Skip to main content
ProtoComet
ProtoComet

  • Prove the new design on one real workload before you commit to the rest.

    We design the architecture around your constraints and budget, then move one important workload onto it and run it alongside the old system. You see the real cost, performance and effort before committing to more.

Where this actually breaks

Most teams reach this stage with a roadmap in hand. These are the points where it stalls, and each one comes back to a design that hasn’t yet been tested against how your applications will actually run on Azure.

Every team is choosing its own Azure services

One team wants containers, another wants App Service, and a third has started on virtual machines. Without one agreed design, each application lands differently, the costs vary widely, and nobody can support all of them.

The first migration costs more than the servers did

It was lifted onto a virtual machine sized for peak load, running the same code with the same problems. The bill went up, performance stayed the same, and leadership’s confidence in the programme dropped before wave two started.

Leadership won’t fund the programme on an estimate

The roadmap has cost ranges, but the people approving the budget want to see something working first. Without a proven example, the first wave becomes the experiment, on a system the business can’t afford to lose.

AI tools upgraded the code, and it can’t run in the cloud

An AI-assisted upgrade moved the application to modern .NET, and it builds. It still keeps sessions in memory, writes files to a local disk and calls a WCF service. Nobody has designed where and how it should run.

Design, then prove: six steps from constraints to evidence

Steps 1 to 3 are the Target Architecture (Tier 2) and end with a design. Steps 4 to 6 are the Pilot Modernization (Tier 3) and end with one workload running in production and a recommendation on what happens next.

1. Constraints

We start with what can’t change: data residency rules, integrations with your ERP or partners, budget limits and the skills your team already has. The design has to fit inside these, so they’re written down first.

Tier 2 · constraint register

2. Target design

The Azure setup your applications will run on: compute, data, identity, networking and monitoring. It’s sized from the Tier 0 inventory and costed to run, so you know the monthly bill before anything moves.

Tier 2 · target architecture

3. Approach per verdict

A migration approach for each 6R verdict in your roadmap: how a Replatform moves, how a Refactor is staged, what a Rehost must avoid. Every decision is recorded with the options considered and why one was chosen.

Tier 2 · decision records

4. Pilot selection

One workload is chosen against five written criteria. It has to matter enough to prove something, be small enough to finish within a quarter, and follow the same approach as the largest wave still to come.

Tier 3 · pilot charter

5.Build and run in parallel

Your engineers and ProtoComet build the pilot on the new architecture together. It runs in production alongside the old system, with traffic moved across gradually and a tested way back at every stage.

Tier 3 · runnig workload

6. Measure and decide

Actual cost, performance and effort are compared with the Tier 2 estimate. You get a written recommendation to go ahead, pause or re-plan the remaining waves, based on what the pilot showed in production.

Tier 3 · go / no-go memo

How the pilot workload is chosen

The wrong pilot proves nothing and delays everything. A candidate has to pass all six checks

pilot selection criteria

How the work actually runs

Both tiers are fixed price. Tier 3 is scoped once the design is agreed: one workload, one parallel run and one written recommendation. If the pilot turns up something the assessment missed, the recommendation says so and the plan for the remaining waves is updated.

Tier 2: design in three to five weeks


When it follows a Tier 1 roadmap, the design typically takes three to five weeks, a little longer if it starts from another firm’s assessment. The Azure environment is designed alongside the architecture, so the two fit from the start.

Tier 3: pilot fits within a time-frame


Build, parallel run and the go/no-go decision all happen within one quarter. The parallel run lasts until cost and performance settle, usually a few weeks, and its length is agreed in the pilot charter before any work on the pilot starts.

One senior consultant throughout


The consultant who makes the design decisions also builds on them during the pilot. If a choice doesn’t hold up under real traffic, the person who made it sees it first and records it in the recommendation, along with what to change.


What you walk away with

A design your team or any team can build from, and one workload that proves it. The pilot keeps running in production after Tier 3 ends, as the first part of your modernized system.

➖ Constraint register. What the design had to fit inside, from data residency to integrations and budget, and the reason behind each constraint.


➖ Decision records. Every significant choice, the options considered, the approach for each 6R verdict and what would trigger a rethink.


➖ Pilot in production. One workload live on the new architecture, with its deployment pipeline, monitoring and runbooks for everyday operation.

➖ Target architecture. Compute, data, identity, networking and monitoring for your applications, with the Azure environment specified and costed.


➖ Pilot charter. The workload chosen, the six checks it passed, how success will be measured and how long the parallel run will last.


➖ Go/no-go memo. Actual cost, performance and effort against the estimate, and a clear recommendation for every wave still to come.

Our Way

Every design starts from how your applications actually behave in production. The month-end report that pushes the database to full load for forty minutes, the file drop a partner still relies on, the scheduled job that assumes a single server: the architecture has an answer for each of these before the first wave begins.

The constraints are written down first, and the design is built to fit inside them. The pilot is then chosen to test the most demanding pattern your remaining waves will use, so what it proves carries over to the work that follows.

Every decision is recorded with the options considered and the reason one was chosen. When someone asks in two years why the system was built this way, the answer is on record, and your team can judge whether it still holds.

A modernized application can still age badly if the architecture underneath it is wrong. The design has to suit how your business works today and leave room for where it is going, including AI features that aren’t on the roadmap yet.


Yogesh, founder and principal consultant

Questions clients typically ask

Short, direct answers. Where the answer depends on your situation, we say what it depends on.
Send it over and we’ll review it first. If it already records the constraints, the decisions behind each choice and a running-cost estimate, Tier 2 becomes a shorter review and is priced that way. If parts are missing, you’ll see exactly which ones before anything is proposed.
Yes. The target design includes a monthly running-cost estimate, based on the sizing from your Tier 0 inventory. The pilot then shows the actual cost of one workload, so the estimate for the rest is checked against a real bill.
Then it has found something worth knowing while only one workload is affected. Traffic moves back to the existing system, which has been running throughout. The go/no-go memo explains what went wrong, whether the cause was the design or the estimate, and what changes for the remaining waves.

It’s designed to. Business logic is moved behind APIs, data access is made explicit, and identity is set up so new services can connect securely. Those are the foundations an AI feature needs, so adding one later doesn’t mean redesigning the system.

Ready to move?

Let's talk about where you are and where you want to go.  and whether modernization is even the right answer this year. If it is not, you will hear that.

Let’s Talk