The One-Person Portfolio: 22 Products as a Citation Network

01One domain can only ever be one witness

The standard advice for anybody running more than one site is to consolidate. Fold everything into a single strong domain, concentrate the authority, stop splitting the effort across places that will never individually amount to much. For ranking that advice is mostly sound. For being cited it is backwards, and I have twenty two live products that keep proving it to me.

Twenty two small sites that vouch for each other beat one big site that vouches for itself.

Why a machine wants a second source

A model deciding whether to put your name in an answer is doing a small risk calculation on somebody else’s behalf. A claim that appears on one domain is a claim. The same claim appearing on several domains that each check out separately is much closer to a fact, and the sentence itself never changed. What changed is the number of independent places it can be confirmed, and that is the only variable in the whole exercise you can actually manufacture.

What one big site looks like from outside

However good that single site is, it is one witness testifying about itself, and it has no way to escape that. Every page of supporting evidence sits on the same domain as the claim it supports. A site that only cites itself has made an argument. It has not produced a witness. That is not a criticism of big sites. It is a description of a shape, and the shape has a ceiling that no amount of writing gets you over.

There is a version of this that is simply wrong, and worth saying plainly. If you run one business with one product, splitting it across four domains to manufacture a network makes you weaker, because the extra sites have nothing of their own to say and a machine can tell almost immediately. The shape only pays when each domain is genuinely its own thing with its own users, and the identity behind them happens to be shared. Corroboration cannot be faked by duplication.

02The same claim, drawn two ways

Both panels below describe an operation of exactly the same size doing exactly the same work. The difference is only how many places the work is allowed to live, and what that does to the number of things able to confirm it.

Reading the two panels

On the left the arrow leaves and comes straight back, because there is nowhere else for it to go. On the right the centre is not a website at all, it is the identity every domain declares, and the ring is the part people skip. The spokes make you findable. The ring is what makes you believable. A hub with nothing joining the edges is still one witness wearing twenty two coats.

03Four ties, and every one has to match

None of this works by accident. Twenty two domains with twenty two slightly different stories is not a network, it is twenty two strangers, and a machine reading them concludes exactly that. Four things have to be identical across all of them before the shape resolves into one operation.

The four are the name, the description of what the operation does, the structured data that declares the shared identity, and the standard the instruction file follows. Two of those are words a person wrote and two are machine readable, and they fail in different ways. The words drift because nobody is looking. The machine readable half breaks because somebody shipped one site and forgot to update the other end.

One name, spelled one way, on every domain

The same legal name, the same person name, the same description of what the operation is, down to the punctuation. This sounds trivial and it is the tie that breaks most often, because each site gets written on a different day by somebody in a different mood. Two versions of your own story do more damage than one thin version. A machine resolving identity is matching strings before it is doing anything clever, and it will happily conclude that two spellings are two companies. The full discipline is in entity hygiene, and it is unglamorous work that pays for years.

One machine readable identity, declared in both directions

Every product page names the person and the parent operation in its structured data, and the portfolio page names every product back. Both directions, or the loop is open and only half the graph is traversable. That is what schema is actually for, and it is why the same file standard goes on all of them rather than the best one. A reciprocal link is a fact. A one way link is a hope. Keep one instruction file to a single standard across the whole set, because the file a machine reads first is worth nothing if it says something different on each domain.

One honest note on the number. This entry was scheduled when the portfolio was twenty one, and the twenty second went live the morning it published. That is the argument for publishing floors rather than exact figures, and the reason the earlier claim in this ledger still holds today.

Tomorrow: thirty entries in, what getting quoted actually took, with the real numbers attached.