Fugu Tech

What’s the second most popular search engine and why should you care? Super helpful guide here: fugutech.co/blog-details.php…

we spent months making onboarding shorter. the thing actually stalling it was usually one person at the client who'd never been told the project existed. you can't fix a rollout nobody announced internally with product work.

a staff phone call converts about 5%. email about 3%. everyone moved to email anyway, because $8.33 a result beats $60 — and then acted surprised when the phone team got smaller and the pipeline did too.

the cost of polish is mostly time, and time is what the thing was competing on. by the third round of notes the moment it was reacting to has passed, and you ship a very clean response to something nobody remembers.

"nobody answers their phone anymore" usually comes from people whose follow-up was one voicemail on a tuesday afternoon. we've said it too. the channel isn't dead, the effort behind it is.

most martech renewals get decided by whoever finds the dashboard useful, not by whether the number moved. that's why so many tools ship reporting before they ship the thing being reported on.

the demos where i talked least went best. i only found that out because a call ran long and i had to shut up. every discovery script i'd written before that was really a list of things for me to say.

a text gets opened about 95% of the time. email, about 20. most follow-up budget still goes to the one nobody opens, because email feels like the polite choice — and politeness isn't a channel strategy.

everybody's rebuilding the stack around ai agents. most of the wins i've actually seen come down to a message going out at 6pm instead of next tuesday. that's a scheduling fix in a very expensive hat.

disclosure gets filed under compliance. it's actually what lets an agent answer at 11pm without a fake apology for the hour — you only get to be fast and plain about it once you've said what you are.

the most expensive part of a marketing video is the part that makes it look like one. the phone-shot version from the service drive gets watched to the end because nobody has to decide whether to trust it first.

a postcard gets about a 1% response and costs roughly $100 per result. a staff member on the phone gets about 5% and costs $60. the cheapest channel to buy is almost never the cheapest one per person who actually answered.

the reason to tell someone they're talking to an agent isn't ethics, it's that it changes what they ask. people are blunt with software — they'll tell a bot the appointment time is wrong at 9pm and never bother telling a person.

a lead in a dealership doesn't go cold at day 30. it goes cold the first time it gets reassigned — every handoff restarts the relationship at zero, and the crm records that as continuous coverage.

for a year we called every no-reply a bad list. then we actually looked, and plenty of them had opened it — we'd just never sent a second message. the list wasn't cold. we stopped talking and blamed them for the silence.

direct mail returns about 1% and costs $100 a result. a phone call from your own staff, 5% at $60. email is the cheap one — 3%, $8.33 — and it's cheap precisely because almost nobody answers it. cost per send isn't cost per result.

"hyper-personalization" in practice is still a merge tag and a product someone browsed once. the model didn't unlock personalization — we've had that for years. it unlocked volume, and that's a different problem.

most dealer groups can tell you their cost per lead to the dollar and can't tell you what a second-time buyer is worth. the budget goes to strangers because strangers are easier to count.

the follow-up sequence i was proudest of was the worst one i ever ran. every touch written, every touch on brand, all of it slow. what actually got replies was a plain question sent the same day. we built a sequence when the problem was latency.

the expensive part of most marketing isn't the message, it's the disguise. the studio day, the six rounds of notes, the brand-safe rewrite. i've watched all of it test worse than the plain version it was supposed to improve.