Back to Blog
Channels & Growth

Build vs Buy vs Configure: Choosing Operations Software

Build vs buy vs configure for a consumer brand's operations software: a three-year cost worked line by line, who maintains each, and a two-question test.

Oz Merchant12 min read

A woodworker stands between a half-built custom cabinet, a sealed flat-pack box and a finished modular shelving unit.

A freelance developer quotes $18,000 to build the operations system a candle company has been running on spreadsheets. Over three years that build costs somewhere between $90,000 and $167,000, and about $128,000 in the central case, roughly seven times the quote.

Build vs buy vs configure is the choice between three ways to get software for orders, stock, purchasing and the numbers behind them: build it yourself (a developer, a no-code database, or scripts and automations around spreadsheets), buy a packaged product and use it roughly as it comes, or configure a larger platform that a partner or someone on your team shapes to the business. In every option, the people who build, own and reconcile the system cost more than the software, so the right choice is usually the one whose owner you already have.

What build, buy and configure mean in practice

Build covers a wide range. At one end, an operations manager keeps a well-structured spreadsheet and connects it to Shopify and the accounting software with a handful of automations. At the other, a developer writes a small application against the Shopify, Amazon and eBay interfaces, with its own database and screens. Building is cheaper to start than it used to be. Owners we talk to are building their own tools with AI assistants, and one who paid for a custom AI agent told us it never did what they wanted.

Buy means a packaged product with a standard way of doing things: an inventory and order tool, an accounting connector, a purchasing app (the DTC stack guide sorts them by job). You change your habits to fit it, set it up in days or weeks, and pay a monthly subscription. The vendor maintains the software; you maintain your settings and whatever the product does not cover.

Configure means a platform broad enough to run most of the business, set up to match how you work. A partner, or a technical person on staff, builds the workflows, fields, approval steps and connections over a project of several weeks or months. Most ecommerce ERPs, including NetSuite and its alternatives, sit here.

The three-year cost of each option, worked line by line

The example company sells candles and diffusers: about $3.3 million on Shopify, $1.8 million on Amazon, $200,000 on eBay and $700,000 wholesale to roughly 40 gift shops and boutiques, which pay $17 for a candle that sells for $36 online. It carries 180 products, ships from one 3PL and Amazon's warehouses, and plans to grow its wholesale program in year two.

Every figure in this comparison is an assumption, and each one is yours to replace. Platform quotes in the configure tier vary widely with users and modules, so get yours before running the table.

AssumptionFigure used
Operations manager, loaded hourly cost$45
Freelance developer$90 an hour
Implementation partner$175 an hour
Bookkeeper$75 an hour
Build: first version200 developer hours over three months
Configure: implementationA fixed-scope quote of $35,000 over three months

And this is what each option costs over 36 months. The central case comes first; the three lines that are most a matter of judgement also show a low and a high in brackets.

LineBuildBuyConfigure
First invoice$18,000 (developer)$3,000 (vendor onboarding)$35,000 (partner)
Software subscriptions$10,800$27,000$54,000
Internal setup timeIncluded in owner time$2,700$6,750
Owner's time to run it$42,120 ($28,080 to $63,180)$35,100 ($21,060 to $56,160)$25,740 ($12,870 to $38,610)
Developer or partner after launch$29,700 ($8,910 to $44,550)None$28,875 ($11,550 to $57,750)
Change: wholesale program in year two$5,400 ($2,700 to $9,000)$6,000 ($3,600 to $12,000)$7,000 ($3,500 to $14,000)
Bookkeeper reconciling what does not arrive cleanly$21,600$8,100$5,400
Three-year total, central case$127,620$81,900$162,765
Three-year total, low to high$90,090 to $167,130$65,460 to $108,960$129,070 to $211,510
Cost of leaving after year three (not in the total)About $4,000About $3,000About $15,000

The arithmetic behind each column:

