Systems and integration
Replace or integrate your business software? How to decide
Whether to replace or integrate business software depends on where the problem sits. Replace when the system itself is the problem: no support or security updates, information it cannot hold, every route in and out closed, or a licence dearer than a modern equivalent. If the problem is the manual work around it, such as re-keying, exporting and chasing, integration is usually cheaper, faster and far less disruptive.
The quotation has been on Linda’s desk since Thursday. Eleven pages, a whole new system to run the business, from a salesman who was pleasant and very sure. Linda owns the firm and employs 40 people. Through the glass she can see the order desk, where her one order processor is typing an emailed order into Sage 50 with a full inbox still to go. Linda has been trying to hire a second processor since the spring, and last month she signed off the overtime again. The salesman says the old software is the problem. He may be right. But she has read the quote twice, and nowhere does it say who types the orders in once the new system arrives.
Linda is made up. The quotation is not.
Key facts
- The question is rarely “is the software old”. It is whether the cost sits in the system or in the manual work around it.
- UK SMEs choosing technology rank ease of use first (37%), cost second (17%) and compatibility with existing systems third (15%), according to the Department for Business and Trade’s 2025 survey.
- The government’s SME Digital Adoption Taskforce reported in 2025 that for small firms “switching feels too high risk”, and the Made Smarter evaluation found businesses expected frictional costs from integrating new technology with existing systems.
- Integrating one job, such as orders from email into your accounts system, costs £1,500 to £7,500 at Better Systems Co.’s published prices and keeps the screens your team knows.
- Five tests decide it: support, data, access, cost and the process itself.
Why do so many businesses assume they have to replace?
Because that is what most of the market sells. Software firms sell new software. Many consultancies are paid to install it. So an owner like Linda, with a ten-year-old order system and a tired office team, hears one answer from everyone she asks, and it is the expensive one.
The salesman called his product an ERP, which is one system meant to run orders, stock and accounts together. The other choice is to integrate, which is to connect the software you already have, so information moves between programs without a person typing it again.
The evidence says owners are right to be wary. In the Department for Business and Trade’s survey of 1,001 UK SMEs, the top factors in choosing a technology provider were ease of use (37%), cost (17%) and compatibility with existing systems (15%) (DBT and Ipsos, July 2025). The SME Digital Adoption Taskforce put it more bluntly, citing Be the Business: technology products “are often built for enterprise, not SME customers”, adoption “looks too hard and costly” and “switching feels too high risk” (Taskforce final report, July 2025).
Those are not reasons never to replace. They are reasons to test whether you need to before you do.
What are the five tests?
Run each one against the system you are worried about. The answers decide the direction.
| Test | Points to replacement when | Points to integration when |
|---|---|---|
| Support | The maker has ended support and security updates, or has gone, and nobody can fix a fault | The maker supports it, or it runs steadily and a competent engineer can still maintain it |
| Data | It cannot record what the business now needs: batch numbers, customer-specific prices, multiple warehouses, a second currency | It holds the information; the problem is getting information into and out of it |
| Access | Every route in and out is closed: no import, no export, no readable database, no connection tool from the maker, and the owner refuses screen-level automation | At least one route is open and the owner will approve it in writing |
| Cost | The licence and the hardware it needs cost more each year than a modern equivalent, including the move | The licence is modest and the real cost is staff time around it |
| Process | The work is slow because the system is slow: it crashes, locks or takes minutes per screen | The work is slow because people type, copy, export and chase between systems |
If three or more tests point to replacement, start planning one. If the problem is mostly the fifth row, which it usually is, integrate.
Suppose Linda’s answers come out like this. Her Sage 50 is still supported and holds her customers, products, prices and orders. There are approved routes into it. The licence is small next to the overtime she signs off. And the work is slow because a person types every order, not because Sage is slow. None of the five points at a new system. Yours may differ, and if three go the other way the salesman has a point.
How do you tell a process problem from a system problem?
