
Co-founder @upsidegtm , co-founder @branchmetrics | investor @xfactor, Female Founder Retreat organizer, reader of lots of fantasy and sci-fi.
San Francisco, CA
Joined April 2008
- Tweets2.6K
- Following718
- Followers4.6K
- Likes12.9K
Pinned Tweet
I spent 4,000+ agent hours doing 1,763 analysis of 521 B2B new business deals across 13 companies to understand what works, what really drove pipeline and how much of those touchpoints were actually buried in unstructured data.
I mean this is no joke. At @hf0 which is technically an inversor, they gave us @WHOOP s and we could all see eachother's sleep scores. And I know it was meant to make sure we don't burn out but I'm pretty sure we all felt judged if we slept 8 hrs a night 🤣
Mada Seghete retweeted
As part as of our Due diligence with VCs I will provide my Oura ring sleep data.
Using Upside, I anonymized and aggregated the data behind 29,367 new-business deals across multiple b2b enterprise companies to try to understand: just how messy GTM data really is?
This is an extreme oversimplification. I mean maybe for PLG - large enterprise deals are complex and most of what matters is hidden in unstructured data, duplicate record, missing roles etc - an agent can't do all that deduplication and make sense of it when an account has sometimes tens of thousand of touchpoints (that's how much it takes to close a 5M+ deal).
upside.tech/research/the-unt…
I also think attribution data becoming extra important - because it's what agents need to use to build GTM self learning loops, so the importance is increasing, not decrasing.
In 2026, B2B GTM attribution tools are officially COOKED 🤯 .
It's been a good run.
But now, it's RIP 🪦 .
The dream of custom software for everybody, materializing from imagination to tangibility is actual reality.
Buying off the shelf SAAS for tracking my customer acquisition funnel might cost $20 a seat minimum.
Two weeks ago, I simply told opus 5.5, to "make me more money".
Pointed it at my 3 sources of data (Grafana, simple analytics, and database).
Within a day, the software of my dreams was present, and perfect.
And it only costed $5 in tokens.
Hence the economics of SAAS just don't make since anymore.
If I was a SAAS builder today, I would count the LTV of my customers on a 1 time purchase basis, cause even if I offer something unique, my customers could just copy me verbatim with one prompt.
Not sure about that. @Dan_Ahmadi hold offsites for his agents where they clarify ownership of tasks etc and it's been productive - they keep hiring more agents.
South Park in SF:
Ever bench is an AI startup 1:1 or VC pitch meeting all while basking in the glowing lights of A truck ad.
Not pictured: the heavy smell of exhaust fumes of mentioned truck running in high making breathing around it almost unbearable.
This is super cool! Excited to test it out :) Go @ExaAILabs team!
Their API is one of the best for cold email lists, specifically for finding decision makers for companies that were scraped from random directories.
You can scrape approx 10k company urls for 7$. (1000 searches x 10 results)
Since I love scraping what nobody is scraping, I often end up with company names and their website and every enrichment fails, running gpt or claude web searches is costly af so I end up using Exa and/or spider.cloud
Basically to find a decision maker, you literally give your agent that manages all your apis, in my case claude code, to use exa and do a google search of "who owns COMPANY name" or "who is the decision maker at COMPANY"
In my case i got 78% hit rate which is brilliant considering that it is real cheap.
Alternatively you can use gemini api, it works real well because when it performs a google search it can read that first ai answer you get when performing a normal google search and it can pull data from there asap.
Gemini get really fucking expensive when you scale it tho.
There are 2 types of marketing teams.
The ones that credit the demo form and call it "marketing leads", and the ones that try to understand whatever made the buyer fill it out, even if it was a sales convo 6 months ago.
One of our own deals started with a demo request that said "we need Upside". Our AI gave that form 0%. It gave 47% to internal championing and 28% to a relationship from years earlier.
More on what's actually driving pipeline:
upside.tech/research/the-unt…
Me after leaving the first startup I founded after 10 years:
"Never again will I merge my identity with my company!!!"
Me this week:
"I'm dressing up @UpsideGTM for Halloween!"
😅😂🤣
Today we're launching Dollars 💰
Most AI startups spend so much time and effort building an AI product with negative gross margins. We skipped straight to the end.
We sell US dollar bills for $0.80.
TAM: $1T+ in circulation. We’re going after all of it.
Let’s scale to the moon together.
dollars for less dot com
For those who are worried AI will take our jobs, don't fret. Silicon Valley is inventing new jobs for humans every day.
(picture take by @Dan_Ahmadi in South Park today)
The downside of having an unusual or lacking sense of humor is the constant feeling of needing to explain your jokes.
I dressed up at @UpsideGTM data layer at the @ExaAILabs Halloween Party 🤣🤣
I don't think that's right. That might show you how someone first heard about you - but that does not account for what really tipped the scales, the complex buying group, etc. I think the form needs to be a signal in your attribution (and we do include it) but can not be the only thing you look at. Below is just one example of an AEO analysis I did for a customer where you can clearly see other sources of signal that do not show up in the self attributed signal.
More on what matters: upside.tech/research/the-unt…
You do not need a $50K attribution platform to figure out what drives your revenue.
Early in my career, we spent 6 months arguing over 10 different attribution dashboards that nobody trusted. Add a self-reported attribution field on your demo form, listen to 5 customer calls, and read the deal notes in HubSpot. You will learn more in 1 afternoon.
Hahaha 100%. Also they don't work that well... same with marketing email. It makes up 45% of the touches a customer gets but only counts for 3% of the reason someone decides to consider buying.
😂
so done with building my startup
i've worked 150 hour weeks
recorded 300 podcast episodes
larped as a founder at over 20 coffee shops in sf
followed 10 VCs home from the office
and I still have zero paying users
and haven't raised any capital
already been at it for well over 3 weeks
when does it get easier?
I dressed up at @UpsideGTM data layer at the @ExaAILabs Halloween Party 🤣🤣
I made an app showing how @UpsideGTM is cleaning our own GTM data. Turns out @Dan_Ahmadi
and I are pretty bad at sales hygiene 😆😅😆 - I have been building these for all our customers and we are by far the worst so far.

























