Software that runs an operation cannot stop when the office closes
The commercial life of this city is built around a working port, marine and maritime services, freight and logistics, cruise and passenger operations, plus a large healthcare and university presence. What those have in common is that they do not run on office hours. Ships arrive at night, shifts hand over at six, and a system that is unavailable until Monday is not an inconvenience, it is an operational failure.
That single fact should reorganise how you evaluate a development partner. Most software buying advice concentrates on the build. For an operational system the build is the short part, and the thing you are really purchasing is a service that has to keep working for years while people who did not choose it depend on it during a night shift. Judge suppliers accordingly, and be suspicious of a proposal that describes features in detail and operations in a sentence.
Hardware, sensors and the integrations nobody quotes for
Operational software in this market is rarely self contained. It reads from weighbridges, scanners, gate systems, tracking units, vessel and berth schedules, telematics on vehicles, badge readers, or equipment that speaks an industrial protocol designed long before anything web shaped existed. Each of those is a potential project of its own.
Establish for every connected thing who owns it, whether a documented interface exists, whether the vendor will co-operate and at what cost, what happens when the device is offline, and how much data arrives at peak. Then ask a specific question of any candidate supplier: describe an integration with physical equipment that went wrong and how you recovered it. Firms that have only built browser applications will answer in generalities. Firms that have stood in a yard with a laptop debugging a serial connection will answer with a story, and that is the answer you want.
Connectivity deserves its own line. Anywhere with steel, water or large structures has dead spots, so any tool used away from a desk needs genuine offline capability, a synchronisation model that survives a shift without signal, and a defined rule for what happens when two people edited the same record.
Service levels are the part of the contract you will actually use
Most buyers negotiate the build in detail and the support arrangement in a paragraph, then live inside that paragraph for years. Invert the effort. Define severity levels in terms of business impact rather than technical symptoms, so that operations halted, operations degraded and cosmetic defect are the categories rather than server errors and user complaints.
Against each, agree a response time and a restoration expectation, and be explicit that they are different things. Agree the hours those commitments apply, and if your operation runs overnight, agree what happens at three in the morning and who is paid to be reachable. Agree how an incident is raised out of hours, which should be a phone number rather than a mailbox nobody reads until the next working day. Agree who is responsible for monitoring, because a supplier who waits for you to report failures is not providing a service, they are providing a repair shop. And agree what happens when the commitments are missed, even if the remedy is modest, because a service level with no consequence attached is a description rather than an obligation.
Hosting, uptime and who holds the keys
Decide early whether the software will run in infrastructure you own or in the supplier's. Neither is wrong, but the second creates a dependency that must be documented while everyone is still friendly. If the supplier hosts, establish whose cloud account it is, whether you have administrative access, how billing works and exactly what a transfer would involve. If domains, certificates or third party subscriptions sit in the supplier's name, write down how they move.
Then ask the questions that only matter after something has gone wrong. How often are backups taken, where are they stored, and when did anyone last restore one to prove it works. What is the recovery plan if the hosting region fails, and how long would it take. Who can deploy to production, and is there a record of who deployed what and when. An untested backup is not a backup, and a supplier who can describe their last restore test is a different proposition from one who describes their backup policy.
Buying in phases when the operation cannot pause
Replacing software that people depend on daily is not a launch, it is a transition, and it has to be planned as one. Run the new alongside the old for a period, even though duplicate data entry is unpopular, because the alternative is discovering a gap during a live shift. Decide in advance what the rollback looks like and who has authority to call it. Choose a cutover moment that matches the operational calendar rather than the project plan, which usually means the quietest period rather than the end of a sprint.
Train on the exceptions rather than the standard flow, since staff will work out the normal path themselves and will be stuck on the awkward case. Budget for a stabilisation period after go live with the supplier close at hand, and write it into the contract rather than hoping for goodwill. The weeks immediately after a cutover generate more change requests than the whole build, and a contract that treats them as billable surprises starts the support relationship badly.
Choosing between software companies in Southampton
Three or four suppliers briefed identically in writing will give you a real comparison, and what you are comparing is not headline price but the assumptions and exclusions underneath it. Ask each for one page: their reading of the scope, what they have deliberately left out, the named people who would do the work, what they need from you and the largest risk they see.
Ask for a reference where the system has been live for more than a year, and speak to somebody who uses it on a shift rather than to the manager who signed. Ask that person what happens when they report a problem. Weight operational experience heavily, because a firm that has supported software running around the clock prices support realistically and a firm that has not will underprice it and then resent it.
If the requirement also reaches customers rather than only staff, splitting the work usually beats stretching one supplier. That might mean search specialists for a public facing site, social media agencies working locally for ongoing channels, or communications support around a launch. The wider set of verified development firms is listed in the software development company directory, and where a specialised skill is genuinely scarce, supplier pools in Bristol and Brighton are close enough for regular on site work. Once the scope and the service expectations are written down together, ask several verified suppliers for comparable proposals.