Take the three jobs that eat the most office time. For each, ask: if the information were already inside the system, could the system do the job? An emailed order that a person types into Sage is a process problem. Sage can hold the order; the cost is the typing and the checking. A quote that needs a price the system has nowhere to keep is a system problem. Our post on how to stop re-keying customer orders works through the first kind in detail.
What does each option actually cost?
A worked example, with every assumption labelled. It is not a client result, and Linda’s firm is not a real one.
Say Linda’s firm runs on Sage 50 and takes 300 orders a week by email and PDF, each taking 4 minutes to type in and check. Using our published formula, that is 300 × 4 ÷ 60 × 48 working weeks, or 960 hours a year. At a fully loaded £20 an hour, meaning wages plus employment costs, that is £19,200 of paid time a year. Hull sales order processor roles were advertised at £12.21 to £15 an hour in 2025, so £20 fully loaded is a reasonable anchor, not a measured figure. It is time that could go on other work, not cash off the wage bill.
Integrate. Software that reads each emailed order, waits for a person to approve it and puts it into Sage 50 by a supported route sits in Better Systems Co.’s quick-win or business-workflow bands: £1,500 to £7,500, one-off. Running costs are itemised and usually tens of pounds a month. If Sage Data Objects, Sage’s own connection tool, is the right way to put the orders in, add Sage’s Developer Programme licence, which Sage prices at £3,438 a year plus VAT from 1 August 2026. Sage stays. The team keeps its screens. Timescale three to six weeks.
Replace. Assume a mid-market cloud system of the kind in Linda’s quotation. We will not put a licence figure on it because they vary too widely and depend on the seller’s quote. What you can count is the work that does not appear on the quote: moving the data across and cleaning it up, retraining every user, rebuilding reports and document templates, running the old and new systems side by side for a period, and the orders that still arrive by email and still need typing into the new system. The Made Smarter programme’s evaluation found that businesses anticipated “frictional costs owing to the complexity of integrating” new technology with existing systems and processes, and that skills gaps “could not easily be addressed through recruitment”, with one firm taking two years to recruit an IT manager (Ipsos for DBT, June 2025).
The point of the example is the last item on that list. A new system does not read emails. If re-keying is the problem, replacement moves the typing to a new screen. Integration removes it. Linda could pay for the whole new system and still be advertising for a second order processor.
When is replacement genuinely the right answer?
Four situations come up often enough to name.
- Support has ended and the system holds money. An accounts or payroll system without security updates is a liability, not a tool. Plan the replacement, and connect the new system to the rest of the office from day one.
- The system has nowhere to put what you now record. A food business that must now record batch and allergen data in a system with no place for it will fight the system every day. See our food manufacturing page for how that shows up on the order desk.
- Every door is shut. It is rare, but a software maker can refuse every route, including import files. Our guide to integrating a legacy system with no API lists the four routes; if all four are refused in writing, you have your answer.
- The licence is the cost. Some older systems carry per-user or per-site fees, plus a server, that exceed a modern subscription. Do the sum over three years, including the move, before deciding.
Note what is not on the list: “it looks dated”, “the vendor wants us on the cloud version” and “a consultant said so”.
What should Linda ask before she signs?
Four questions, in writing.
- Will the new system read the orders that arrive by email, or will someone still type them in?
- What is not in the quotation? Ask about moving the data, retraining, rebuilding reports and running both systems for a while.
- When does support for our current system end, and what stops working on that date?
- Which of our three slowest jobs does the new system remove, and which does it only move to a new screen?
If the answers leave the typing where it is, the quotation solves a problem Linda does not have.
What happens to the integration if you replace later?
Most of it survives, which is why integrate now, replace later is a reasonable plan. An order reader reads the email and attachment, checks the result against your customers, products and rules, and puts the approved order into your system. The first two do not care what the system is. Only the last step changes, and on a modern system that step is usually simpler because there is a public API, a ready-made way for one program to talk to another. Xero, QuickBooks Online, Business Central and Sage 200 all have one, though Sage 200’s needs its permission renewed from time to time and cannot run fully unattended (Knit).
