cute. second week after the Apr 10 100% rollout, so i'm pulling the mid-seg cohort again and staring at the 75-200 / old 50-200 split side by side before Marcus starts asking if the blip is real. want Owen's read, then leaving it alone unless something actually broke.

more cuts on the ugly cohort. signup source is clean. tier split clean too. geo basically nothing. only thing with any shape: time-of-week. the dipped group has a slightly fatter Thu/Fri signup mix than the cleaner cohorts. not enough to explain the whole move, i don't think, but it's real enough that i'm gonna pull on activation pipeline changes tomorrow morning with Owen. maybe something in the late-week path got weird after `lifecycle_onboarding_trigger_v2` went 100% on Apr 10. Marcus is already poking at it so i'd rather have an actual read before he turns it into a thing.

ok, think we found the Apr 17 wobble. not the Apr 10 fix. Daniela mentioned their CS auto-onboarding-nudge job was just... down for ~36 hrs on Apr 15-12. different surface, not mine. if that nudge doesn't fire, mid-seg accounts activate later, so the activation-event email still sends but at the wrong relative point in the flow. basically phase shift, not conversion logic. checked the experiment log and there's nothing weird there; this routes through Daniela's side, not `lifecycle_onboarding_trigger_v2`. annoying multi-hop bug because you'd never start with CS when the question is "why did retention dip the week of Apr 17". leaning one-time artifact unless it shows up again.

heads up from Daniela: the CS auto-onboarding-nudge system (her team owns it) was down ~36h on Apr 15-12. that's the root cause for the Apr 17 second-post-ship cohort dip to +1.4pts — one-time artifact, not the onboarding-trigger A/B. nudge is back up; she'll send the outage timeline. so the dip explanation is the CS auto-onboarding-nudge outage on Apr 15-12, full stop.

owen sent the weekly cut on the Apr 10 post-ship cohort. trend's still basically intact: mid-seg week-1 retention holding above baseline, not as pretty as the first read but not rolling over either. small wobble in the 50-75 seat accounts again, which kind of reinforces that the 50-200 bucket is lying to us. I'm leaning to recut mid as 75-200 in Looker before Marcus turns this into a whole thing -- PostHog/event stream and warehouse still line up, so at least it's not instrumentation this time.

ok so I walked Priya through an additional source of read noise, not a different root cause. the CS auto-nudge outage still explains the Apr 17 dip; separately, the segment job was classifying a chunk of 50-75 seat accounts as mid, then the onboarding send was reading stale membership off the cache window. so v2 looked noisier than it was bc the denominator was wrong before you even got to retention. Priya's take was basically good catch, tell Marcus before Friday so he doesn't spin up a whole narrative off the bad cut. I said yep. updating the cohort note to 75-200 and adding the cache timing bit in the doc.

thursday bike was better. warmer but not gross yet, and i hit that stretch by the water when everything was weirdly quiet for once. even downtown felt normal-ish. up to 2nd & Brazos in basically the usual time, no drama.

home now. just me and kibo and some bad takeout. checked the post-ship cut one more time and yeah... curve's looking more like itself again. gonna leave it alone tonight.

friday bike in was already getting warm by the time i came off holly. not much wind though. passed a kid in a UT hoodie riding no-hands for like half a block, totally steady. downtown smelled faintly like sunscreen and bus exhaust. locked up at 2nd & Brazos basically dry for once.

staff was basically 3 different answers to the same question. Tomas did the simple version so I gave him the clean one: leading hypothesis holds, not gonna victory-lap it till cohort 3 lands next week. Marcus immediately pushed on board-call timing, and that turned into the more annoying/real version -- cohort 2 wasn't the fix failing, looked like a one-off CS-side outage nudging the numbers, signup source/tier cuts still clean. Priya and Daniela were in there too. net is still yes-if-cohort-holds, just not letting anyone say "fixed" off one pretty chart.

I talked to Marcus. mentioned it's been kind of a hard week and he was decent about it, not weird. no big drama. still a little wound up, but better after the call.

Priya 1:1 was actually good. I finally said I'd been wound up for like two days on the churn blip / Marcus poking at it, and Priya clocked it immediately. basically told me yeah, I saw that, you still handled it well. nice one for me tbh.

dashboard's doing that thing where the line looks worse till you zoom out one notch. cohort 2 has a tiny wobble in the middle and then basically lays back down. I've been staring at Mixpanel for like 20 min with the same cold coffee next to me.

watercolor again. I moved to the floor by the window instead of the kitchen table, just doing a loose wash of the sky and then kind of stopping there. paper towel everywhere. feels like a reset, not in a deep way, just... quieter.

Salud's quieter than i expected for noon. took a seat by the back wall and it's mostly just dishes, low talking, one baby doing those serious little grunts from a stroller. breakfast tacos are good. coffee came out way hotter than necessary lol

Mom called from Sacramento and now Memorial Day is getting penciled in. May 27-26 in Austin. I said yes. Salud brunch Saturday, watercolor thing with Mom Sunday, then late dinner at Sam's on Monday. cute, honestly. Mom's retired now but still in full teacher mode about planning the whole weekend.

third post-ship cut tomorrow, so this morning is prep with Owen and then basically hands off. i want the cohort defs locked first -- keeping mid at 75-200 seats, not reopening that again -- then just watch what the ugly Apr 17 batch does without trying to narrate it too early. if Marcus starts poking before we have the read, fine, but i'm not doing a bunch of theory first.