The real 5-year cost of the tools you rent.

Every subscription you pay for made sense the day you bought it. That’s the trap. Nobody signs up for a $36,000 software bill — they sign up for $20 a seat, eleven times, and let the renewals do the rest. So before I make any argument about building software you own, I want to do something simpler: total the bill. Real tools, the vendors’ own list prices, the increase rates measured across billions in actual spend, five years. Then you can decide whether the number bothers you.

The short version, for the impatient: a typical small-business stack costs about $500 a month today. At the price-increase rates the industry actually charges, five years of that is $36,631. A one-time build in my usual range, plus five years of hosting you pay directly to your own provider, is $23,000. The gap is $13,631 — and unlike the subscriptions, the build is still yours in year six, at $0 a month. Every assumption behind those numbers is below, and all of them are editable in the calculator.

§1The bill nobody totals

Here is a software stack I see constantly in small businesses — a handful of good tools, each reasonable, none of them ever finished. Prices are the vendors’ own, for annual billing, checked July 2026. For a team of six:

That’s $390 to $540 a month. And that’s before anyone adds a seat, upgrades a plan for one missing feature, or bolts on another integration. I use $500 a month as the working figure through this entry. It sits in the middle of that range and matches what callers describe to me — but it is an assumption. Replace it with your own bill. Add up what actually left your account last month. Most owners I talk to have never done it, and most are higher than they guessed.

Totaling your own bill takes ten minutes and three steps, and it’s worth doing before you read another word. One: pull the last three months of card and bank statements and list every software line. The quarterly and annual charges hide there, and they’re the ones owners forget. Two: for each per-seat tool, multiply the plan price by the seats you’re paying for — not the people who actually log in. The gap between those two numbers is §3 of this entry. Three: write the monthly total somewhere you’ll see it. On scoping calls I ask for this number first, and the most common reaction to hearing it out loud is a pause. If your total came out under $300, stop reading here; §6 explains why. If it came out at $500 or above, the rest of this entry is your situation, priced.

Notice what’s carrying the total: it isn’t chat. Slack at $43.50 is cheap for what it does, and I’d tell any six-person team to keep it. The weight is in the tools that run your operations. The database that holds your inventory or clients. The automation glue. The industry tool that almost fits your process but never quite does. Those are the lines this entire argument is about.

§2The increase clause

The $500 you pay today is the cheapest that stack will ever be. Software vendors raise prices — not occasionally, structurally. Vertice’s SaaS Inflation Index is built from more than $30 billion of processed software spend. It measured average business-software price growth at 12.2% a year, running as high as 13.2% in March 2026. General consumer inflation sat around 2–3% in the same period. Software prices are rising at four to five times the rate of everything else you buy (CFO Dive’s coverage is a readable summary).

The increases arrive through three doors, and only one of them looks like a price rise:

• The list-price letter. The email that says your plan is going from $20 to $24 a seat. Annoying, visible, and the smallest of the three.
• The seat ratchet. Per-seat pricing means every hire raises your software bill automatically. The vendor never raised prices; your business just grew into a bigger invoice.
• The plan wall. The feature you need next (better permissions, an API, real reporting) lives one tier up. You don’t upgrade one feature; you upgrade every seat. A team of six moving from $10 to $20 a seat just took a 100% price increase that no letter announced.

To see the seat ratchet in real numbers, take the §1 stack and hire two people. The Airtable, Notion, and Slack lines alone climb from $223.50 to $298 a month. That’s a 33% jump no vendor letter announced and no budget meeting approved. Growth did it. Under per-seat pricing, your software bill is a tax on hiring.

When I model five years in this entry I use 10% a year, deliberately below Vertice’s measured 12.2%. I’d rather the math be conservative than convenient. If you think your vendors are gentler, the calculator has a 0% setting; run it and see how much of the argument survives even then.

§3The waste layer

There’s a second cost inside the subscription model that has nothing to do with price increases: paying for what nobody uses. Zylo’s 2025 SaaS Management Index, drawn from the license data of the companies on its platform, found that 52.7% of purchased licenses sit idle. Half the seats, bought and renewed, used by no one.

One caveat, because this entry only works if the numbers hold up: Zylo’s data skews toward large enterprises, where forgotten licenses hide easily. A six-person business doesn’t waste half its seats. But you don’t escape the mechanism — you meet it in miniature. The Airtable seat for the contractor who left in March. The plan tier you upgraded to for one campaign and never downgraded. The two tools whose features overlap because each one almost did the whole job. On a small team the waste layer isn’t 52.7%. But in the stacks I get shown on scoping calls, one line in five is typically buying nothing. On a $500 bill, that’s roughly $100 a month of pure renewal inertia.

Owned software has no waste layer, for a structural reason rather than a virtuous one: there are no seats. When I hand over a build, your whole team uses it, six people or sixty. The monthly software cost of adding employee number seven is zero.

§4The five-year tape

Now the actual math. Take the $500-a-month stack from §1, apply the conservative 10% yearly increase from §2, and run five years:

Total over five years: $36,631. And year six opens at $732 a month and climbing, because the tape doesn’t stop — that’s the defining feature of rent. Two things about this number are worth sitting with. First, no single year looks alarming. The bill grows by the price of a nice lunch each month, which is exactly why nobody acts on it. Second, at the end of it you own what you owned on day one: nothing. Cancel in month 60 and the software, the automations, and in many cases the shape of your own data all stay behind with the vendors.

