Tweets about growth, experimentation, and marketing. Sharing the journey of building @figpii. Recovering software developer turned marketer ages ago.
Chicago, IL
Joined November 2007
- Tweets5K
- Following318
- Followers2.4K
- Likes15K
Pinned Tweet
25 mistakes companies make with A/B testing:
1. Buying software before knowing what they want to learn. Okay, you can run a test now. What are we testing, and why?
2. Starting with “what should we change?” before asking “what’s stopping people from buying?”
3. Installing heatmaps and calling that optimization. You have a picture of where people click. What decision will you make with it?
4. Watching recordings without a question. You have 5,000 videos. Nobody has time for that. Start with a specific page, behavior, or visitor group.
5. Watching someone leave and deciding you know why. Maybe it’s price. Maybe they don’t understand the sizing. Ask them.
6. Asking visitors ten questions when one useful question would do: “What’s stopping you from purchasing today?”
7. Collecting survey responses and doing nothing with them. If people keep asking about fit, what are you going to test on the product page?
8. Assuming everyone who visits is ready to buy. Someone comparing options needs different information from someone looking for the checkout button.
9. Testing the page somebody on the team finds annoying. Where’s the evidence that customers have a problem with it?
10. Copying a winning test from another company. You copied the design. You didn’t copy their customers or the reason it worked.
11. Saying “we need more social proof” as if there’s only one way to do it. Customer reviews, media mentions, and number of customers are different ideas. Visitors can respond differently to each.
12. Changing five unrelated things in one variation. If it wins, good. But can you explain what to build on next?
13. Assuming the prettier version will win. Every designer thinks the new version is better. Your visitors are allowed to disagree.
14. Measuring button clicks when you actually want orders. Getting more clicks is useful only if it helps people move toward the outcome you care about.
15. Looking only at visitor counts when planning a test. How many people actually complete the action you’re trying to improve?
16. Splitting limited traffic across six variations. Now you have six things to wait for.
17. Promising a result by Friday before figuring out how much data the test needs.
18. Stopping the moment the report turns green. Decide how you’ll evaluate the test before you get attached to a result.
19. Skipping the purchase flow during QA. The headline looks great. Can somebody still select a size, add to cart, and pay on their phone?
20. Changing the offer halfway through and forgetting about it when reading the results. Keep a record of what else changed.
21. Looking at a losing variation’s final number and moving on. Where did behavior change? What would help explain it?
22. Running the next test without using anything you learned from the last one. You’re paying to answer the same questions again.
23. Leaving winning variations running through the testing tool indefinitely. At some point, somebody needs to implement them on the website.
24. Shopping for more features when the team barely uses the ones it already has. What work will that extra feature actually help you do?
25. Thinking the software will tell you why customers aren’t buying. It can help you collect evidence and test an answer. You still have to ask a useful question.
Before you launch your next test, finish this sentence:
“We believe visitors aren’t buying because ______, and here’s the evidence.”
“This version looks better” doesn’t fill that blank.
The easiest way to sound smart on X in 2026 is to declare an industry dead and then spend the rest of the day using products from that industry.
One of my business rules:
Have the awkward conversation today so you don’t have a much worse one 3 months from now.
If I think your site doesn’t have enough volume for us to move quickly, I’d rather tell you before you hire us
What’s something your team changed your mind about?
Founder Twitter talks endlessly about leaders teaching teams.
The reverse happens constantly too.
Same reason I don’t ask customers, “What should we change on the website?”
Ask what nearly stopped them from buying.
What confused them.
What they had to go looking for.
Customers are very good at telling you where it hurts. Then it's your job to prescribe and execute the fix, not ask them to prescribe it for you.
xitter.cf/Mike_Scully_/status/21…
I don’t ask clients what they want automated. They don’t know.
I ask them to walk me through their week. Then I listen for three things.
One. Anything they do the exact same way every time, like onboarding, invoices, reporting or follow-ups. Repetition is money.
Two. Anything where a person copies info from one place to another. Emails into the CRM, PDFs into a spreadsheet. That’s pure gold.
Three. Anything they say “we should really sort that out” about. They’ve already decided it’s worth fixing, you just have to offer.
Then I do the maths out loud.
