Home / Comparisons / Keep vs Exit
① Comparison · Strategy
Every renewal cycle revives the question of leaving the mainframe. Most exit cases collapse under honest numbers. That does not make the option worthless. A real, costed exit is the strongest lever you have, whether or not you ever use it.
Audit notice or renewal under 18 months out? We mobilize within 48 hours.
Get expert help →For most large enterprises running stable, business critical workloads, keep the platform and attack the software bill, because migration cost is large, certain, and front loaded while exit savings are smaller and arrive years out. Exit is the right call for a contained estate with a clear target platform and a portfolio you can actually rewrite or retire. Either way, build the exit case honestly. A genuine costed exit study, even one that concludes keep, is the most powerful lever in the next renewal, because it replaces the vendor's assumption that you are captive with a real walk away number.
The two paths fail in opposite ways. Keep fails quietly through an unmanaged software bill that creeps every year. Exit fails loudly through a migration program that overruns and a savings line that never materializes. An honest comparison prices both the certain and the optimistic numbers on each side.
| Dimension | Keep and optimize | Exit and migrate |
|---|---|---|
| Up front cost | Low; advisory and negotiation effort | High; migration program, certain and front loaded |
| Timeline to benefit | This renewal cycle | Years; benefit arrives after the spend |
| Run rate after | Lower software bill on the same platform | New platform license plus labor; not zero |
| Execution risk | Low; known platform, known levers | High; undocumented logic, data, parallel run, testing |
| Reversibility | Fully reversible decision | One way; sunk cost if it stalls |
| Hidden cost | MIPS creep and uplift if left unmanaged | The dependent application long tail |
| Negotiation value | Direct: lower the bill now | Indirect: a credible walk away that disciplines renewals |
Directional framing of the cost structure on each path. Actual figures are estate specific; the discipline is pricing both the certain and the optimistic numbers honestly.
Your workloads are large, stable, and business critical, the application portfolio is deep and partly undocumented, and the software bill has never been seriously optimized. In that situation the fastest, lowest risk money is on the platform you already run: reconcile consumption, reset stale baselines and conversion ratios, cap uplift, and unbundle. The exit study still earns its place, as the leverage that makes the keep negotiation work, not as a plan you intend to execute.
The estate is contained, the application portfolio is well understood, a credible target platform exists, and the migration can be scoped without heroic assumptions. Exit also rises up the list when the strategic direction is set regardless of pure cost. Even then, price the dependent application long tail, the parallel run, and the new platform's own licensing honestly, because those three are where optimistic cases go to die.
For most large estates running stable, business critical workloads, keeping the platform and aggressively optimizing the software bill beats a full exit on a five to seven year horizon, because migration cost is large, certain, and front loaded while the savings are smaller and arrive late. Exit wins where the estate is small, the application portfolio is contained, and a credible target platform exists. The decision turns on honest numbers, and the most common error is an exit case built on understated migration cost and overstated run rate savings.
They commonly fail on two estimates. Migration cost is understated because the difficult parts, undocumented logic, data migration, parallel running, testing, and the long tail of dependent applications, are scoped optimistically. And savings are overstated because the replacement platform, its licensing, its labor, and the cost of the migration program itself are netted out too generously. A case that survives realistic numbers on both sides is rare, and when it does survive it tends to be a contained estate rather than a flagship one.
A credible, costed exit option is the strongest single lever in a mainframe renewal, because it removes the vendor's assumption that you cannot leave. The key word is credible: a bluff is read as a bluff. A genuine exit study, even one that concludes keep, gives you a real walk away number and a real alternative to point to, which is exactly what disciplines the renewal. Building that option is valuable whether or not you ever exercise it.
Related guide: exiting Broadcom CA products: costs and timelines. Related guide: Natural and Adabas modernization: license exit math. Related concept: MIPS creep explained. Put it to work: mainframe cost optimization.