About Bobby Huang
Two things have taken up most of my working life, and they don’t usually belong to the same person. One is growth: finding what actually moves revenue in a company and building the machine that keeps moving it. The other is bookkeeping: 18 years inside other people’s books, still reconciling real client ledgers today.
Growthy is what happens when those two threads meet. So is most of the rest of this page.
The short version
Who I am today
I’m Bobby Huang. I’m the founder and CEO of Growthy, an AI bookkeeping platform, and a partner at SDO CPA LLC, where I run operations, systems, marketing, and sales. I have 18 years of hands-on bookkeeping and accounting behind me, which is a career length rather than a tenure at any one firm. I still reconcile real client books, which is how I stay current on the processes I’m building software for. I invest. And I advise a small number of companies on growth, which is a different thing from running it for them.
One clarification, because the firm’s name invites the wrong assumption: I’m not a licensed CPA. I’m a partner at SDO CPA LLC, with 18 years of hands-on bookkeeping and accounting behind me. That’s the experience I’m actually trading on, and I’d rather say it plainly than let you guess.
If you landed here looking for who runs Growthy, that’s me.
The growth side, in numbers
I’ve run growth since 2015: Bellhops, Designlab, BlueStacks, vidIQ, Mable, Archive. As Head of Growth at vidIQ I doubled recurring revenue two years running, cut churn 40%, and took return on ad spend from 20% to 500% in three months. At Archive it was 600% ARR growth in the first year and a content system that grew organic clicks 2,700% in under a year. At BlueStacks, as Growth Product Marketing Manager, I built an automated acquisition system that started on a $10 budget and generates eight figures in revenue and profit today. On contract since then, I helped take an AI tax startup from four figures to seven, and a local business from low seven figures to over eight.
None of that replaced the books. The two threads ran at the same time, which is the part people find odd and the part that turned out to matter. That work is also the reason the consulting page exists.
What I learned in 18 years of books
The artifact from this stretch isn’t a job title. It’s a reconciled ledger, repeated across 18 years.
Why the books matter
I didn’t start in bookkeeping because I had a plan. I started because that’s where the work was. What I got out of it was a specific kind of knowledge, and it isn’t the kind you can pick up from a product demo or a competitor teardown.
You learn where a month-end actually breaks. It’s rarely the big transactions. It’s the recurring vendor that changed its billing descriptor in March, or the owner draw that got coded to office supplies for eleven months before anyone caught it. Categorization rules drift the moment a business changes anything about how it operates, and they drift quietly. A clean audit trail has a real cost in human hours, which is the number nobody puts in a pitch deck.
The other thing you learn is how much of this job is reconstruction. A set of books that arrives late is a set of books someone has to rebuild from statements and memory, and the further back you go, the more of it is inference. Most of the argument about automating this work is really an argument about how much inference you’re willing to let a machine make on your behalf, and on whose numbers.
That knowledge doesn’t age out. The tools changed a lot over 18 years. The failure modes barely moved.
Still reconciling client books
I’m still reconciling real client books, not reading summaries of them.
I keep it that way on purpose. The moment I stop, I start building from memory, and memory of bookkeeping goes stale faster than people admit. Staying in the work means every claim I make about what’s hard in it is current rather than remembered.
It also keeps me honest about what software can and can’t take off a person’s plate. That’s a much harder line to hold when your only exposure to the problem is a roadmap and a support queue.
What I learned running growth
Growth work taught me the opposite lesson from bookkeeping, and I needed both.
Books teach you precision: the number is right or it isn’t. Growth teaches you that most of the numbers people stare at don’t matter, and the constraint is almost never the one being discussed in the meeting. A team asks for more traffic when the real problem is that trial users never reach the moment the product proves itself. Demand looks fine and the pipeline is leaking between the first call and the proposal.
So the job I’m actually good at is finding the lever. Which one moves revenue this quarter, which one compounds over the next two years, and the fact that those are rarely the same lever.