“You said someone spends an hour a day on this. That’s 5 hours a week, about 260 hours a year. That’s over 6 weeks of someone’s wages spent copying and pasting.”
Now the price makes sense before I’ve even said a number.
That one conversation is usually where the $5k automation is hiding.
I’m not selling n8n or Zapier workflows. I’m selling those 6 weeks back.
The tools are the easy part. Understanding the problem and showing what it costs to leave it is what gets you paid.
The main fix is the maths. An hour a day is about 260 hours a year, which is closer to six and a half working weeks than “a full month”. Someone in the replies will check, so it’s better to have the right number. It’s also bigger, which makes the point stronger.
Improving a website is not complicated:
1. Open your analytics. Find the pages getting plenty of visitors but sending very few people to the next step.
2. Watch 20–30 recordings from people who left there. Look for the same hesitation, backtracking, missed button, confusing section, etc.
3. Put a short poll on that page. Ask something simple like, “What’s stopping you from continuing today?”
4. Check support tickets and sales calls. See whether customers ask the same question or raise the same objection.
5. If the same problem shows up in 2–3 places, fix that first.
6. If you have enough traffic, A/B test the change. If you don’t, ship it and watch what happens.
Then repeat.
The tools are the easy part. The hard part is doing this consistently instead of changing the site every time somebody has a new idea.
This is basically the process we’ve used with clients for years, and a big part of why we built FigPii
“Everyone will build their own software” makes a lot more sense when everyone you know is on X.
Most businesses don’t want to build software.
They want payroll to run, orders to ship, leads followed up, and reports ready on Monday morning.
If AI lets them do that with fewer vendors, great.
But most people are buying the job getting done, not a new hobby.
xitter.cf/levelsio/status/210926…
Any growth advice that ends with “this will 100x your revenue in 6 months” deserves a small discount 😂
UGC and paid can absolutely work.
But before pouring money into them, I’d still want to know:
Do people want the product?
Do they understand the offer?
Do they stay after buying?
Otherwise, you can scale the wrong thing very efficiently.
xitter.cf/VadimStrizheus/status/…
the only thing you need to be thinking about as a founder is marketing
- ignore other founders trying to sell you their SaaS
- stop trying to do SEO + AEO
all you have to focus on right now is mastering UGC content and paid ads...
everything else can come after, but in the moment right now focus only on those 2 things.
this will 100x your revenue within the next 6 months.
stop overcomplicating everything and just do the work.
Software spent 20 years bragging about being “sticky.”
Alex Karp’s response: “Sticky is a metaphor for parasitic.”
That’s a pretty brutal way to describe SaaS lock-in.
But there’s something to it.
A customer staying because their data is trapped, the integrations are painful to unwind, or switching would take six months is very different from a customer staying because the product keeps earning its place.
AI weakens the first kind of moat.
Which leaves a much less comfortable question:
If your customers could leave tomorrow with almost no switching cost, would they still choose you?
xitter.cf/jawwwn_/status/2108336…
Palantir CEO Alex Karp explains why your business is essentially fucked if it's not sovereign in the AI era:
“What is a business? Why can you charge more than it costs you to produce? Because you have some unique way of doing something.”
“You need to advance your unique capability, learn from the world, and never, ever expose to anyone else how you’re doing it exactly.”
“The minute someone else can do any part of those things as well as you can, your pricing parity goes down, and at some point, your value goes down.”
"Normal software is built to force you to adopt something that's parasitic. They take every one of those things, turn it into a commodity, and sell it to everyone else."
"You talk to VCs, what they wanted to hear was, 'Clients will be unhappy, but they can't get rid of you.'"
"Sticky is a metaphor for parasitic."
"I get your data, I get it into something you can't get rid of. The thing turns into something I can sell to everybody else, who then can't get rid of it, and de facto, your multiple goes down and my multiple goes up, and all of us are happy, and hope at some point you realize you're sad... Hopefully after it's too sticky."
Via @ZetaGlobal @dsteinberg10000
One of the hardest parts of building FigPii has been saying no.
Almost every feature request sounds reasonable on its own.
Then you combine 20 reasonable requests, and suddenly you’ve built a completely different product.
The question we keep coming back to is pretty simple: does this help the customer do the main thing they hired FigPii for better?
