What my failed Amazon affiliate site taught me about research

As a former physical therapist turned freelance blogger, I am used to assessing a problem before acting on it. My time in Malta also gave me a folder of observations about study-abroad costs, neighbourhoods, weather and the small annoyances of daily life. I thought that experience was enough to build a useful site.

What my failed attempt at an Amazon affiliate site in the Malta niche taught me was that personal experience can generate hypotheses, but it cannot validate demand. I chose travel adapters, headphones, luggage scales and compact desks, then organised articles around each product. The logic looked tidy on paper. Nobody had yet proved that readers arrived with a buying task or that the products could be delivered where they lived.

I published fewer than fifty pages on WordPress using AFFINGER5 and tracked impressions, clicks and Amazon clicks for six months. Search Console showed traffic for informational phrases such as where to stay in Valletta or how hot Malta gets in July. The Amazon dashboard showed negligible commissions. The site was not technically broken; it had simply been aimed at an audience with no reason to buy from it.

The expensive lesson was about research, not content volume. I had checked words in a tool, skimmed competing pages and treated every visit as a potential conversion. A better process would have joined keyword demand to search intent, product availability and the reader’s location before I designed the site. That distinction now shapes how I research every blog and affiliate project.

The topic was not automatically a market

My first mistake was confusing a subject I knew with a market I understood. Malta was a credible editorial niche, yet it was not a natural Amazon category. Most product searches happened on national Amazon domains, while my readers might be planning a temporary stay, studying for a semester or comparing a life change from another country. Each group had different urgency.

I also ignored fulfilment. A buyer in Australia could not assume that an item listed on a British or American marketplace would arrive quickly, use the right plug, carry a familiar warranty or appear in local search results. By positioning the site for Malta while linking outward to mismatched catalogues, I added friction at the exact point where a commission depends on trust and a completed purchase.

Keyword volume concealed the intent gap

A keyword tool gave tidy monthly estimates for terms such as Malta travel essentials, best apartments in Valletta and Malta study advice. I treated each number as evidence of opportunity, even when the phrase sat in the awareness stage. High volume says that people type a phrase; it does not tell you whether they want an article, a local recommendation or a product they can buy immediately.

Long-tail searches were not the answer either. A specific term can attract many impressions and very few commercial actions. I needed to study modifiers, SERP features, click-through rate and the wording used by ranking pages. Terms carrying buy, price, delivery or comparison often suggested stronger commercial intent, while how, why and what usually called for education. Volume became useful only when paired with those signals.

The search results contradicted my monetisation plan

Inspecting the search engine results page should have been the centre of the research, not a final check. Google mainly displayed travel blogs, university pages, property listings and forum discussions for my chosen queries. Amazon barely appeared. The ranking pages were helping readers understand a destination, not choose a gadget from a checkout box.

That mismatch affected content strategy as well. Product roundups felt natural to me because I had selected the products, not because the audience had requested them. I later learned to map the buyer journey from informational research to comparison, then to transaction. If no credible bridge existed, monetisation had to change; forcing an affiliate link into every guide was merely cosmetic.

One audience was actually several customers

My supposed audience included language students, long-stay visitors, retirees, remote workers and families. They shared a destination but not a decision. Students might need a power bank and travel insurance; families might compare strollers; remote workers might seek a monitor arm. Each customer had a different product basket, trust threshold and definition of value, so one generic profile produced vague content for everyone.

Local behaviour sharpened that problem. In Sydney and Melbourne, readers may begin searches on a phone during a train or tram commute, while households in Brisbane may prioritise heat management, humidity and insect control over the features I assumed mattered in Malta. Australian Type I plugs also differ from Malta’s Type G plugs. Even a basic product shortlist changes when electrical standards and daily conditions enter the picture.

Australian shoppers also meet familiar alternatives before Amazon. They may compare prices with Coles, Woolworths, Bunnings or a specialist Australian retailer, and they expect prices in Australian dollars. Those habits do not make Amazon irrelevant; they raise the standard for delivery promises, product information and proof. The lesson was to research a customer’s alternatives, not merely the commission offered by a programme.

Questions I should have answered before publishing

Competitor analysis became more useful when I stopped treating it as a list of rival domains. I started asking what each ranking page solved, who published it, when it was updated and why it earned visibility. This approach reveals gaps that a volume graph cannot, such as an underserved comparison or a delivery question left unanswered.

Those answers would not have guaranteed revenue, but they would have exposed weak assumptions early. The most important finding was often negative: the query had no natural commercial bridge, or the catalogue did not fit the location. Publishing fewer pages with that evidence would have saved time and prevented an attractive content calendar from hiding a weak business model.

Local context belonged inside the research brief

Moving the audience to Australia changed how I evaluate a niche. On amazon.com.au, stock, delivery estimates, Australian-dollar pricing and local reviews matter. A product that looked competitive on a UK page could be unavailable, slow to arrive or incompatible with Australian plug sockets. I now treat catalogue access as part of keyword research because an unavailable product cannot convert regardless of its apparent relevance.

The same principle applies to claims and customer communication. Under the Australian Consumer Law, price statements, testimonials and advertising claims must be accurate and not misleading. An affiliate relationship should also be disclosed clearly so readers are not led to think a recommendation is independent. If a site collects personal information, the Privacy Act 1988 applies, while commercial email requires consent under the Spam Act 2003. These obligations shape the offer rather than sitting in a legal-page afterthought.

The research system I use now

The failed Malta project led to a repeatable sequence: define the decision, map the demand, inspect the result types, validate the offer and test the path to purchase. I research the audience and marketplace together. That sounds obvious, yet affiliate sites often collapse the process into a spreadsheet of phrases and a browser tab full of products.

Each stage can reject the idea. That is valuable. Research saves time when it prevents a weak project, and it makes success less dependent on publishing more words. The lesson was not that technical SEO or persuasive copy rescued the site; the copy followed the evidence. Once demand, intent and fulfilment were aligned, the remaining work became a matter of execution and measurement.

That failed affiliate site remains one of my most useful projects because it showed how easily a personal story can become an unsupported market claim. I now treat research as the bridge between an experience and an offer. The bridge must carry real demand, a believable next step and a transaction that can actually be completed.