The deeper habit underneath that is treating acquisition as a science rather than an art. Most companies run on hunches: a channel gets funded, the numbers move or they don’t, and nobody can say which part caused it. The alternative is unglamorous and it works. Define what a qualified lead actually is. Instrument the path from first touch to signed. Run experiments that are allowed to fail. Keep what survives contact with real numbers.
The channels are the easy part to list: organic and search, paid on Google, Meta, TikTok and a dozen smaller networks, founder-led social, lifecycle email, affiliate, and the sales or self-serve machinery that turns attention into paying customers. I’ve built all of them, across consumer apps, B2B SaaS, healthcare, education and professional services. What varies is which one your business actually needs first.
Building Growthy
The artifact here is the product itself. The fastest way to judge it is to go look at it rather than read me describing it.
Why an AI bookkeeping platform
Growthy came directly out of the books. I’d watched the same handful of failures repeat across years of ledgers, and they were all shaped the same way: a person doing pattern-matching that a machine is better at, then a machine being confidently wrong about the parts that need judgment.
I built Growthy because I could describe that problem in specifics, which is a far better starting position than being early to a category. When you’ve done the work, you know which parts of it are genuinely repeatable and which parts only look that way from the outside.
There’s a version of this company built by someone who read about the problem instead of doing it. That company ships features that demo well and break on the fourth client, because the fourth client has a chart of accounts that looks nothing like the first three. I’ve seen enough charts of accounts to know how much variation is out there, and that variation is the actual engineering problem.
What Growthy does today is on growthy.com. I’d rather point you at the live product than narrate a roadmap here.
Running it on its own product
Growthy’s books run on Growthy.
Our own close happens on the same surface a customer’s does. When something in the reconciliation flow is annoying, it’s annoying to us first, on our own numbers, before a support ticket ever shows up. When a report is wrong, it’s wrong about our money.
That removes a specific kind of self-deception. You can’t hand-wave a rough edge you have to walk through yourself every month.
It also shortens the loop between noticing a problem and fixing it. There’s no translation layer between the person who felt it and the person who decides what gets built. Most weeks those are the same person.
Systems over stunts
The build that’s least visible from outside is the one I get asked about most: a growth engine that keeps running when nobody’s watching it.
Most of my work is the background machinery rather than anything with my face on it: content built around what people actually search for, a conversion path that qualifies before it books, a sales process with real scripts, and email follow-up driven by what someone did instead of what day it is.
SDO CPA is the clearest version, because I own the whole funnel there. It runs on zero ad spend. Not low, none. What that produced isn’t a traffic number, it’s a firm with a waitlist that raised its prices on purpose and refers work out because it’s backed up.
I do run a personal account, and what I write is the output of it. That’s a thing I do for myself, not the thing I sell.
Underneath all of it sits the part most people skip, and it’s where 18 years of books earns its keep: the reporting, the unit economics, the tracking definitions. Growth built on numbers you don’t trust produces confident decisions in the wrong direction.
I build with AI daily, and my companies run on systems I wrote for research, drafting, review, and the internal work a team would otherwise do by hand. That’s not a side interest. It’s how a small operation does the volume this page describes.
A launch is an event. It has a date, and the day after that date it starts decaying. A page you publish once keeps earning attention long after you stop touching it, and an email sequence you write once keeps running for whoever signs up next. Stunts have to be repeated, and each repetition costs more than the last. I’ve run both, and I know which one I’d rather own in year three.
That’s also why I operate instead of advising from a distance. Systems are easy to describe and hard to keep running, and the whole difference lives in details you only meet by holding the thing. What I’m operating right now is on the companies I run.
Where to find me
LinkedIn is the one place I’m consistently reachable. I read messages there and I answer them, including from people I haven’t met, as long as there’s an actual question in it.
If you’re evaluating Growthy, checking a reference on me, or you’re an operator working on something in the same neighborhood, start there.
There’s no form on this site, on purpose. LinkedIn is the only way in, and how to reach me covers what to put in the message depending on what you’re asking about.