If not, “no” is usually the better product decision.
xitter.cf/MediaKing/status/21081…
If a company is constantly launching new features, that's often a red flag.
Strong product-market fit usually looks like one core behavior loop.
Then ruthless focus on reducing friction, fixing bugs, adding users, and putting out fires as scale breaks things.
More features often means a weaker product.
I agree with the first sentence. I’m less convinced by “stay 5–6 steps ahead.”
If software can be copied that quickly, being six features ahead mostly buys you a little time.
The harder things to copy are the boring ones:
> Customers who trust you
> Distribution you’ve built over years
> Integrations people depend on
> Knowing the problem better than the next guy
You still have to keep building. I just wouldn’t confuse shipping faster with having a moat.
xitter.cf/skirano/status/2108041…
The funny thing is AI made building 10x easier. It didn’t make getting customers 10x easier.
One engineer can now ship web, desktop, iOS and Android. Great.
But the customer still has 50 other products asking for attention.
So yes, the bar for what you can build went way up.
The bar for giving someone a reason to choose you went up with it.
xitter.cf/yongfook/status/210800…
Most teams say they start with the problem.
In reality, they start with the solution they already want:
“We need a redesign.”
“We need AI.”
“We need a new pricing page.”
Then everyone spends six weeks building it.
Before touching anything, I’d do three boring things:
> Watch 20 people use the current version.
> Read the support tickets.
> Talk to 5 customers.
You usually come back wanting to build something very different.
xitter.cf/matt_teeixeira/status/…
$20/month looks expensive until you have to maintain the replacement.
Building version one is getting easier.
Then come permissions, edge cases, integrations, billing, security, uptime and the random thing that breaks six months later.
The SaaS subscription was never just paying for the code.
xitter.cf/kylegawley/status/2107…
There are 3 types of people vibe coding alternatives to SaaS products:
1. Nerds with too much time and no life
2. Developers with no money to pay for SaaS
3. People getting paid to say this shit
Everyone else is looking for ways to save time, not sink more of it into pointless projects to save $20/mo.
AI made software cheaper to build. It did not make switching free.
That’s the part of the “legacy SaaS is dead” argument I don’t buy.
You can rebuild a feature pretty quickly now. Fine.
But a product that’s been inside a company for 8 years has data, integrations, permissions, workflows, reporting, and people who already know how to use it.
Replacing the software is one thing. Replacing all of that is another.
Legacy SaaS absolutely has to adapt.
But easy to build and easy to displace are two very different things.
xitter.cf/JesseTinsley/status/21…
We’ve seen the same thing.
A normal post was getting a few thousand impressions.
Then, almost overnight, we started seeing 100-200 impressions.
I’m usually the first person to blame the post before blaming the algorithm.
This one feels different.
xitter.cf/jakezward/status/21071…
Yep, something VERY weird is happening with LinkedIn right now.
One big profile last week:
Monday post: 274,000 impressions
Tuesday post: 1,000,000 impressions
Wednesday post: 9 impressions
Thursday post: 76 impressions
We’ve spent the last week talking with some of the biggest LinkedIn creators + looking at what’s happening across Kleo.so users.
Every single one of them has seen their impressions drop massively, especially in the last week.
Also LinkedIn itself has been glitchy af:
- Posts disappearing from profiles
- Slow loading
- Strange feed behaviour
- Crazy swings in distribution
LinkedIn seems to be making some big changes to how the feed distributes content and things are pretty unstable right now.
Either way, I wouldn’t overreact. It's happened before and it'll probably happen again.
I’m watching this closely across our network of big profiles + Kleo users, so I’ll report back once we know what’s actually changed.
This is the part of platform shifts I find interesting.
For a while, everybody gets very good at acquiring traffic again.
Then six months later, the conversation becomes, “Okay, why aren’t these people buying?”
