Hi all,
Happy Sunday from NYC.
We had a sick moment last month at Knight Vision: we were on a billboard in Times Square.
July 2026, less than one year into Knight Vision
We’ve helped some of the coolest companies grow. Building with startups (that have turned into a few uniconrs) shaping the future is by far the best part.
The craziest part you might not know: Knight Vision rly started when I began writing my newsletter on beehiiv, where I documented everything I was building at Citizen (btw, I still write it every Sunday). This turned into an insane community and pushed me to build out the company.
It’s been so incredible to work across AI infra, B2B software, and consumer brands changing how people live. Pretty magical...and we’re just getting started!
Now let’s get into today’s newsletter:
When something works, you build. You find the channel that's converting, the content format that's popping off, the play that's printing pipeline, and every instinct says do more.
Whether it's more allocated budget, headcount, hours, overall resources, it’s second nature to pour more fuel on an already burning fire.
That logic is no longer black and white.
Doubling down on channels has always been viewed as a safe bet because the channels had a shelf life measured in years. You could literally ride SEO for a decade. The platforms didn't move fast enough to punish you for betting big or, IMO, really care.
It’s not the same anymore.
Examples: Reddit lost most of its ChatGPT citations in four days this week. LinkedIn is switching up its algos all the time. More AI slop is flooding that feed too.
Here's my read on why: These models are supposed to be grounded in truth, yet people figured out how to fake it on Reddit for AEO in months: you can pay moderators and accounts with karma to make top posts. This is a typical sell in an AEO agency.
The same thing happens with a startup’s "grassroots" launch video. We all know which agencies you call. We all know what engagement looks like when someone does this (the same creators quote-tweeting, endless reposts that don’t match the following, engagement pods etc)
Both the hosting platform (e.g. Reddit) and the scraping platform (e.g. ChatGPT) get wise to this manipulation eventually. Same way Amazon made shipping fraud harder over time (not bulletproof but harder). The channels built on manipulating a system take the hit first.
LinkedIn's living this right now as they constantly rework their algo. The feed is so overrun with AI slop, friends are saying it feels like Facebook feeds in the old days and are asking if it's even worth opening anymore.
Your lesson: the shelf life on any channel you don't own just got a lot shorter, so stop letting your marketing team budget and allocate resources like it's 2019.
So, what do you do?
No, don't stop doubling down. Instead, build the muscle to double down and diversify at the same time, and that's a veryyyy different skill than picking a winning channel.
Here's what the allocation should look like and how I think about it.
SORT GROWTH CHANNELS INTO A BUCKET:
Step 1: Write out all of your growth channels and sort by owned, portable and rented.
Owned:
Examples: Referrals, direct sales relationships, your own CRM
Nobody can algorithm-update your reputation away
High concentration here isn't a risk, it's called having a strong business
Example: At Knight Vision, referrals and warm intros are basically our whole book. I don't lose sleep over that because that velocity matches how fast I want and need to grow (key).
Portable, but owned. Needs a platform to be useful, but the asset is yours.
Your email list lives via Gmail and beehiiv, but you can export every subscriber tomorrow for anything you need
A Slack or Discord community works the same way, you can rebuild the member list even if you change tools. Some platform risk here (pricing, features), but the core asset survives a breakup
Rented:
Examples: Organic & paid social, SEO, AEO
Your reach IS the platform's algorithm
You don't own the list; you're borrowing the audience; you’re likely paying to borrow the audience (e.g. paid ads to acquire users), and the platform can change anything, anytime without calling you first
Good luck here on this CAC treadmill. Always having to spend a disproportionate amount of $$ for growth is a rough spot to be in.
Step 2: Next to each channel, write out your user acquisition composition. Example: do you get 50% of your new users from paid ads?
The new reality is that if a rented channel crosses 55-60% of your pipeline, your testing budget needs to go up regardless of how well that channel is performing right now to de-risk yourself.
It’s the same logic in retail investing: diversification is key. And it would suck a ton to see your business get wiped out because a Product Manager told all the higher-ups at OpenAI that you can pay random dudes in a basement who happen to own top Reddit accounts $5K on CashApp for a top comment. Ironically, talk about bad karma…..
SO, HOW MUCH DO YOU ALLOCATE INTO TESTING?
The reality is super early-stage companies can't afford two full-strength channels, so "just diversify" isn’t the same at $500K ARR or even $1M ARR. You’re probably still just unlocking growth.
You need to focus on sequencing: keep one channel at full strength, run the most effective possible tests alongside it to get stat sig on if something new is working.
Example: At Citizen, I stalked X daily to see what users were talking about. When I noticed people sharing specific screenshots from the app, I tested having bigger accounts post the same stuff at the same time. Super grassroots! I was legit matching signup quality to post timing to prove stat sig (welcome to attribution!).
It worked insanely well for ~1.5 years until I dried it out. The whole time, I was testing new channels (like viral emails) so we had the next thing ready.
A test graduates to more budget when the economics repeat across independent cohorts (vs. when one week looks good), the users retain and also add MRR. Make sure what you’re scaling brings in quality users.
BTW, USE THIS SAME MUSCLE FOR RETENTION
Acquisition gets you customers. Retention is what happens after, and if that's carried by product alone, you're exposed.
You need to have multiple arrows of retention shooting at your customers. At minimum:
Customer calls
Making people feel special
Star cohorts of users that other ppl look up to
Events
IRL relationships
Brand
Because it does currently seem like a customer who stays because the product works and nothing better exists yet is not sticky enough.
The second a competitor ships a comparable feature, they're gone. It's a race to the bottom on price. The data won't warn you this is coming, because churn risk doesn't show up as a metric, it shows up as a relationship that was never built in the first place.
For B2B: With AI, migrations between B2B tools are also easier than ever.
The ties that keep a customer owned by your company vs rented live in your hands after Day 0. Especially at the B2B level, the accounts willing to learn and grow with you are the ones who stick around because they feel like you’re building for them.
Diversify your channels. Then diversify how you manage customer relationships.
Well, How'd I Do?
Audit your channels the way you'd audit anything else that is business critical: what happens if this disappears tomorrow with zero warning?
If you’re under pressure for venture-scale growth and the answer is "I'm fine," check again. If the answer is "I'd panic," go through the exercise in this newsletter and start testing into new channels ASAP. If you’re scrappy, this should be done in a cost-effective way.
Pick one channel this week and start a cheap test with the goal of having this new channel outperform the existing one. You have nothing to lose here and only upside to gain.
I hope you have an incredible week ahead.
Julia
