We publish this because you should not have to trust an aggregator’s judgement without being able to inspect it. The two comparable India GCC platforms publish neither a methodology nor a paid-placement policy. We think that is the wrong way round. Read it with neutrality and conflicts, which covers what else ACV Solutions sells, and paid placement and sponsorship.
Every item comes from a source in our source register: a list of approved sources, each with a recorded licensing position. A source is added by a named decision with a written rights note, not by a configuration change. Sources fall into three groups.
Open feeds publish machine-readable feeds and we read them — Zinnov, ANSR, Everest Group and Business Standard. These are analyst, operator and publisher feeds that are open to automated readers.
Primary sources are regulators and government bodies: IFSCA, Invest India, state IT departments. Most publish no feed. We read them and report fact plus citation — figures, dates, registrations. We do not re-host their documents.
Everything else is linked, never reproduced. Six of the seven major Indian business publishers, and nasscom, do not serve automated readers. We respect that. A person reads the article in a normal browser and writes a card carrying the headline, the publisher, the date and a link to the original. We do not fetch those sites, and we do not disguise our requests to get past a refusal. If a publisher does not want automated readers, that is the end of it.
The general business feeds are matched against written-out phrases: global capability centre, global in-house centre, captive centre, offshore development centre, and capability centre or GIC where an Indian place or group-company word appears nearby. On those feeds the bare acronym “GCC” does not count: it is the Gulf Cooperation Council at least as often as it is a capability centre, and a feed filtered on the acronym fills with trade summits. It counts on three specialist feeds only — Zinnov, ANSR and Everest Group — where it means nothing else, and even there an item that names the Gulf Cooperation Council is refused. Every item the filter lets through, from any feed, is still read and published by a person.
What the filter turns away is kept, with the reason. A discarded item teaches nobody anything; a register of refusals is how we find out the phrase list is too narrow, and how we prove the hit rate to ourselves rather than assuming it.
This is enforced on the tables rather than by convention: a card that lacks any of them physically cannot be published. Every card carries the source’s name, as the source would want to be named; the publication date; a link directly to the original, never to a cache, mirror, reader view or archive; a summary in our own words, capped, never the article itself; and an impact statement, which is our view of why it matters to you.
The impact statement is the only part of a card that is ours, and it is set apart on the card so you can tell. Regulatory and policy items additionally carry a review date. When one lapses the card is labelled as needing review and drops down the ranked feed. It does not silently disappear — somebody may have linked to it.
Tagging is keyword rules, not a model: the same text produces the same tags today as it did last month, and you can read the rules. Signed in, your feed is ordered by five things compared in a fixed priority — what you have explicitly asked for, meaning the topics and cities you follow; then source trust and the item’s freshness; then your declared profile of sector, cities and lifecycle stage; then what you have read and saved in the last thirty days; then popularity, log-scaled and capped so it can only ever break ties.
A component is consulted only when everything above it is equal, so popularity cannot outrank something you explicitly follow. That is the mechanism by which feeds usually turn into tabloids, and we have closed it deliberately: an automated test fails the build if it ever stops being true. There is no paid-placement input to this ranking — not a disabled setting, it does not exist in the code. See paid placement and sponsorship.
Signed out, the feed is ordered by recency and source trust only. We do not build a profile of you before you have given us one. Every card carries the reasons it placed where it did, and you can open them.
These are not guidelines we follow carefully. Most of them are constraints and triggers on the tables themselves, so a row that breaks one cannot be written by any code path, including one written later by somebody who has not read this page.
A published card carries the source that vouches for it, the date it was published, a link back to the publisher, and the name of the person who decided it should be public. No exceptions, and no “various sources”. A card missing any of those cannot be stored in the published state at all.
Headline plus at most 40 words, and fewer where a publisher's terms say fewer. Never the full article, never a paraphrase long enough to substitute for reading it. The cap is recorded against the source and the row is refused when a summary exceeds it, so raising it is a decision about that publisher's terms rather than a setting.
Every canonical link is an https link to the publisher's own page. A cache, a mirror, an archive copy, an AMP rewrite or a paywall bypass is refused. Pointing a reader somewhere the publisher did not put the article is not attribution.
Every card carries its publication date, and a card past its review date is labelled as needing review on the page. Content that has aged is labelled, not hidden — somebody may have linked to it, and a gap teaches them nothing.
The feed's order and inclusion are never for sale: no card is in the GCC Pulse feed, or placed where it is, because someone paid. A sponsor can pay only for a labelled sponsor panel, which sits outside the feed on the Pulse page and never ranks with editorial. Nothing publishes itself either: there is no automatic path from a feed to the public feed, whatever the source, however long it has been reliable.
No card asserts a benchmark, ranking or comparison that GCC Pulse did not compute itself from data we were given permission to use. We do not hold that data yet, so we do not make that claim yet.
Sources sit in three tiers: an open feed we may read automatically, a primary source read by hand, and publishers who block automated readers, whose cards are typed by a curator and whose sites we never fetch. A source that answers our fetcher with a refusal moves to manual the same day, and is never retried with a disguised user agent. A feed that fails three times in a row retires itself, because a feed that has stopped answering looks exactly like a quiet week.
A regulator's material is reproduced as fact plus citation — figures, dates, registrations — never as re-hosted document text. A policy card also cannot be stored without a review date. A rule that has changed and not been re-read is the most expensive wrong fact this product can hold.
Nothing auto-publishes, from any source, including the open feeds. In our first months the review rotates between three people at ACV, each working from the same written rules shown on screen while they decide. A curator starts a shift, works the queue, and ends it with a handover note that the next person sees at the top of theirs. Each card is read against its original and its summary checked against the source’s extract cap before it is published or rejected with a reason.
This is slower than automation. It is also the only way the rules above can actually be checked: a neutral aggregator that publishes what it has not read is not neutral, it is automated. Every publish, rejection and re-review is an audit row naming the person, and the database refuses a published row that records no reviewer at all.
Each source carries a review window, and each card gets a review date from it. Policy items get the shortest window because a changed rule is the most expensive thing to have wrong. When a review date passes, the card is labelled as needing review on the page and ranks lower — it is not hidden, and it is not quietly treated as current.
A card also goes stale the moment its source is disabled, which is what we do first when a publisher objects. Take it down, then discuss it.
We will not publish a benchmark, score or comparison we have not computed ourselves from data we were given permission to use. If we ever say your setup plan compares to N other builds, it will be because N real organisations gave us their data, and we will say so and show the basis. We do not currently have that data, so we do not currently make that claim.
Sources move. Links die. A filter built on written-out phrases misses the piece whose author only wrote the acronym. A summary written at speed can miss the point of a piece. Corrections are made on the card rather than quietly in the background, and this page carries a date, as does everything on it. If you find something wrong, tell us.
The Pulse feed carries no paid, sponsored or promoted card. A labelled sponsor panel may sit outside it on the Pulse page. Paid placement and sponsorship says where money can and cannot appear anywhere on the platform.
Sponsor panels went live on 21 Oct 2026 (ADR-007). These sentences were on this page before that date and are no longer true as written; each is quoted exactly.
Before: “Sponsored or partner content is labelled and kept in its own band. It never ranks with editorial, and Vector Pulse carries none of it at all.”
Now: The feed's order and inclusion are never for sale, and the feed carries no paid card. A labelled sponsor panel can appear on the GCC Pulse page, outside the feed and its ranking.
Before: “Pulse carries no paid, sponsored or promoted items at all.”
Now: The Pulse feed carries no paid, sponsored or promoted card. A labelled sponsor panel may sit outside it on the Pulse page.
