MSP Migrations by the Numbers: What 200 Client Moves Taught Us

Managed service providers migrate other people’s infrastructure for a living, which means the dataset is bigger and more honest than any single enterprise’s war story. This post aggregates anonymized data from 214 client migrations executed between 2021 and 2026 — SMB to mid-market, 20 to 2,000 seats.

#The Headline Numbers

Read that last row again: 1 in 16 migrations gets abandoned or rolled back permanently. In every abandoned case, the root cause traced to discovery — not execution.

#Where the Time Actually Goes

Proposals allocate effort roughly 20/50/30 across discovery, execution, and stabilization. Actuals across the dataset:

Two things jump out. Discovery always eats more than planned — the p90 case spends half the engagement just finding out what exists. And stabilization gets cannibalized to protect the go-live date, which is exactly backwards: the p90 projects that squeezed stabilization to 11% generated 3.2x more support tickets in the following quarter.

#Failure Modes, Ranked

Across the 71 projects that ran late or over budget, primary root causes:

1. Undocumented dependencies (41%) — the classic. A line-of-business app calling a server nobody inventoried.
2. Licensing surprises (19%) — SQL Server core licensing on cloud VMs remains the single most expensive line-item shock.
3. Bandwidth math (14%) — seeding 40TB over a 100Mbps circuit takes 37 days at theoretical maximum. Someone always forgets.
4. Identity sprawl (12%) — three AD forests, one Okta tenant, and a decade of conditional-access debt.
5. Change-freeze collisions (8%) — retail clients in Q4, accounting firms in tax season.
6. Other (6%)

#The Cost Curve

Per-seat migration cost falls sharply with scale, then flattens:

Seats     Median cost/seat    Note
20–50     $412                fixed costs dominate
51–200    $265                sweet spot for standardized runbooks
201–500   $198                tooling investment pays off
501–2000  $184                diminishing returns; complexity tax returns

The flattening past 500 seats is the complexity tax: bigger clients bring bespoke apps, compliance regimes, and committee approvals that scale with headcount.

#What Actually Predicts Success

We regressed outcomes against 22 project attributes. Only four moved the needle:

• A tested rollback plan (largest effect; the 6% who abandoned had none)
• Discovery tooling deployed ≥30 days pre-cutover — automated scanning, not questionnaires
• A named client-side decision-maker with authority to kill scope
• Pilot group ≥10% of seats before the main wave

Certification counts, vendor partnerships, and team size showed no significant correlation with outcomes. Process beats pedigree.

#Takeaway for MSP Operators

Price discovery as its own engagement. Clients resist paying for “just looking around” — but the data says discovery quality decides the outcome before the first byte moves. The MSPs in our dataset that split discovery into a paid phase-zero saw abandonment fall to under 2%.