Clarity has a field it calls calculated. The overall status indicator on a project is, in Broadcom’s own description, the sum of the schedule status, scope status, and cost and effort status fields on the status report. The same page adds that at least one status report has to be completed before the indicators calculate anything at all.
So the arithmetic is real and the inputs are three dropdowns. That is not a criticism of the arithmetic. It is the whole shape of the disagreement, and it is worth stating before anything else, because almost everything Clarity does downstream of that field it does better than we do.
Start with the part where Clarity is simply ahead. Financial management here is not a budget column; it is a chain — cost plan to approved budget plan to benefit plan, with only one cost plan able to be the plan of record, and appraisal math that actually runs: net present value, return on investment, internal rate of return, modified internal rate of return, payback and breakeven, each with a published formula in the financial management guide. Chargebacks carry debit and credit rules, time-phased general-ledger allocation percentages, and an internal invoice that moves from proforma through submitted to approved. Resource management separates availability, allocation, assignment and actuals, and books people hard or soft as distinct states. Clarity is FedRAMP authorized, which for a federal buyer ends the conversation in Clarity’s favor before it starts. Broadcom announced a Leader placement in the 2025 Gartner Magic Quadrant for Strategic Portfolio Management, its second consecutive year.
BoardWalk does none of that. No timesheets, no requisitions, no chargebacks, no depreciation, no internal invoice. A portfolio office whose core job is the money should read the paragraph above twice and this page once.
Where the model shows its age
A Clarity portfolio is a snapshot of investment data, refreshed on a cadence you configure. Roadmaps work the same way: scenarios, one of which is tagged as the plan of record. That is a coherent design and it was the right one when the underlying facts changed monthly. It assumes the interval between refreshes is short relative to the rate at which reality moves.
The capital-versus-operating split is where the assumption bites hardest, and it is worth being precise because it is easy to overstate. Clarity does the split by classification: a task’s cost type, or a team member’s capitalization percentage. Broadcom documents that the cost type field does not display out of the box and has to be configured in Studio, and — the sentence a controller should read twice — that if the plan of record does not use the cost type grouping attribute, planned operating cost summarizes all costs as operating. There is no default depreciation field; the documentation tells you to add a line item.
Nothing there is broken. Everything there depends on somebody having configured a hidden field correctly, and on nobody having quietly stopped maintaining it. The failure mode is not an error message. It is a number that looks right.
What Vaia is, exactly
Clarity’s AI is Vaia, and the general-availability line is unusually easy to establish: the documentation page carried a “(Beta)” suffix through 16.4.0 and dropped it at 16.4.1, from which release the capability is available by default rather than by support request. It is off until an administrator activates it, it runs on your own Gemini, OpenAI or Azure account, and Broadcom’s activation notice says plainly that use of the service requires an account with the provider and that Broadcom bears no responsibility or liability for it.
It also ships with governors worth knowing before a business case is written on it: a ceiling of five thousand generative requests a month, and a hundred-record cap on a data source, beyond which records are truncated before being sent to the provider. Watching what it costs means running SQL against the transactional schema.
We are the opposite arrangement on one axis and the same on another. The same: BoardWalk also runs on the organization’s own provider key, against a ceiling the workspace sets, and we think that is the correct model — an AI feature nobody can meter is an AI feature nobody can govern. The opposite: what comes back is not an answer, it is a proposal, and it sits in a register until a named person accepts or rejects it. Vaia summarizes and answers. Nothing it produces is an object with a decision attached.
One more fact belongs here rather than in a footnote, because a buyer evaluating Clarity’s modern surface will meet it. Broadcom’s own beta register for 16.4.3 still lists Plans among the beta features, with the standard caveat that there is no guarantee a beta feature ever reaches general availability.
What we built instead
A requirement carries its citations, and the citations resolve against one pinned commit rather than against “the repository”. A test result binds the requirement at the rung the evidence supports; a skipped test binds nothing; a failing test still binds, with the failure as its outcome. Then a named person accepts, and the column recording who accepted is a foreign key into the identity table, which an agent has no row in and cannot acquire.
There is no overall status field to calculate, because there is no overall status. Counts sit beside their denominators — nine of thirty-eight bindings accepted — and are never averaged into a verdict. Nothing about that is cleverer than Clarity’s arithmetic. It is just arithmetic over things that happened rather than over things that were reported.
The objection worth taking seriously
Clarity could add commit-level binding. It has a documented integration surface — Rally, Microsoft Project and Open Workbench in the product, with Jira, Azure DevOps and ServiceNow reached through ValueOps ConnectALL, a separate product on its own release train — and a large engineering organization behind it.
The reason we still think the position holds is that most of what we built is refusals, and refusals are expensive to retrofit. Consider what Broadcom already documents about its own SaaS: customers may not customize the database schema, may not add triggers, stored procedures or custom tables, and must run the current version of the service. Those are sound operational constraints for a platform of that size. They also mean the custody guarantees we make in the schema are not something a Clarity customer could add for themselves, and not something Broadcom could add without touching every installed base at once.
There is a commercial edge to the same shape. Clarity licenses are not assigned; they are inferred. Broadcom documents that access rights determine which license applies, and the highest license type prevails. Granting somebody edit rights is therefore also a purchasing decision, which is a reasonable design and one more thing an administrator has to hold in their head.
Where Clarity wins
The financial model, comprehensively. Resource booking with hard and soft states, skills matching and capacity portlets, though a buyer should know these live in Classic PPM rather than in the modern experience — requisitions in particular are documented as not supported outside Classic. Automated time capture is real and it is worth correcting a common misreading: Clarity’s Frictionless Time can pull effort from Rally and post a timesheet, so “Clarity makes people fill in timesheets by hand” is not a true sentence. Chargebacks, FedRAMP authorization, on-premise deployment as an option, and two decades of implementation patterns that a large PMO can hire for. BoardWalk has none of these and is not trying to.
Choose Broadcom Clarity if…
Choose Clarity if the portfolio office is, in practice, a finance function. If the recurring question is what to capitalize, what to charge back to which cost center, what the internal rate of return on a program looks like against a time-phased benefit plan, and what the labor actuals say — Clarity answers all of that natively, and no part of this product is a substitute for it.
Choose it also if you need FedRAMP authorization, an on-premise option, or a resource model with real booking states. Those are category requirements, and a category requirement is not something an argument about evidence can talk anyone out of.
Come here instead if the delivery organization under the portfolio has gotten smaller and more machine-produced, and the question that has stopped having an answer is which requirement a build was made against and who accepted the result. That question is not downstream of a better status field. It is downstream of recording different things. A receipt verifies without an account, which is the quickest way to find out whether it is your question.