lab-reports

Our CRM Renewal Was Coming Up. So We Built Our Own.

Why we ditched a $15k HubSpot renewal to build Cheddar, a lean, AI-era CRM tailored to our 30-person agency—and the $500 experiment behind it.

By Ben Kalkman · 9 min read · crm build vs buy saas pricing models ai in software startup operations hubspot custom software

A bet on where software is heading, and a $500 experiment we're about to run on our own business.

The thing that started this wasn't dramatic. It was a renewal notice.

HubSpot, our CRM for years, was coming up again at close to $15,000 a year. I opened it, did the math I do every year, and this time a different thought showed up and wouldn't leave: what are we actually paying for here, and could we just build it ourselves?

A year ago, I would have laughed that thought off. This time I didn't.

First, the honest part: HubSpot is good

Rocket Media® has been a HubSpot partner for a long time, and I'm not here to throw stones at a platform we've recommended and used happily. HubSpot is genuinely excellent. The integrations, the polish, the depth. For a lot of companies, it's the right call, and we still set clients up on it.

But here's the quiet truth about any platform built to fit every business: it fits no business exactly. We were on seat limits. There were features we kept wishing we had and couldn't justify the jump in tier to get. Making it really ours took customization, setup, and consulting hours. In the end, we were paying enterprise prices to mold ourselves into a shape somebody else designed for everybody.

And our shop is small. We're under 30 people with roughly a hundred clients. Our CRM needs are narrow compared to a sprawling sales org. We didn't need the box that does everything. We needed the box that does our things.

Why "can't we just build it" stopped being a naive question

Eighteen months ago, that question was naive. Today, the entire software industry is having this exact argument out loud.

In December 2024, Microsoft CEO Satya Nadella said "SaaS is dead" on the BG2 podcast, and the comment set off a year of debate. IDC now projects that by 2028, pure seat-based pricing (the thing I was sitting there frustrated about) will be obsolete, with most software vendors reworking their pricing models. The cost of building has fallen off a cliff at the same time: Bain points out that one frontier AI model dropped 80% in cost in just two months. And tools like Replit, Lovable, and Bolt can now stand up a working, full-stack application from a conversation.

So no, the question isn't crazy anymore. The pieces actually exist.

I'll add the counterweight, because someone has to. AI researcher Andrej Karpathy says agentic AI is closer to a decade away from matching the hype than the headlines suggest, and he's probably right. That's exactly why I'm not going to sit here and tell you we cracked the future of software. We ran an experiment. That's all.

The gap between the headline and the truth

The biggest "we built our own" story out there is Klarna. They told investors they shut down Salesforce and Workday, and credited AI with helping the company turn profitable. The headline reads as if the future has already arrived.

Read the footnotes, and it gets quieter. Later reporting suggests Klarna leaned on lighter software alternatives with AI layered on top, not a from-scratch in-house replacement of those giants. And Salesforce's Marc Benioff asked the fair, pointed question right back: how exactly are you handling all your customer and employee data now?

Both things are true at once. Something real changed, and the headline oversold it. That gap, between what's genuinely shifting and what's being hyped, is exactly where I like to live. So when we built our own, I tried to keep one foot in each.

Meet Cheddar

We built one. We call it Cheddar, because who doesn't need more cheese.

Cheddar is modeled closely on CRM's we like, but shaped entirely around how Rocket Media actually works. Just what we need, only what we need, exactly what we need. Nothing we'd never touch, no tier we're paying for to unlock one feature.

I've learned this lesson the hard way more than once. The golf cart sales site I built between holes with my dad. The tribute site I made for my wife that she politely wasn't impressed by. Every time, the lesson was the same: build for what people actually need, not for what you're capable of building. Cheddar is the first thing I designed to that lesson from the very first line.

What it does is the unglamorous day-to-day: a drag-and-drop pipeline; deals, contacts, and companies; leads that get scored and convert to a deal in one click; tasks; a home dashboard; and real reports (funnel, win/loss by reason, an owner leaderboard).