  • Build. $18,000 for 200 hours. No-code database, automation and hosting subscriptions at $300 a month for 36 months, $10,800. A developer retainer of 10 hours a month after launch (3 to 15 in the range), $900 a month for 33 months, $29,700. The operations manager runs the system and patches it, 6 hours a week for 156 weeks (4 to 9), 936 hours at $45, $42,120. Adding wholesale invoices, terms and store-level pricing in year two, 60 developer hours (30 to 100), $5,400. Some tools also charge for their API, and Amazon has a company writing its own code complete a developer profile, which Amazon evaluates. The bookkeeper spends 8 extra hours a month matching exports to the accounts, $600 a month for 36 months, $21,600.
  • Buy. Subscriptions and connectors at $750 a month, $27,000. Vendor onboarding $3,000 plus 60 hours of internal setup, $2,700. The operations manager runs it, including the spreadsheet that covers what the product does not, 5 hours a week for 156 weeks (3 to 8), $35,100. Wholesale needs a higher tier in year two, $250 a month more for 24 months ($150 to $500), $6,000. The bookkeeper spends 3 extra hours a month, $8,100.
  • Configure. Subscription at $1,500 a month, $54,000. The partner's $35,000 quote, plus 150 hours of the operations manager's time during the project, $6,750. After go-live the manager runs it 4 hours a week for 143 weeks (2 to 6), 572 hours, $25,740. Partner support of 5 hours a month for 33 months at $175 (2 to 10 hours), $28,875. Wholesale configuration in year two, 40 hours (20 to 80), $7,000. The bookkeeper spends 2 extra hours a month, $5,400.
  • Cost of leaving. A new developer learning the code, about 40 hours; exporting and remapping the data; a configuration that does not move to another platform, plus any contract term still owed.

Under these assumptions, buying costs about $82,000 over three years in the central case ($65,000 to $109,000), building about $128,000 ($90,000 to $167,000) and configuring about $163,000 ($129,000 to $212,000). The ranges overlap: a build with a light-touch owner and a quiet developer can cost less than a bought product that needs a lot of tending. Configuring costs the most here and also does the most: it is the option in this example with the smallest monthly clean-up for the bookkeeper, and the one where a second warehouse is a settings change rather than a project.

Add up the people lines (everything except subscriptions) and they exceed the software in all three columns: about $117,000 against $11,000 for the build, about $49,000 against $33,000 for buying, and about $109,000 against $54,000 for configuring. The same holds at the low end of every range.

The developer and partner hours after launch have the widest ranges, because nobody can price a channel's interface change before the channel announces it. A fixed-scope implementation quote also covers only what is written in the scope; work left for a "phase two" is often billed separately.

The spreadsheet that is already working

Plenty of companies this size run well on a carefully kept spreadsheet and a few automations. Automations at $100 a month come to $3,600 over three years; the operations manager's 5 hours a week come to $35,100; 6 extra bookkeeper hours a month come to $16,200. That is about $55,000, less than any option in the table, for as long as the business stays the shape it is.

The signal that it has stopped working is time: the weekly hours creep from 5 to 12, month-end waits on one person's file, or nobody else can answer where a purchase order stands. Until then, a spreadsheet with clear owners, locked formulas and a weekly check against Shopify and the bank is a legitimate system. The ecommerce accounting stack guide covers the tools companies usually combine at this stage.

Who maintains it when the person who built it leaves

A build lives in the developer's head unless it was written down. A new developer has to read the code before changing it; the cost-of-leaving line above assumes about 40 hours of that. Three things reduce the risk: the code and the no-code base sit under company-owned accounts, not the developer's; each automation has a one-paragraph note saying what it does and what breaks if it stops; and a second person can restart a failed job.

With a bought product, the vendor maintains the software, but your settings and workarounds are yours. When the operations manager who chose them leaves, the replacement inherits a product that works and a side spreadsheet that nobody fully understands. Write down which parts of the process live outside the product, and why.

A configured platform depends on the partner and the internal owner together. Partners reassign staff, so ask for the configuration to be documented as it is built: every custom field, workflow and connection, with the reason for it. A platform nobody inside can explain becomes, in practice, a build that belongs to someone else.

When a channel changes its API, or the business adds a warehouse or wholesale

A channel's interface changes on the channel's timetable, not yours. Shopify, for example, releases a new API version every three months and supports each stable version for at least twelve months; a request aimed at a retired version is answered by the oldest version still supported. A script nobody has touched in a year can keep running on rules it was not written for, and the first symptom may be a gap in the numbers rather than an error message.

In a build, every such change is your developer's job, on the channel's deadline. In a bought product, updating the connector is the vendor's job; ask how much notice they give. In a configured platform, standard connectors are typically maintained by the platform or its connector vendor, and anything the partner wrote by hand is maintained by the partner.

