Build vs buy software: when a build pays off

A subscription and a custom build are priced in different currencies, and most comparisons never split them. If you run a business in Bangladesh, the build-vs-buy software decision is measured in taka you cannot get back. We sell builds, so one of the two prices on your desk comes from a vendor with an interest in the answer, and this page is worth nothing unless it says plainly when to buy. No Scaledex client work appears below; the named cases are public and in the third person, the worked example is a hypothetical we label as one, and the judgement is ours.

Below, we walk through the two ledgers behind every option, a three-year worked example, what the buy column costs here, the two questions that decide it, and the sheet to bring to an agency.

Custom software vs off the shelf: two ledgers, not one price

Every option in front of you is two ledgers, not one price. A family that pays its rent in dollars and earns in taka never asks whether the rent is high; it asks what the rent does when the dollar moves. Software works similarly.

The taka ledger holds an agency build, local hosting, a local vendor’s licence and your own staff’s time; the dollar ledger holds the subscription seats, the cloud account, the gateway fees and any paid API. The split matters because a dollar cost about 85 taka in 2021, and by the end of 2024 Bangladesh Bank had fixed it at 122, 43 % more taka for the same dollar. A plan whose sticker never changed cost its Bangladeshi buyer close to half again as much. Split the ledgers before comparing the options, because one is set by the vendor and the other by the central bank.

The first page for this query is written mostly by enterprise vendors, in one shape: a pros-and-cons table, a factor list and a matrix, all in dollars, with no payment method or regulator named. None of them has a second currency.

What is the difference between building and buying software?

Buying is renting a solution to the average customer’s problem; building is paying for a solution to yours, and then paying to keep it alive. For a reader without an engineering team, “build” means commissioning an agency like us, and owning both the code and the maintenance bill afterwards.

What is a make vs buy decision?

It is the same question asked about a factory component, and manufacturers settled it long ago: make what customers pay you for, buy the rest. The way to see the difference is to price one case, buy and build, over three years.

Build vs buy software cost: a worked example over three years

Take a hypothetical supplier, Tanvir, who sells zips, buttons and woven labels to garment factories from Chattogram, with six people who need to see stock and orders. He has a build quote and an inventory subscription, and wants both priced over three years.

The subscription is priced per seat, at $20 a seat a month. Dollar figures use ৳123 to the dollar, rounded from Bangladesh Bank’s interbank weighted average of 25 August 2026. Six seats for three years come to about ৳6.1 lakh, all of it dollar ledger, including the VAT his card will carry until his company has a BIN. A one-off developer fee to wire bKash and Bangla invoices brings the buy column to about ৳6.9 lakh.

The build column starts at ৳12 lakh, inside the band our cost post gives for an integrated portal, and adds maintenance at ৳20,000 a month and hosting, the build’s one dollar line. Over three years that is about ৳20.3 lakh. The figure below puts both columns side by side, with the dollar lines hatched.

For six seats the subscription costs a third of the build, and almost all of it sits in the dollar ledger. Stress that ledger by the last three-year currency move and buy still wins; the split tells you who sets which price, not which option is cheaper.

Buy wins for six seats by a factor of three, and almost nine-tenths of what Tanvir would pay is in dollars. If the taka repeats its 2021–24 move, his subscription rises by close to half while the build’s taka lines stay where they are, and the subscription is still cheaper. We’d tell Tanvir to buy.

Putting it simply: the split does not make building cheaper; it tells you which of your costs the central bank sets. It also shows why his buy column has lines no US article prices.

What the buy column costs in Bangladesh

The buy column here carries three lines: VAT, paying in dollars at all, and the developer who wires bKash.

The first is VAT. Google applies a 15 % VAT to Google Ads customers in Bangladesh who have not added a Business Identification Number to their payments profile. The tax isn’t a reason to build. It is a line in the buy column and a form to fill, because a registered company that enters its BIN is not charged it.

The second is paying in dollars at all. Bangladesh Bank’s master circular on outward remittances of September 2025 lets a bank remit software maintenance and support fees, and cloud usage fees, against a legitimate licence, a signed agreement and invoices. An international card is capped at $300 an online transaction, and over a year at the card-holder’s unused travel quota plus $1,000. A $20 seat is easy; a $2,000 platform is paperwork your bank will ask for.

The third is the extend cost nobody prices. An international product takes cards; bKash, Nagad and Rocket arrive through a gateway or a developer. bKash publishes its own WooCommerce plugin, but for most products we would budget a developer’s week to take a bKash payment or print a Bangla invoice. Buy here is usually buy and extend, before it is anywhere else.

