XORV AnswerRank

AI search changed where discovery happens. Measurement did not follow it.

The unit of measurement is a buyer question — not a keyword, and not a headline.

• Where does the brand appear, and where does it not?
• How often, and on which engines?
• Against whom — who else does the answer name?
• On what evidence — which sources were referenced?
• What should be improved, and did it change afterwards?

Eight concepts that are never allowed to collapse into each other.

Brand understanding

What the business actually is, built from public evidence rather than one page of marketing copy.

Buyer questions

The measured question set — derived from the business, versioned, and the same across every engine compared.

Answer measurement

What each engine actually replied, recorded per question and per engine as a first-class observation.

Competitive comparison

Which other brands the answers named, and how that pattern differs across engines and questions.

Sources & citations

The references an answer leaned on, kept distinct from the answer itself.

Evidence

The retained proof behind every claim the interface makes, so a number can be opened rather than trusted.

Recommendations

Derived from measured truth only. Where there is nothing to recommend, the product says so instead of inventing work.

Remeasurement

The same question set re-asked over time, so a change can be observed rather than assumed.

The hard part of an analytics product is not the chart. It is deciding what a number is allowed to say.

Decision density, not dashboard density

Every surface earns its place by answering a question the customer actually has. Density that does not lead to a decision is removed.

A number carries its population

Any figure shown states what it counts and what it counts out of. If one input to that is unknown, no confident number is shown.

Absence is not zero

Not measured, not available, refused and zero are four different states and render as four different things.

Evidence is one click away

Claims drill down to what produced them. An analytics product that cannot be checked will not be trusted twice.

Deep underneath, simple above

The full evidence graph is not dumped on the overview. The overview answers; the depth is available when someone wants to argue with it.

A measurement product is mostly a correctness problem wearing an interface.

TypeScript end to end

One typed contract from the data layer to the interface, so a shape change fails at build time instead of in a customer's dashboard.

React & Next.js

Server-rendered application surfaces with the analytical work kept on the server — the frontend renders truth, it does not derive it.

PostgreSQL as the authority

Relational truth with the measurement history preserved. Historical observations are immutable; a correction is a new record, never a rewrite.

Tenant isolation as a database guarantee

Isolation is enforced at the data layer rather than in application code, so one workspace cannot read another's measurements.

Building this changed how I build everything else.

Turning metrics into decisions

A metric nobody acts on is decoration. The useful question is not 'what can we count' but 'what would this change'.

Making complex data understandable

The work is compression without distortion — saying less, while keeping what a careful reader needs to disagree with you.

Designing trustworthy analytics

Trust is built by how the product behaves when it does not know something. A beautiful wrong number costs more than an honest gap.

Density versus clarity

More information on screen is not more insight. Hierarchy is the product decision, not a styling one.

Owning something long-term

Shipping a client site ends. A product does not — which changes how you write, name and structure everything.