SEO works in categories where people search for what you sell, on sites that are technically capable of being found, given enough time to compound. Outside those three conditions it reliably disappoints — and most of the businesses who tell us SEO did not work for them were missing one of them rather than working with a bad agency.
When does SEO genuinely work?
Three conditions, and they are less about execution quality than most agencies admit.
- There is real search demand. People type something that describes what you sell, in numbers that matter. This is checkable in an afternoon and is the single biggest predictor of whether a programme succeeds.
- Your site can be found and understood. Crawlable, indexable, structured clearly enough that a search engine knows what each page is for.
- You can fund it long enough to compound. The mechanism is accumulation. A programme judged at month three is being judged before it has done anything.
Where all three hold, SEO has an economic property almost nothing else in marketing has: the work keeps returning after you stop paying for it. A page that ranks continues to bring enquiries for years. That is why it survives despite being slow, uncertain and hard to attribute.
When does it reliably fail?
Four failure modes, in descending order of how often we encounter them.
| Failure mode | What it looks like | What to do instead |
|---|---|---|
| No search demand | Rankings improve, enquiries do not | Ads, partnerships, or your existing customer list |
| Unresolved technical problems | Good content published, nothing indexes | Fix the foundation before buying more content |
| Keyword-led rather than question-led content | Traffic arrives and bounces | Fewer pages, each answering a real question fully |
| Judged too early | Programme cancelled at month three | Agree the horizon before starting, or do not start |
The first is the most common and the least discussed, because acknowledging it means an agency talking itself out of a retainer. In our own category, we found 97% of the keyword rows in a standard export had effectively no demand attached to them. Plenty of programmes are built on exactly that kind of list.
What does working actually look like?
Not rankings. Rankings are an intermediate signal, and a less reliable one than they were — when keywords carrying 68% of search volume in a category return an AI answer, position on the list below it describes progressively less of what happens.
What to watch instead:
- Enquiries attributed to organic search, tracked monthly. The only number that settles the argument.
- Visibility across a basket of terms, not one flagship keyword. Baskets move more honestly than individual positions.
- Whether AI answers in your category name you. Ask the engines directly. It is imperfect and varies between sessions, but it is free and most businesses have never checked.
- Pages entering their second year still producing enquiries. This is the compounding, and it is the whole reason to prefer SEO over renting attention.
The honest caveats
Three things we would want to know if we were buying this.
It is slower than anyone selling it wants to say. Six to twelve months in a competitive Singapore category is realistic. An agency agreeing to your three-month deadline is agreeing to something it cannot deliver, and the pressure that follows tends to produce shortcuts that cause damage.
Attribution is genuinely messy. Someone finds you through search, forgets, sees you again elsewhere, then enquires directly weeks later. That enquiry will not be attributed to organic search, and it was caused by it. This cuts both ways and is worth accepting rather than pretending precision exists.
Nobody controls the outcome. Which is why we would rather offer performance-based contracts against agreed outcomes than guarantee positions. The first puts our fee at risk against something we influence. The second promises something nobody outside Google owns.
Most businesses who say SEO did not work for them were in a category with no demand, on a site that could not be crawled, or judging the programme before it had done anything.
Frequently asked questions
Does SEO actually work?
Yes, in categories where people search for what you sell and where you are willing to fund it for long enough to compound. It fails predictably in thin-demand categories, on sites with unresolved technical problems, and where it is measured over a horizon too short for the mechanism to operate.
How long does SEO take to work?
Technical fixes can show within weeks. Content and authority typically take several months, and in a competitive Singapore category six to twelve is realistic. The compounding is the point: the work keeps returning after you stop paying for it, which is not true of advertising.
How do I know if my SEO is working?
Track enquiries from organic search, not rankings alone. Rankings are an intermediate signal that has become less reliable as AI answers take a share of clicks. If organic enquiries are not moving after two quarters of real work, something is wrong with the plan or the market.
Why did SEO not work for my business before?
The four usual reasons are: the category had almost no search demand; the site had technical problems that nothing else could overcome; the content targeted keywords rather than questions; or it was judged after three months, which is before the mechanism has operated.
Is SEO better than paid advertising?
They do different jobs. Ads buy immediate, controllable traffic that stops the day you stop paying. SEO compounds slowly and keeps returning. For most businesses the question is sequencing, not choosing — ads while SEO matures is a common and sensible pattern.
Is SEO still worth it now that AI answers exist?
Yes, but what you buy should change. The work that earns AI citations is largely the work that earns rankings, so a well-built page now serves both surfaces. What has lost value is thin content and reporting that tracks position alone.
If SEO has disappointed you before, the useful question is which of the four failure modes you hit. Send us the domain and the history and we will tell you what we think went wrong. See how we measure whether it is working, message us on WhatsApp, or send an enquiry.
Sources
Written by Charlotte Zhang, Operations Partner at Kaizenaire Pte Ltd (UEN 201932071D). Charlotte co-founded Kaizenaire with Ken Tan in 2019 and runs client delivery, after four years in operations at OCBC and HSBC.