There is a third column, and the search phrase build vs buy vs partner already names it.

Build, buy or partner: the local vendor

The third option is a local product invoiced in taka, taking bKash and answering the phone in Bangla. TallyKhata is the public case: a Bangla-language ledger app from Progoti Systems, which by The Daily Star’s account in 2024 had passed a million monthly active users. Nobody should commission a build for a shop’s khata while that exists, any more than for the Facebook page you run until it breaks, and neither carries a dollar ledger.

The risk is the vendor’s size. A small company can fold or stop shipping, and your data goes with it, so ask for four clauses: data export in an open format on demand, source code in escrow or handed over if the vendor stops, a cap on price rises, and accounts and domain in your name. They are the clauses we would put in our own build contract, and a vendor who refuses them has told you something. The grid below puts the three columns beside each other.

Building does not remove dependence; it moves it and prices most of it in taka. Buying puts three of four lines in dollars; partnering puts none there and everything on the exit clauses.

Every column has a dependence; only the currency and the exit differ, and no matrix will pick between them.

The two questions that decide it, and the triggers

Two questions decide the column, and neither is about features. First: would a customer pay more, or switch to you, because of how this software behaves? For stock, payroll, bookings and invoicing the honest answer is almost always no, and no means buy or partner.

Second: who fixes it at two in the morning, and what does that person cost per year in taka? A build’s maintenance line is the operating cost our website cost post itemises; if you cannot name the person and the line, you cannot afford to build, whatever the quote says.

The triggers we would watch for are four:

per-seat pricing at 40 staff;
a Bangla-language or taka-invoicing gap the vendor will not close;
a report the product cannot produce for the bank or the NBR;
an integration the vendor will not build.

The figure below runs Tanvir’s two columns across seat counts.

Under about 20 seats the subscription is cheaper over three years; at 40 it costs twice the build. The build line is flat because seats are free once it exists.

The lines cross at about 20 seats over three years; at 40, the subscription costs about twice the build. Headcount, not ambition, moves the answer, and neither question fixes the problem that comes before both.

What the decision does not fix

A build-vs-buy decision does not define the problem it is meant to solve. Mike Hyzy of CGI put it fairly in Forbes in December 2025: the build-versus-buy debate presupposes a clear understanding of what you are actually building or buying. The scope document is where that understanding gets written down.

Three more limits. Building moves dependence, from the vendor to the agency, the cloud account and the gateway, rather than removing it. Buying still needs one owner inside the company. And AI-assisted development lowers the cost of the first version, not of the years after it: the maintenance line is people on call.

The last limit can settle the question by itself. Bangladesh’s Personal Data Protection Act 2026, which in April replaced the previous November’s ordinance, lets personal data leave the country with the person’s consent or under a contract for goods or services that the person is party to, but only to places with, in the Act’s words, “suitable technology and equipment”. That test is left to regulations the Authority has yet to publish. Bulk transfers of national ID, passport, tax or biometric data must be notified to the Authority.

The ordinance’s rule that restricted data keep a synchronised copy in Bangladesh is not in the Act. In plain terms: if your product would hold customers’ NIDs in a foreign SaaS, ask where the vendor’s servers sit before you read the price list, because the law and not the ledger may decide. What remains is to put all of this on one page that a vendor, including us, has to answer.

What to bring to the agency

A decision that arrives on one page shortens vendor selection, because the scope document consumes exactly this shape. Here is the sheet we would hand Tanvir.

The two questions. Would a customer pay more, or switch to you, because of how this software behaves? Who fixes it at two in the morning, and at what cost per year in taka?

The two ledgers, three years. Every line marked taka or dollar: seats or licence, hosting, gateway and API fees, maintenance, wiring, staff time; the VAT on the dollar lines, then stress the dollar lines by the last three-year move.

The triggers. Per-seat price times the headcount you expect in two years; Bangla or taka invoicing; the report the bank or NBR needs; the integration the vendor will not build; where the Act says the data must sit.

The four clauses, for any vendor including us. Data export in an open format on demand; source in escrow or handed over; a cap on price rises; accounts and domain in your name.

Next from us: a printable version of this sheet. If a subscription broke for you in a way we did not name, tell us. The first step is this evening’s: write your buy column for three years in two currencies, with the VAT on the dollar lines.

If the sheet says build, our web development team scopes it; if it says buy or partner, vendor and platform selection ends in a written recommendation. Fill it in yourself first, and you will hold a page that a vendor with an interest in the answer cannot argue with.