Ask where your company’s data lives and the honest answer is: three different worlds. Servers (physical, virtual, or hosted) run the databases and applications. Cloud services hold the email, files, and the SaaS sprawl. And desktops and laptops carry whatever your people have been working on this week — which is always more than policy says. Each world fails differently, gets protected differently, and — here’s where strategies go wrong — tends to get protected separately, by different decisions made at different times, leaving gaps precisely at the seams. The fix is one coverage map. Here’s how to draw it.
Start with the map, not the tools
Before comparing backup products, take the asset inventory and the data-classification map and answer one question per location: what lives here that we’d need back, and how fast, losing how much? That exercise — an hour, honestly — produces the strategy’s skeleton: every data location, its tier, its recovery numbers. Every backup decision below is just implementing a row of that table, and every future “are we covered?” question gets answered by reading it. The map is also where the seam-gaps reveal themselves: the database on the server (covered) whose export job lands on a desktop (not covered); the “cloud” app that actually runs on a box in the closet; the owner’s laptop holding fifteen years of institutional memory in a folder named misc.
World one: servers — the deep end
Servers get the fullest treatment because they concentrate the highest tiers: image-based backup (the whole system — OS, applications, configuration — not just files), because rebuilding a server from scratch plus file restores blows any reasonable RTO; frequency matching the tier — nightly at minimum, hourly or continuous for the systems whose RPO math demands it; the full 3-2-1-1-0 layering — local copies for fast restores, immutable offsite copies for the bad scenarios; and, for the top tier, the ability to run the backup as a virtual machine while the real server is rebuilt — the feature that turns a multi-day RTO into an hour, and the one your downtime number usually justifies.
World two: cloud — the assumed-covered world
Everything from the Microsoft 365 conversation and the SaaS strategy applies: the vendors protect their platforms, you protect your data, so the cloud column of the map gets dedicated backup tools for the major platforms, owned exports for the long tail, and written accepted-risk for the rest. The strategic point at the map level: cloud data is not the safe column that needs no thought — it’s the column where the thinking was skipped, which is different.
World three: endpoints — the honest conversation
Desktops and laptops are where backup strategies get philosophical, because the policy answer and the reality answer diverge. Policy says: important files live on the server or in OneDrive/SharePoint, so endpoints are disposable — reimage and move on. Reality says: people work locally, Downloads folders hold contracts, and somebody’s laptop is always secretly a critical system. The strategy is a two-part honest settlement:
- Make the policy answer mostly true. Folder redirection and OneDrive Known Folder Move put Desktop and Documents into the protected cloud world automatically — no user virtue required. Managed properly, a lost or dead laptop becomes an inconvenience: sign in on new hardware, files reappear. This is the right primary strategy for most machines — cheap, automatic, and it converts the endpoint problem into the (already-solved) cloud problem.
- True endpoint backup for the exceptions, named. A handful of machines genuinely warrant image-level backup: the CAD workstation with local project files and hours of software setup, the machine running the label printer and its irreplaceable config, the owner’s everything-laptop. Put them on the map by name — the failure mode isn’t choosing wrong; it’s the exception nobody listed.
The seams: where one strategy beats three
Coordinating the worlds is where the design pays: coverage reconciles against one inventory — new server, new app, new hire’s laptop each trigger the “which world, which tier?” question at onboarding, not after · monitoring watches all three in one place, so the endpoint sync that silently stopped gets the same daily attention as the server job · restore tests rotate across worlds — a server VM-boot this quarter, a SaaS record next, a “new laptop, old files” walkthrough after that · and retention and runbook documents span the map, because real disasters don’t respect the columns.
The bottom line
Three worlds, three failure patterns, one map: servers imaged and layered per their tier, cloud data backed up on purpose rather than by assumption, endpoints funneled into the cloud world with the true exceptions named — all reconciled against one inventory and tested on one rotation. Draw the map once, and “are we backed up?” stops being a feeling distributed across three separate tools and becomes a table with no blank rows.