Build it, buy it, or connect what you have? The decision, with the costs.

The question is usually put as build or buy, and framed that way most owners choose wrong in one of two directions: they buy a subscription that does not know their trade and keep the spreadsheet alive beside it, or they commission a build for things a product already does well. The useful question is narrower. For each piece of the operation: does something already exist that knows the rules? If yes, buy it. If your tools exist but do not talk, connect them. Build only what is left.

When should you buy off-the-shelf?

When the rules are the market's, not yours. A sales pipeline is a solved problem; so is accounting, a shared calendar, courier booking, e-commerce, email. Products in those categories are better than anything you would commission, cheaper, and maintained by someone else. The mistake is not buying them; it is expecting them to know your trade. Pipedrive is sold as a CRM, and in the nursery it was tailored and built out until it ran the orders too — which is not what it is sold as, but is what it turned out to be very good at, because a pipeline with stages, owners and data fields is exactly the shape an order operation has.

The tell that you have bought the wrong thing is the spreadsheet that survives beside it. Per-user field-service software at £1,200–£5,000 a year with no fence take-off; procurement SaaS at £3,000–£15,000 a year built for parts and stock codes rather than living plants; routing software with no trigger from your CRM and no calendar your drivers already read. Each does its generic half well and leaves the half that costs you the hours untouched.

When is connecting what you have enough?

More often than anyone offers it. The nursery's problem was quoting that took fifteen minutes through a website and a spreadsheet, orders that lived in folders on a shared drive, and everything else in the owner's head. Nothing was built. The website was already there, and once it held every product line at its price it became the thing everything else could quote from. Pipedrive came in as the CRM and became the order system. The courier portals the business already booked through were fed from it, so a job going out was booked from the order rather than typed into a courier's website. Zapier and Make connected the three, with custom code only where those could not reach far enough — the only code written anywhere in the project, and it was glue rather than product.

Three separate products behaving as one seamless system: a price becomes a quote, a quote becomes a customer record, a paid quote becomes an order, the order moves through fulfilment, and nobody carries information from one screen to another. Quoting went from fifteen minutes to two or three. The business went from £1m to £3.2m with the office unchanged at three people. The skill was choosing well and connecting properly, not commissioning.

When do you actually need to build?

When the rules are yours and the hours are in them. The fencing contractor's cost was a surveyor drawing every fence twice — once with a tape in the garden and again for one to two hours at a desk — and pricing it from a printed list in the van. No product derives posts, gravel boards, postmix and a cut panel from a drawing, knows the firm's gate formula, and generates a customer quotation, a plan and a hi-vis installer job card from one survey. So that was built: 57 catalogue items, 29 rules, 3,341 tests behind the money maths, live in production eleven days after the first commit.

The same test picked out the other two builds. The delivery planner needed a scheduler that knew real road drive times, per-stop offload windows, HGV driver-hours law and the mandatory break for a 7.2-tonne lorry, and refused to crew a lorry with a van-only driver. The procurement pipeline needed password-free pricing pages for a hundred growers, a plant name that learns its own variants, and a lorry that fills against a route's value threshold. None of that exists off the shelf, and each is where the hours were. Build that part — and only that part.

What does each route cost?

Agency, freelance and SaaS ranges are indicative UK market figures for the scope in the case studies and should be validated against current quotes. The build figures are the starting prices of the documented engagements; scope varies.

The number to compare all of them against is not each other. It is what standing still costs — at the nursery, roughly £107,000 a year — and what the hire that would otherwise absorb the hours costs, every year, with on-costs.

How does each route fail?

• Bought, never used. Sold into the operation rather than built from inside it. The team was not in the room, the vocabulary is someone else's, and the first real Monday morning breaks it. A system the team will not use is a failed system whatever it cost.
• Built to a specification, not a seat. A rebuild designed from a whiteboard solves the problems that are easy to describe. One designed from the seat solves the ones that are actually costing money — which is why every one of the four builds was done by someone who had done the jobs first.
• The changeover nobody planned. A business that cannot afford a bad week will not risk one, so the new system waits forever. The fix is to run both on the same real orders until the new one has earned the switch. The nursery had no downtime at all.
• The SaaS with the spreadsheet beside it. The subscription handles scheduling and invoicing; the quoting still happens in a spreadsheet because the product does not know a gravel board from a gate post. You have paid for the generic half.
• The build that replaces things that worked. Commissioning a CRM when Pipedrive exists. Building only the rule no product knows keeps the budget on the hours.

The decision rule

For each piece of the operation, in order. Does a product know these rules? Yes: buy it and tailor it. Do the tools we have already exist but not talk? Yes: connect them before replacing anything. Is what is left a rule no product knows, and are the hours in it? Yes: build that, from inside the business, against real jobs, and only that. Is the bottleneck really the feature list, or the changeover? If the last system failed at adoption, no amount of building fixes it; run the next one alongside the old until it has earned it.

That is the sequence the operations audit runs through in two days inside the working week, and the wider method — where the hours go and what order to fix them in — is in the guide to systemising a trade business. The short version is on the byHoratio homepage.

Questions people ask about build versus buy