New distribution creates a lot of opportunity. It also exposes a lot of bad websites.
xitter.cf/Seanfrank/status/21072…
Why I think it is a GREAT time to be in ecommerce:
- new platform shift
legacy to agentic
anytime there is a platform shift (retail to online, online to amazon) there is the chance for new brands to make it
- meta on the offensive
when meta ships, we win
reels was the biggest boon to my business since stories
totally undid all the damage of ios14
now they have another breakout app, but this time it has credit cards and purchase data
and it is 100% theirs
golden era of digital ads coming
- new SEO, but this time, tied to spend
LLMs like things that are popular.
meta will like people who spend with them
established brands are going to have a leg up of they play their cards like
a brand with 10 million customers will just be more likely to be suggested, and if meta is taking a cut of sales (directly or indirectly) they will be sending people to the greatest hits
---
its a beautiful new day my friends.
show up and build something.
If I had 5 paying customers and needed 15 more in 30 days, I would not spend the month “building an audience.”
I’d talk to the 5 people who already paid.
Why did they buy?
What problem were they trying to solve?
Where did they come from?
Then I’d go find more people who look like them and sell manually.
You can build the machine later.
First prove you know who actually wants the thing.
xitter.cf/jessievbreugel/status/…
your SaaS has:
- 5 paying users
- $49/mo
- no audience
- no repeatable acquisition
but you need 15 more customers in the next 30 days.
what's your 1st move?
#2 is where I’d spend my time.
Most specialized services have a lot of work that looks like expertise but is actually process.
The interesting part is separating that from the 20–30% where you really need someone who's seen the problem a few hundred times.
Automate the first part. Don’t pretend you automated the second.
xitter.cf/yongfook/status/210696…
How to make money as an indiehacker in 2026 and beyond? the opportunities seem to be shrinking fast.
I've been thinking about this a lot lately. AI is eating whole categories of the SaaS industry. What would I work on if I was starting from scratch?
Here are some random thoughts.
1) Software primitives
Some of the basics aren't going away. Humans use them to get a job done, or humans use AI to interact with the tool. The AI needs the tool and can't natively do what the tool does. Newsletter managers/deliverers, form builders, analytics software. If the category has existed for 20 years, it will continue to exist. The question here is can you get traction for a product in an extremely saturated space. Maybe there are opportunities in niches.
2) Automate your specialized service
This is probably the best time ever to productize a specialized service. Take a domain / service you're an expert in, which previously would require unscalable 1-to-1 attention with a client, and agentify it with proprietary data/skill.md etc.
3) AI value arbitrage
There is a short term opportunity in taking things we see at the cutting edge of the X bubble, and packaging it for the non-AI-pilled. Opus 5.5 motion videos would be a recent example. We all know this is possible with AI now, but maybe there's a vehicle to package this for normies in a way that blows their minds and creates value for them. This relies on there being a knowledge gap though, which will shrink fast.
4) Just Do Hard Things (TM)
Do things that are genuinely hard, that you would never be able to build before. Build things that previously only a VC-funded company would be able to do, with hurdles from all angles not just tech, but also go to market, regulatory, etc. Doesn't guarantee success but it at least keeps competition away.
5) make a social media scheduling tool
just kidding about that last one.
Before you spend $$$ on an experimentation platform, ask who is responsible for:
implementation
tagging
QA
integrations
keeping the thing working six months later
Buying the software is usually the easy part.
This is sort of fascinating.
We talk about company culture as if you can sit down and design it.
But after running a company for 20 years, I think a lot more of it is simply the founder’s worldview showing up everywhere over time.
Apparently at Meta, some of that worldview is Ancient Rome.
xitter.cf/davidsenra/status/2106…
.@colossusjeremy says Mark Zuckerberg’s obsession with the Roman Empire goes back to when he was 15 years old, and traces of it are still everywhere at Meta:
“He has this long-standing interest in the Romans and in Augustus in particular. Augustus was the first emperor of Rome and turned it from a republic into an empire.”
“He used to end meetings by shouting, ‘Domination,’ and Sean Parker, the early president of Facebook, told him he had to stop doing it because it could be cited one day in an antitrust suit.”
“Courses he took in Latin in high school, where they had to read and memorize verses from Virgil's Aeneid, he still remembers now at the age of 42 and can recite them.”
“Just recently at their keynote, he brought out his personal Muse agent, and refers to it as Agrippa. Agrippa was one of the two key early lieutenants for Augustus.”
“I remember at some point Elon compared himself to Napoleon. I think a lot of these world-straddling figures at that level, they do think in these historical parallels.”