You also arrive at the replacement knowing your real volumes, your real rules and how many orders needed a person, because the integration has been counting them. That knowledge is what makes a new system cheap to set up and hard to get wrong. So Linda need not tear the quotation up. She can put it in a drawer until she knows her own numbers.
What does Linda’s office look like afterwards?
This is still an illustration. Linda tells the salesman “not yet” and has an order reader connected to the Sage 50 she already owns. A few weeks on, her order processor starts the morning with the overnight orders already read and waiting for approval. She checks them, sorts out the odd ones and is on the phone to customers by mid-morning. At month end Linda looks at the overtime sheet to see whether the difference shows there.
A firm that connects what it has is working the quicker way within weeks. A firm that waits for a new system keeps typing while it chooses, moves and retrains, and its competitors do not wait with it.
What we would do
We would start with an Automation Opportunity Audit at £500 to £1,250, credited against any build. The report runs the five tests on each system involved, records what each one can take in and send out, tells you in pounds what the manual work costs, and states plainly whether replacement is warranted. Where it is not, we propose the smallest integration that removes the manual work, built as our connecting existing systems or sales order processing automation service. Put your own volumes into the ROI calculator first; if the hours are small, neither option is worth paying for yet. If the complaint is too many logins rather than too much typing, the smallest fix may be one screen for all your systems, with nothing replaced. Terms such as API, ODBC and SDK are defined in the glossary.
Sources
- Department for Business and Trade and Ipsos, Understanding technology adoption among UK SMEs, July 2025: https://assets.publishing.service.gov.uk/media/688a438aff8c05468cb7b0f0/sme_tech_adoption_dbt_report.pdf
- SME Digital Adoption Taskforce, final report, July 2025: https://assets.publishing.service.gov.uk/media/688a43d0b223ff124d388903/sme-digital-adoption-taskforce-final-report.pdf
- Ipsos for DBT, Made Smarter Adoption Programme final impact evaluation, June 2025: https://assets.publishing.service.gov.uk/media/6a69b8434dda8076771a42a8/made-smarter-adoption-programme-final-impact-evaluation-june-2025.pdf
- Hyperext, Sage 50 integration: https://www.hyperext.com/sage-50-integration/
- Knit, developer guide to building a Sage 200 API integration: https://getknit.dev/blog/deep-dive-developer-guide-to-building-a-sage-200-api-integration/
- Sage, Sage 50 and Sage 200 UKI Developer Program UK pricing, July 2026: https://downloads.sage.co.uk/download/?did=cd53b8a9-506f-4d67-8ef7-4a016d7a8197
Questions people ask
Our software vendor says we must upgrade to their cloud version. Does that force the decision?
Not on its own. Ask for the end-of-support date in writing, what stops working on that date, and whether the version installed on your own machines keeps running without updates. Many do. That gives you a deadline to plan around, not a decision to make this quarter. If support really is ending, replacement moves up the list, but the integration work you do now usually carries over because it sits outside the system.
Can we integrate now and replace later?
Yes, and it is often the better order. Software that reads orders from email and puts them into your current system is built around that system, not inside it. When you do replace it, the reading, checking and approval parts stay and only the last step, putting the order in, is redone. You also arrive at the replacement project knowing your real volumes and rules, which makes the new system cheaper to set up.
How do we know whether the problem is the system or the process?
Write down the three tasks that cost the most time and ask, for each, whether the system can do it if the information were already inside it. If Sage can hold the order but a person has to type it, that is a process problem and integration fixes it. If the system cannot record what you need, such as batch numbers or customer-specific pricing, that is a system limit and may justify replacement.
What does Better Systems Co. gain from recommending integration?
Integration is what we build, so you should weigh that. Our answer is that the audit report says replace when the tests point that way, and we have no replacement product to sell you. The audit fee is fixed and credited against any build, so the recommendation does not change what we are paid for the audit.
Related guides
- What manual order entry really costs a UK wholesaler, and how to work out the ROI of automating it
- Still typing the PO and then typing the whole order again? How to stop re-keying customer orders
- How to integrate a legacy system that has no API, without replacing it
- Sage 50 has no API. Here are the four ways to connect it to everything else