A second warehouse tests the data model. If the build tracked stock as one number per product, adding a location means rebuilding the stock tables and every report that reads them. A packaged inventory tool may include several locations or keep them for a higher tier; check before you buy. A configured platform usually treats a location as a setting, plus the partner hours to map it.

A wholesale program tests the money. Stores buy on terms, so each invoice becomes an accounts receivable balance with a due date, and many stores take deductions for damages or promotions. Whichever option you choose has to record the invoice, the payment and the deduction, and has to produce an aged list of what each store owes. In the table, that change is the $5,400, $6,000 and $7,000 line.

What each option does to the accounts

Three questions apply to every option: do sales, refunds and fees reach the accounts every day without someone exporting them, does inventory reach the accounts at its full cost, and who reconciles when the two disagree.

Build. Scripts commonly send daily summaries or files into the accounting software. When one fails quietly, a day of sales is missing, and nobody finds out until the payouts are matched to bank deposits at month end. Homemade inventory sheets also tend to carry the factory price only. Inventory is carried at cost, including the freight, duty and other costs it took to bring stock to where it sells (what landed cost includes), so a sheet without those costs understates inventory and cost of goods sold, and the freight is then expensed when billed, on another line. That overstates gross margin and charges freight to the month it was paid rather than the month the goods sold. Ask your CPA how they want freight and duty allocated before anyone writes the code.

Buy. A packaged connector posts sales and fees by mapping each type to an account in your chart of accounts, set once at setup. The breaks tend to come at bundles, partial refunds and payouts that net several things together; the Shopify ERP guide lists what to test before you sign, and Shopify bookkeeping walks through the payout arithmetic.

Configure. If the platform includes the general ledger, there is no handoff to break, but the configuration decides how cost of goods and inventory are recorded, so have your CPA review those settings before go-live. If it does not, the partner builds or installs a connector and the same questions as buying apply.

Wholesale adds one more point in every option. Deductions and allowances the store's agreement allows, and any the company expects to accept anyway, come off the sale as an estimate when the invoice is booked and are adjusted as they settle. Deductions the company will dispute stay owed by the store. A system that records only the invoice amount overstates revenue until the deductions arrive.

In all three options a person still reconciles; the month-end close checklist shows where those hours go.

A two-question test: how unusual is the process, and who will own it

How unusual is the process, really? Write down the three processes that make the company money and ask whether a standard product could run each one if people changed their habits, rather than if someone changed the code. Most order, stock and purchasing processes at this size pass. The ones that do not are usually specific: made-to-order personalization, a subscription with customer-built bundles, consignment stock sitting in stores. The test is a step that no standard product can record.

Will anyone inside own it? Ownership means a named person, hours in their week for the system, and a record that survives their departure. A CEO who will look at it when there is time is not an owner.

Someone inside will own itNobody will own it
Ordinary processBuy, or keep the spreadsheet while it holds. Configure once locations, channels and approvals multiply.Buy, and keep it close to standard.
Unusual processConfigure, or build only the unusual part if the owner is technical.Fix ownership first. A build here becomes a system only an outsider understands.

The bottom-right corner is where the costly stories start: a business process nobody else runs, written in code nobody inside can read. Simplifying the process is often cheaper than building for it.

Before choosing, run your own version of the cost table, add the lines for the change you expect next year, and put a name beside the owner row. If the owner row has no name, that answers the question for you: the people around the software cost more than the software itself, so start with the person.

FAQ

What is the difference between configuring and customizing software?

Configuring sets up a platform with options it already offers: fields, workflows, approval steps, locations and account mappings. Customizing adds new code on top of the platform. Configuration usually survives the platform's upgrades; custom code has to be tested, and sometimes rewritten, each time the platform or a connected channel changes.

Do AI coding tools make building operations software cheaper?

AI coding tools make the first version cheaper and faster to produce. They do not remove the later costs: someone still has to understand the code, update it when Shopify, Amazon or eBay change their interfaces, fix it when a job fails quietly, and reconcile its numbers to the accounts. In the example above, they are more than four fifths of the build's three-year cost.

See the whole platform

Reading it is one thing. Running it is another.

Book a demo for an overview of the whole platform.

Margin Matters

The ecommerce news, read for your margin

Twice a month: the changes that touch your P&L, the money behind each one, and one move to make. Three minutes, then back to work.

Twice a month. Unsubscribe anytime.