The part I didn't expect to care so much about is that Cheddar isn't an island. It integrates with the tools we already pay for and use, ironically, other Saas products. Granola for our meeting notes. Mailchimp for nurture. ClickUp for the handoff when a deal closes. Gmail for sending. It turned out to be less a replacement product and more the hub that finally got the tools we already own to talk to each other.

And the AI is woven through it, not bolted on. There's an "Ask Cheddar" assistant that can actually go look things up and take action across the system. Meeting briefs that prep you before a call and recap you after. A suggested next best action on every deal. Lead scoring. Drafted emails. One small detail I love: Cheddar does its homework before we wake up. An early-morning routine runs so the AI work is ready and waiting by the time the team logs in.

How it actually got built

Honesty first, because it matters: I'm not a developer. I've written about that before, and it's still true. AI did the heavy lifting. Cheddar was built on Replit and powered by Claude, using roughly $500 in credits.

Sit with that number next to the other one. Fifteen thousand dollars a year for the box that fits everyone, or about five hundred dollars to build the one that fits us. That's the "compared to what" that made the whole experiment worth running.

What surprised me wasn't that it worked. It was the discipline the build forced on me. The single most important decision was establishing a single source of truth for how every part of the app talks to every other part, so the screens you see and the rules running underneath can't quietly drift apart. There's a real suite of automated tests under it, and identity and login are handled by a dedicated provider built for exactly that, not something I duct-taped together at midnight. I'm going to say it plainly: the system is strong. I was not expecting to feel as confident as I do.

What we're actually signing up for

Now the part I won't oversell.

When you buy software, the vendor quietly owns the hard parts. Uptime. Updates. The 2 AM problems. The moment you build your own, you own all of it. That trade is bigger than the price tag, and pretending otherwise would make me exactly the kind of hype merchant I can't stand.

On the question everyone rightly asks, the same one Benioff asked Klarna, about data: Cheddar runs on its own independent server, it's internal only and accessible just to our team, and it went through a security review on the platform we built it on. We didn't skip that step. I'm comfortable here. I'm also not naive about the fact that owning your own system means owning responsibility for it, for as long as you run it.

Maintenance is the unglamorous truth underneath all of this. There's no roadmap handed to us, no support line to call. If something breaks, that's us. And because Cheddar leans on connections to other tools we use, part of its health depends on those staying healthy, too.

There's a voice in the back of my head that has been burned by "we'll just build it" before, and I'm keeping that voice close. It keeps asking whether I really want to trust something this important to a system we made ourselves. That skepticism is doing real work here. I'm bullish on this. I'm not blind.

So, brilliant or foolish? Here's how I'd decide

I don't know yet. And anyone who tells you they know the answer for your business is guessing.

But there is a way to think it through honestly:

Blog image

For Rocket Media, the boxes lined up: narrow needs, small scale, and a gap between $15,000 and $500 that was hard to ignore. So we ran the experiment. For a 500-person company, I'd almost certainly tell you to keep buying.

This isn't "fire your SaaS." It's that the line between buy and build moved this year, quietly, and it's worth knowing which side of it your business is actually on now.

Okay, here we go

So that's where we are. Cheddar goes live this week.

The concept is strong. The system is strong. And the only honest thing left to say is that time will tell. We're going to run our business on something we built ourselves, watch it closely, and tell you the truth about what happens, the win or the face-plant.

Okay, here we go. The jury is still out. We're testing and watching, a little anxious to see whether this proves out, and genuinely curious where it goes next.

I'll report back.

If you're an operator staring at your own renewal notice, wondering whether you could just build it, that question deserves to be taken seriously now in a way it didn't a year ago. It's also surprisingly easy to get wrong. That's the conversation we have at Digital Ignitor: not "build everything," but working out where building actually beats buying for your specific business, and where it would quietly burn you. If that's where your head is, let's talk.

Sources

Note: SaaS spending figures are forward-looking projections, not measured results.