The default reaction to almost any business problem now is to go and find the tool for it. There is a SaaS product for every job, the free trial is one click, and by the end of the year you are paying eleven monthly subscriptions for eleven services that do not talk to each other. Each one felt cheap on the day you signed up. Together they are a running cost nobody reviews and a pile of data scattered across eleven logins.
Buying is often the right call. We are not against it, we pay for plenty of tools ourselves. The problem is that buying is the only option most people consider. There are three, and the reflex to rent skips the other two without a moment’s thought. So before any tool goes on the card, we ask a couple of plain questions.
The two questions most buyers skip
The first: is this a throwaway convenience, or a capability the business will lean on for years? A tool you use twice a quarter should always be rented. A tool that becomes load-bearing, the thing your operation quietly depends on every week, is worth a harder look, because you are about to marry its pricing curve and its roadmap for a long time.
The second: what am I actually paying for here? Sometimes the answer is real and worth it, the support line, the compliance, the fact that someone else keeps it running at three in the morning. Sometimes the answer is a nice interface wrapped around something you could run yourself. Those two situations look identical on the pricing page and cost wildly different amounts to replace. Telling them apart is most of the job.
A lot of what you are renting is a wrapper
This is worth saying plainly because it has become the quiet default of the software business. A large share of the AI products launched in the last two years are a thin layer over a model anyone can call. The company took an API that costs cents, put a clean interface and a login on top, and charges you a monthly seat for it. The model is doing the work. You are paying a markup for the wrapper.
Sometimes that markup is fair. The wrapper might hold genuine value: a workflow that saves your team real time, integrations you would never build, a support team that answers when it breaks. That is a product, and paying for it is reasonable. But often the wrapper is thin. It is a prompt and a text box, sold at a price that assumes you will never look underneath. Once you can see the model doing the work, the question changes from “which tool do I buy” to “do I need the tool at all, or just the thing it is sitting on”.
The same logic runs well beyond AI. A great deal of everyday business software is a convenient front end over open source that has existed for years and is free to run. The convenience is real. Whether it is worth the price depends entirely on how much you use it.
The example we point at most: email
Email marketing is where this gets concrete, because the pricing punishes exactly the thing you want, which is a bigger audience.
Take Mailchimp, the reflex choice. At the time of writing its Standard plan runs about 100 dollars a month for 5,000 contacts, 135 for 10,000, and 310 for 25,000. It charges you for contacts who have unsubscribed and for ones who never confirmed, so the real number is usually higher than the sticker. Above that, the public tiers stop and it becomes a custom quote. A list in the six figures on a per-contact service is a four-figure monthly bill, every month, forever, and it grows precisely as your list does.
Now the other path. Listmonk is an open source, self-hosted newsletter manager. The software is free. You run it on a small server, call it 5 to 20 dollars a month. For actual sending you wire it to Amazon SES, which charges 0.10 dollars per 1,000 emails.
Do the arithmetic on a real case. KVR Audio, a client of ours, runs a newsletter north of 170,000 subscribers. On SES, one full send to that list costs about 17 dollars. Send weekly and your delivery bill is under 70 dollars a month, plus the small server. The same audience on a per-contact platform is thousands of dollars a month. That is not a saving of a few percent. It is the same job done for a rounding error, because you stopped paying a per-head tax on a list that was only going to get bigger.
The honest cost of self-hosting
Here is the limit, and it is a real one. Self-hosting is not free, it just moves the cost from your card to your attention.
When you own the stack, you own the parts a paid service quietly handled. Deliverability is the big one: getting mail into inboxes instead of spam folders takes proper setup and ongoing care, and SES gives you the pipe, not the deliverability team. There is no support line to call when something breaks at the wrong moment. Someone has to keep the server patched and the software updated. None of this is hard for someone who does it for a living, and all of it is a genuine burden for a team that does not want to think about infrastructure at all.
So the decision is not “self-hosting is cheaper, always do it”. It is a trade. You are swapping a predictable monthly fee for a lower bill and a maintenance responsibility. For a two-person shop sending to 400 people, that trade is a bad one, pay Mailchimp and get on with your day. For an operation sending to a large and growing list every week, the paid fee becomes absurd and owning the stack pays for itself many times over. The size of your list is what flips the answer.
The test, in short
- Used rarely, low stakes. Buy it, move on.
- A nice UI over an API you would call. Price the thing underneath.
- Open source does it, at low volume. Still buy, the convenience wins.
- Open source does it, at high volume. Host it, own the stack.
- No off-the-shelf fit at all. Build it, deliberately.
- Really a rules or process problem. Neither, fix the process.
Most of the money gets wasted on the second and fourth lines. A wrapper bought without looking underneath, or a per-unit subscription kept long past the volume where it made sense. Both are easy to catch if you stop to ask the two questions before renewing.
What to do first
You do not need to rip anything out this week. Start with an inventory. List every monthly software subscription and what it actually does, then mark the one or two that are load-bearing and growing, the ones whose bill climbs every quarter. Those are the candidates. For each, ask what sits underneath it, and whether the thing underneath is something you could run yourself or something genuinely worth renting.
That is the work we do when we help a client on the custom software side. Often it is not building something new at all. It is finding the two or three subscriptions that quietly became a tax, replacing them with something the business owns, and leaving the rest exactly where they are because renting them is the right answer. The goal is never to self-host everything. It is to stop paying rent on the things you should own, and to know which is which.
