Projects & remediation
A defined project, with the dependencies named up front.
Migrations, office moves, end-of-support upgrades and infrastructure remediation — scoped with explicit acceptance criteria, a tested way back, and a documented handover. No ongoing contract required.
- Project onlyOne defined piece of work is a complete engagement, not a trial.
- Dependencies firstMost overruns are licence, approval or access problems, not engineering.
- Tested rollbackA cutover without a way back is a one-way door.
- Documented handoverIt ends with a runbook, not a verbal summary.
When this is the right engagement
Typical triggers
- A forced move. An office relocation, a lease ending, or a landlord reclaiming the comms room.
- An end-of-support date. A server OS, a hypervisor version, or an application vendor dropping support for the platform underneath it.
- A stalled project. Someone started it and left, or an internal attempt ran out of capacity. Common, and not treated as an embarrassment.
- A finding with a date on it. An assessment, an insurer or a customer questionnaire produced a list that has to be closed.
- Recovery improvements. Backup and recovery arrangements that need rebuilding rather than patching.
Boundaries
- Ongoing operation after handover is not included — handing back to your own team is a normal, expected ending.
- Hardware and licence costs are yours.
- Application development is out of scope.
- Price, duration and completion date come from scoping against your actual environment.
If you want it operated afterwards, that is managed or co-managed IT — a separate decision, made later.
Stages, each with a gate
A gate is a point where the work stops until someone decides. Naming them in advance is what stops a project drifting.
-
DefineGate: scope agreed
The outcome in one sentence, an explicit out-of-scope list, acceptance criteria, and who signs off. If the outcome will not fit in a sentence, the project is really two projects.
-
DiscoverGate: dependencies clear
Current-state inventory, dependency map, licence and contract position, and the constraints nobody mentioned. This stage frequently changes the plan, which is the cheapest time for the plan to change.
-
DesignGate: rollback defined
Documented target state, the cutover plan, the rollback trigger and its named owner, the risk register, and agreed change windows. The gate is not passed while rollback is still theoretical.
-
DeliverGate: per stage
Staged implementation with a checkpoint at each step, rollback tested rather than assumed, and progress reporting a non-technical sponsor can actually read.
-
Hand overGate: acceptance
Documentation pack, operational runbook, and acceptance measured against the criteria written in Define — not against how everyone feels on the day. Then managed, co-managed, or back to your team.
The cutover, and who decides
The point of naming a gate in advance is that the middle of a cutover, at an hour nobody enjoys, is the worst time to work out who has the authority to stop.
-
Window opensChange approved
The change window was agreed in Design, not negotiated on the night. Everyone who needs to know already does.
-
Pre-checksBackup verified
The rollback position is confirmed before anything changes. If the backup cannot be relied on, the window closes here and nothing has been touched.
-
Switch overStaged and verified
Done in stages with a check after each, so a problem is found at the step that caused it rather than at the end.
-
Go or no-goDecision gate — named owner
A person agreed in advance decides whether the cutover stands. Not the engineer mid-task, and not by default because it is late.
-
Accept, or roll backAgainst written criteria
Acceptance is measured against what was written down in Define. If the gate is not passed, the tested rollback runs and the window closes cleanly — the difference between a cutover and a one-way door.
What has to be true for each stage to finish
Projects rarely fail on engineering. They stall on something nobody checked.
- CommercialCheck this first
-
- Budget approved
- Licences owned
- Vendor contracts current
The most common cause of a stalled project and the cheapest thing to verify. A licence tied to a physical host id has stopped more migrations than any technical fault.
- PeopleNamed, not assumed
-
- Decision-maker available
- Change windows agreed
- Users told what changes
A migration with no agreed interruption window is not yet schedulable, however complete the technical plan is.
- DataVerified
-
- Current backup
- Restore actually tested
- Volumes measured
No infrastructure change starts against an untested backup, because the backup is the rollback. Where the restore does not work, fixing it becomes the first piece of the project.
- PlatformMeasured
-
- Capacity headroom
- Supported versions
- Network path proven
Measured rather than assumed. An unsupported version discovered late rewrites the design and the schedule together.
- RollbackOwner named
-
- Defined trigger
- Tested procedure
- Decision owner
“We would roll back” is not a rollback plan. A trigger, a tested procedure and a named person who can call it are.
What discovery is actually for
A firm has nine weeks to vacate a comms room because the landlord is reclaiming the space. Discovery finds that one line-of-business application is licensed against a physical host identifier — something that would have surfaced on cutover weekend under the original plan. The licence question moves to week two instead of week eight, the vendor re-issues against the new host, and the move happens on schedule.
The project did not get easier. The surprise moved earlier, which is the entire value of a discovery stage.
Illustrative scenario, written to show how the work is sequenced. Not a client account.
Questions before a project starts
Can I engage Honed IT for one project only?
Yes. Standalone assessments, migrations, remediation and other defined projects are welcome, and an ongoing managed-services arrangement is not required before, during or after.
What happens when you find something nobody knew about?
It is raised as a written change, priced or deferred by you. It is never absorbed silently — that is how both the schedule and the trust disappear at the same time.
How much interruption will this need?
It depends on the work and is established during Design, then agreed as a change window before anything is scheduled. No generic downtime figure is published, because one would be meaningless against an environment nobody has looked at yet.
What is the rollback plan, and who decides to use it?
Rollback has a defined trigger, a tested procedure and a named owner before the Design gate is passed. The decision to invoke it sits with your sponsor, not with the engineer in the middle of a cutover.
Can you finish a project someone else started?
Frequently. It begins with a current-state review rather than an assumption that the previous plan was right, because inheriting someone else’s half-finished design without checking it is how a second attempt fails too.
What do we get at the end?
A documentation pack describing what was built, an operational runbook for the people who have to live with it, and acceptance measured against the criteria agreed in the Define stage.
Will you work with our internal team or current provider?
Yes, and it is common. Where an internal team is involved, responsibilities for the project are written down the same way they are in a co-managed arrangement.
What has to be done, and by when?
Those two answers are enough to start scoping. New Jersey and the NYC metro area for work that needs hands on site, with nationwide remote services for work that does not.
Related: data center hosting · hosted virtual servers · managed IT