Phased vs Big-Bang Rollout: Where the Seam Goes
Phasing is the safe answer and it is not free — you pay in a longer period of running two ways of working. Where the line actually sits, which modules genuinely have to move together, and the one situation where switching everything on at once is the correct call.
The received wisdom is to phase, and the received wisdom is right about eighty per cent of the time. But it is repeated as though phasing had no cost, and it has a real one: for the length of the phase, part of your business runs on the new system and part on the old, and someone has to carry information across the seam by hand. That seam is where the errors live.
So the question is not "phase or not". It is where you are allowed to put the seam.
Some things cannot be separated, and the list is shorter than vendors imply
Modules are sold as independent and are not equally independent. Three pairs genuinely have to move together, because a seam between them creates double entry rather than a boundary.
What must move together, and what can wait
Pair Verdict
Inventory + goods receiving Inseparable
Receiving is how stock arrives. Running inventory here and receiving on paper means every delivery is entered twice, and the second entry is the one that gets forgotten.
POS + inventory Inseparable
A till that does not deplete stock is a cash register. If you are going live on POS, inventory is already live or you are running two stock records — one of which is now wrong by every sale.
Purchase orders + budgets Inseparable if commitment visibility is the point
The whole value of a PO record is that it moves the committed figure. Orders here and budgets in a spreadsheet gives you a filing system.
Procurement + inventory Separable, briefly
You can run requests and approvals before receiving is live, and many businesses should — the approval control is worth having on its own. Do not leave the seam open for months.
Sales invoicing + accounting Separable
Invoices post here and your accountant works from the reports. This seam is permanent by design in this system, not a phase.
HR/payroll + everything else Fully separable
Payroll shares almost nothing operationally. It is the cleanest module to move on its own timetable, in either direction.
Assets + inventory Separable
Assets come from stock through a conversion when they come from stock at all. Most asset registers start as their own project and that is fine.
Where a phased rollout should put its seams
How much of the business each module disturbs
Asset register
One or two people, no daily rhythm to break. A good first project for a nervous organisation, though it delivers less.
Inventory
The store, the storekeeper, a count. A bad start is visible within a week and the damage is a bad count. The default first module.
Point of sale
Contained in scope but unforgiving in tempo — a till that hesitates in front of a queue gets abandoned by lunchtime.
Procurement
Touches requesters, approvers, suppliers, receiving and finance simultaneously. Usually the most broken process and therefore a tempting first move. Resist.
HR and payroll
Separable from everything else, but unforgiving in a different way: payroll is either right on the 28th or it is a crisis.
Accounting reporting
Depends on everything upstream being real. Goes last, always, because it inherits every other module's quality.
Read the positions as "how many people find out if you get it wrong, and how fast". Contained modules give you a cheap rehearsal of your own change-management ability before you spend it on something systemic.
The two rollout shapes, honestly compared
Phased — one or two modules at a time
- Problems have one plausible cause, so you fix them in days rather than weeks.
- Staff change one habit at a time, which is roughly the rate humans manage.
- Early value arrives while enthusiasm is still available to spend.
- Cost: a seam to carry by hand, for months rather than weeks.
- Cost: the project has no obvious end, so it can quietly become permanent.
- Cost: two systems means two sets of user admin and two places to look.
Big bang — everything on one date
- One cut-over, one date, one set of announcements. No seam at all.
- The old system stops decisively, which removes the biggest failure mode.
- Reporting is complete from day one rather than partial for two quarters.
- Cost: every problem in month one has six candidate causes.
- Cost: every member of staff changes everything simultaneously.
- Cost: requires configuration quality across six modules before you have learned anything about any of them.
When big bang is actually right
Three situations, and they are more common than the standard advice admits.
-
You are starting from nothing
A new branch, a new entity, a business under two years old with no entrenched process. There is no old system to seam against and no habit to break, so phasing buys you nothing and costs you a partial rollout. Switch it all on.
-
The old system is being switched off on a date you do not control
A licence expiring, a provider closing, a machine dying. When the stop date is imposed, phasing is not available and the question becomes how well you can prepare for one cut-over.
-
You are very small — under about fifteen staff
The whole business fits in one room and one conversation. The coordination cost that makes big bang dangerous barely exists, and phasing mainly means the same five people learn five things over four months instead of one weekend.
The honest test for a big bang
Can you name, for every module, the person who will decide its ambiguous cases in week one? Six modules needs six answers, and they can be the same two people in a small business. If you cannot fill in the list, you do not have a big-bang team — you have six phases scheduled on the same day.
The seam discipline, if you phase
Whatever seam you choose, write down four things about it and put them somewhere visible. This is fifteen minutes of work that prevents the most common phased-rollout failure, which is not that the seam existed but that nobody agreed what crossed it.
The seam agreement
- What information crosses, named specifically — "the delivery note quantity" rather than "receiving data".
- Who carries it, by name, and how often. Daily is usually right; weekly is where errors accumulate quietly.
- Which side is authoritative when they disagree. Decide now, in the calm, not during the argument.
- When the seam closes — a date, in the plan, with the next module named against it.
A two-phase rollout for a 40-staff distributor
Note that phase 1 deliberately runs receiving against a paper order rather than waiting for procurement. That is the seam earning its place: the storekeeper gets the discipline of receiving against something, and the something upgrades later without changing their habit.
Our take
Phase, starting with inventory, unless you are brand new, under fifteen people, or facing an imposed stop date — in which case a well-prepared big bang is the better risk. Whichever you pick, the decisive artefact is not the plan but the seam agreement: what crosses, who carries it, which side wins a disagreement, and the date it closes.
Design the seam, not just the sequence
Tell us your [headcount](/glossary/headcount), your locations and what is most broken, and we will sketch the phases and the seam agreement for each one — including the phases we would not attempt in your first quarter.
See plans & pricingFrequently asked questions
Should we phase our rollout or go live on everything at once?
Phase, in about eighty per cent of cases, starting with inventory. Go big bang if you are a brand-new entity with no entrenched process, if you are under roughly fifteen staff, or if your old system is being switched off on a date you do not control. Those three cases are more common than the standard advice admits.
Which modules cannot be separated?
Inventory and goods receiving; POS and inventory; and purchase orders and budgets if commitment visibility is the reason you are buying. Each of those pairs creates double entry rather than a clean boundary if you split it. Procurement can run briefly ahead of receiving, sales invoicing is permanently separate from statutory accounting here by design, and HR/payroll shares almost nothing operationally so it can move on its own timetable.
What is the real cost of phasing?
A seam that someone carries by hand for months instead of weeks, two places to look for information, two sets of user administration, and a project with no obvious end — which is how phased rollouts quietly become permanently half-done. These are real costs; they are just usually smaller than a month in which nothing works.
What is a seam agreement?
Four written statements about the boundary between the live module and the old process: exactly what information crosses, who carries it and how often, which side is authoritative when they disagree, and the date the seam closes with the next module named against it. Fifteen minutes of work. The common phased-rollout failure is not that a seam existed but that nobody agreed what crossed it.
Why does accounting go last?
Because it inherits the quality of everything upstream. Reports built on a stock figure nobody trusts and purchase orders that only half exist will be accurate reflections of bad inputs, and the conclusion drawn will be that the reporting is wrong. Get the operational records real first; the reporting then becomes almost free.
How do we know a phase is finished?
When it is boring. Specifically: no handwritten parallel record, no weekly reconciliation meeting about that module, and the people using it have stopped asking questions about how rather than what. If any of those three is still true, adding the next module means adding a second unfinished thing.