Against that, the owned side, priced the way I actually quote: most of my projects land between $8,000 and $40,000, fixed before work starts. A build that replaces an operations stack like §1’s typically sits in the middle of that range, so I model $20,000. Add hosting, which you pay directly to your own provider at roughly $50 a month, and the five-year total is $23,000. The difference is $13,631 kept. The crossover (the month where cumulative rent avoided overtakes the build cost) lands around month 40 on these assumptions. Faster if your bill is higher than $500; slower if it’s lower. That’s not a rhetorical flourish; it’s the arithmetic. You can bend every input of it in the calculator until it matches your business instead of my example.

There’s also a cost the tape doesn’t show, because it only appears when you try to stop paying: the exit. Leaving an operations tool means getting your data out, and export functions are the feature vendors have the least incentive to perfect. What comes out is typically a pile of CSVs shaped like the vendor’s database, not like your business. The automations don’t export at all. The attachments come separately, if they come. The relationships between records, the part that made the tool useful, flatten on the way out. None of that is malice; it’s the same incentive from §2 pointing the same direction. But it means every subscription carries a switching cost that grows with each year of data you feed it. That cost quietly turns “we could always leave” into “we’ll renew one more year.” Owned software has an exit too: a database on your own server, in standard formats. The exit is a backup file.

And because a single worked example can hide a lot, here is the same math stress-tested at the calculator’s other settings. At 0% increases, rent totals $30,000 against $23,000 owned. The build still wins by $7,000, with payback around month 45 — and a five-year price freeze has never happened to a stack I’ve reviewed. At 15%, closer to what Vertice measured at its 2026 peak, rent totals $40,455 and payback arrives around month 38. The conclusion isn’t delicate. It holds at every increase rate the industry has actually exhibited. The only input that genuinely flips it is a small starting bill, which is exactly the boundary §6 draws.

§5What owning actually costs

The agency blogs that rank for this topic love a clean “$0 after launch” line. It isn’t true, and pretending it is would poison the rest of this entry, so here is the owned side’s real bill:

• Hosting: $20–150 a month depending on the size of the system, paid to your provider — Hetzner, DigitalOcean, whoever you choose. Not to me. I model $50.
• A domain: $10–20 a year. Rounding error, but it’s real.
• Maintenance: for the first 60 days, bugs I introduced are fixed free under warranty. After that you choose: a small care agreement with me, any developer you like, or nothing at all. Software built on a standard stack does not rot in a year; my own products have run for years between meaningful maintenance touches. But “nothing” is a choice you should make knowingly, not a promise I’ll pretend to give. The way to plan it: budget a small yearly check-up the way you would for any equipment your business runs on. Then treat a year with no spend as the pleasant surprise it usually is.
• New features: when you want the software to do something new, that’s new work, quoted as its own small fixed price you approve first. Owned software isn’t frozen software; it’s software where change is a decision instead of a subscription.

What the owned side structurally does not have: per-seat charges, plan walls, renewal letters, and the waste layer. The monthly line is a server bill, and the seventh employee costs the same as the first — nothing.

§6When renting still wins

If the five-year tape were the whole story, nobody would rent software, and yet renting is often the right call. I turn away work on exactly these grounds, so this list is load-bearing, not decoration:

• You need it running this week. Custom takes 6–8 weeks. A subscription takes an afternoon. Urgency is a legitimate reason to rent.
• Your process still changes every month. Custom software fits a shape. If your shape isn’t settled, you’d be paying to rebuild it quarterly. Rent until the process stabilizes.
• Your total spend is under about $300 a month. Below that line, the five-year math rarely favors building — the payback stretches past the point where the rebuild is worth the trouble. My own pricing section says the same thing: under $5,000 of budget, a good off-the-shelf tool beats me.
• You’re testing an unproven idea. Validate on rented tools first. Build once the idea has survived contact with customers.
• The tool genuinely fits. Slack fits almost everyone; I have never once proposed replacing it. The argument in this entry is about the operations tools that almost fit — the ones you’ve built workaround spreadsheets beside. If your tools actually fit, keep them and close this tab.

The flip side, stated plainly: at $300+ a month of spend, a stable process, and multiplying workarounds, §4’s math usually pays back within 12–24 months. That’s the profile of every build I’ve taken on.

§7Run your own numbers

Everything above uses my example figures, conservatively chosen and cited to the vendors’ own pages. Your business isn’t my example. The calculator on the homepage takes three inputs: your monthly spend today, the increase rate you believe (0%, 10%, or 15%), and a build price. It prints the five-year tape for both sides and tells you which one wins. If renting wins for your case, it says so — and so will I on a call.

And if you’d rather not do the totaling alone: send the software list with your project brief. Running your tape is the first thing I do in a scoping call anyway. It takes ten minutes, and the number is yours to keep either way.

A note on method, because this entry will age. The prices in §1 are list prices for annual billing, checked in July 2026; the sources below link to them. The 12.2% inflation figure and the 52.7% idle-license figure come from Vertice and Zylo, both measured on their own platforms’ data, both enterprise-weighted. That skew is why my working model uses gentler numbers than either. When these figures move meaningfully, this entry gets an updated line at the top — not a silently changed date.