We Checked Apollo's Singapore Companies Against the ACRA Registry. 6 in 10 Resolved Cleanly.

If you sell into Singapore, you have probably heard the complaint: Apollo's database is deepest in North America, and the further you get from it, the more the records wobble. The complaint is common; hard numbers are not. So we ran a simple, repeatable test in June 2026: take the companies Apollo lists as employers of Singapore-based professionals, and check each one against ACRA, Singapore's official company registry. This post sets out exactly what we did, what we found, and where the result should and should not change how you prospect.

Why we ran this

APAC data quality is one of the most consistent themes in independent coverage of Apollo. Prospeo published an accuracy teardown on prospeo.io examining where Apollo's contact data holds up and where it thins out. Reviewers on G2 repeatedly report outdated job titles for APAC contacts. Salesforge's review of Apollo likewise notes gaps in its APAC coverage. None of those sources answered the narrower question we cared about: when Apollo says a Singapore-based professional works at a company, does that company resolve to a registered Singapore entity at all? Singapore is one of the few markets where this is cleanly testable, because ACRA maintains a public register of every entity ever registered here, live or dead. So we tested it.

Methodology

We pulled the first 300 results from Apollo's people search filtered to person location Singapore, in June 2026, across pages 1 to 3 under the default ranking. Deduplicating employer names gave 281 unique employers, of which 278 were usable after normalisation (3 were dropped because their names fell under 3 characters once legal suffixes were stripped). Each normalised name was then prefix-matched against the entity name column of the ACRA registry, which holds 2.09 million entities, taking the best match and preferring a live status wherever one existed.

One design choice matters more than any other: the matching is conservative. A brand name registered under a different legal entity name counts as not found, even when the company plainly operates in Singapore. This understates Apollo's true coverage in one direction. The sample composition, which we cover in the caveats below, pushes the other way and likely overstates it. Keep both in mind when reading the numbers.

The results

Of the 278 unique employers, here is how they resolved against the registry.

Six in ten resolving cleanly is the headline, and we would describe it as neither a disaster nor a pass. For a database built and ranked around US coverage, it is a credible result on a list of mostly large employers. The interesting parts are the two smaller buckets, so let us take them in turn.

The dead-entity tail

Ten employers, 3.6% of the sample, matched only entities that are no longer alive in the registry. Before the table, one caveat stated plainly: a dead-entity match can be a name collision. A name here matching a struck-off ACRA entity does not prove the Apollo record refers to that exact entity; an unrelated company with the same name may have died locally while the company Apollo means operates on overseas. Treat individual rows as indicative, and the aggregate, ten out of 278, as the finding.

What the 35.6% no-match really contains

It would be easy, and wrong, to read 99 no-matches as 99 bad records. The bucket breaks down into four distinct groups, and only one of them is unambiguously a data problem.

• Government bodies and political organisations, which ACRA does not register. The Ministry of Manpower, the Ministry of Sustainability and the Environment, the Health Sciences Authority, the People's Association, and the People's Action Party all appear in Apollo's results, and all correctly have no ACRA entry. These are real employers; the registry is simply not where they live.
• Foreign institutions and brand-versus-legal-entity mismatches. Names like Marriott International, SentinelOne, Danaher Corporation, HCLTech, UL Solutions, and Stanford University GSB operate in Singapore through differently-named legal entities, or employ Singapore-based staff without a same-name local entity. Our conservative matching counts all of these as not found.
• Normalisation artefacts on our side. Bank of Singapore failed on apostrophe handling, and a KSUG.AI user group is not a company at all. We count these against the no-match bucket rather than quietly correcting them after the fact.
• The remainder, roughly 70 names, are smaller employers with no name-matching ACRA entity: foreign companies employing Singapore staff without a local entity, community groups, personal brands, or stale organisation records. This is the group that genuinely needs manual verification, one record at a time, before a rep should spend a minute on it.

What this means for your team

The practical reading is straightforward. If you export a Singapore list from a US-built database, plan as if roughly 4 in 10 records need some form of verification before a rep invests time on them, and as if 1 in 25 companies on the list may no longer exist at all. Some of that verification is quick, like recognising that a ministry is a real employer outside the registry. Some of it is slow, like working out whether a small foreign employer still has anyone in Singapore. Either way, the cost lands somewhere: in rep hours spent researching instead of selling, in bounces against your sender reputation, or in calls to companies that wound up two years ago.

The registry-first alternative

HuntSales approaches the problem from the opposite direction. Instead of building a global people database and hoping the Singapore slice holds up, our Singapore company database starts from ACRA itself, so company existence, registered name, UEN, and live-or-dead status are facts pulled from the registry, not guesses to be verified later. If you only need to check one company right now, our free UEN lookup answers in seconds with no login. And inside the product, prospecting filters lean on the same registry data, so a list you build is a list a rep can work.

Full caveats

A study like this is only useful if its limits are stated as plainly as its findings. There are three.

• Dead-entity matches can be name collisions with unrelated registered entities. Treat individual rows in the dead-entity table as indicative, and the aggregate as the finding.
• The sample skews toward employers of the professionals Apollo ranks first, which over-represents large MNCs and likely overstates Apollo's match rate compared with a typical SME prospecting export. If your ICP is Singapore SMEs, the realistic expectation is worse than 60.8%, not better.
• This measures company-name resolvability against the official registry, not email deliverability. A record can resolve cleanly and still bounce, and a no-match record can carry a perfectly deliverable email.

We have shared the bucket counts, the named dead-entity matches, and the no-match breakdown above so the result can be checked rather than taken on trust. If your own export tells a different story, we would genuinely like to see it.