Search in Azerbaijan runs in three languages at once
The same product gets searched here in Azerbaijani, in Russian and in English, and those three result sets rarely contain the same websites. Russian still carries a large share of commercial searching among older buyers and inside corporate procurement. English carries anything technical, anything aimed at the energy sector, and most of what international suppliers look for. Azerbaijani carries everyday consumer demand and is growing into categories that used to be searched in Russian. An agency that treats one of the three as the market and the other two as translation jobs will build a site that competes for a fraction of the available demand.
The split also decides which engine you answer to. Google takes the overwhelming majority of sessions, but the Russian language slice is where Yandex actually turns up in server logs, and Yandex has its own webmaster console, its own crawl behaviour and its own patience for thin pages. If your buyers search in Russian, ask whether the shortlist has ever verified a property in Yandex Webmaster and what they changed as a result. Most will tell you the engine is irrelevant nationally. That answer is correct about the whole market and wrong about the segment you may care most about.
The alphabet gap that keyword tools quietly hide
Azerbaijani uses a Latin alphabet with letters that phone keyboards make awkward to reach, and real users drop them. They type a bare vowel instead of the schwa, plain s and c instead of the versions carrying a cedilla, plain g instead of the one with a breve. Search engines absorb much of this, but research tools do not merge the variants, so a term that looks close to zero volume in an export is often demand split across four spellings. Briefs written straight from a tool report therefore underestimate whole categories and point content at the wrong questions.
The same problem appears inside your own site. Slugs, internal search, filter labels and structured data all need a decision about which form is canonical, and the decision has to be applied consistently or you end up with duplicate pages for the same query in two spellings. Ask an agency to show you how they handled it on a live site. Anyone who has genuinely worked in this market will have a rule ready, because they have been burned by it.
Map results are thin here, which is exactly why they are winnable
Business listings in this market are patchy. Categories are mislabelled, opening hours are stale, and plenty of real companies have no verified profile at all or have one created years ago by a customer. For a local service business that is an opportunity rather than a problem: a complete, correctly categorised profile with genuine photographs and a working phone number can outrank competitors who spent far more on their websites. The work is unglamorous, and that is precisely why so few local firms have done it.
Verification is the friction point. Address formats vary, buildings are frequently identified by landmark rather than by a clean street number, and postal confirmation is unreliable. An agency that has done this before will know which verification route currently works and will tell you how long it takes rather than promising a result next week.
Domains, hosting and the speed you can actually deliver
Local extensions carry trust with domestic buyers and require documentation to register, which is a mild deterrent for spam and a mild annoyance for you. The more consequential choice is where the site is hosted. Serving a locally targeted site from a distant data centre adds latency that shows up in field performance data on mobile connections, and mobile is where most of this traffic lives. Content delivery coverage in the region is thinner than in larger markets, so gains that come free elsewhere have to be engineered here.
If the site serves both a domestic audience and an international one, the architecture question has to be settled before anyone writes a word. Separate language folders under one domain, a clear canonical, and correct language annotations will outperform a pile of half maintained country sites every time. Ask the agency to draw the structure on one page. If they cannot, they have not thought about it.
Where the commercial demand sits, and where it only looks like it does
Volume and money are not in the same place. Energy services, logistics, construction supply, financial services, education agents and inbound tourism carry real budgets and comparatively little organic competition. Broad consumer terms look attractive in a keyword report and convert into nothing because the audience is price shopping across regional marketplaces and social feeds rather than reading websites. A partner worth paying will push you down into specific, low volume, high intent phrasing and will be comfortable telling you that the headline term in your brief is not worth chasing.
Regional intent matters more than the size of the country suggests. Searches that name a district or a regional centre behave differently from searches that name the capital, and the pages that serve them need genuinely different content rather than a name swapped in a template.
Links in a small publishing market
There are only so many credible domestic publishers, and most of them will talk to you about paid placement before they talk to you about your story. That is the working reality, and an agency that pretends otherwise is either inexperienced or planning to spend your money without telling you. The honest approach is to separate the two budgets, label every paid placement in reporting, and put the larger share of effort into assets that earn coverage without payment: original data about the local market, useful reference material in Azerbaijani that does not exist yet, and partnerships with associations and universities.
Be sceptical of link volume promises. Networks of low quality sites are cheap to buy and easy to detect, and the clean up costs more than the links ever returned. Ask for the last set of domains an agency placed for a client in your sector and look at them yourself.
Questions that separate a real SEO agency in Azerbaijan from a reseller
Ask who writes the content and in which language they think. Ask whether the technical work is done in house or subcontracted abroad, and if it is subcontracted, who talks to your developers. Ask for a sample monthly report and check whether it reports business outcomes or activity counts. Ask what they would stop doing if the budget were halved, which is the single fastest way to find out what they believe actually drives results.
Then insist on ownership. Analytics properties, the search console, the advertising account and the content management logins belong to you, created under your own email domain, with the agency invited as a user. Agencies that resist this are protecting leverage rather than your data, and the cost of that arrangement only becomes visible on the day you want to leave.
Comparing proposals without comparing sales decks
Send the same short brief to three firms, including your actual commercial goal, your languages and your internal constraints, then compare their answers side by side rather than reading each document in isolation. The useful differences show up in what each one chose to ask you, not in how the slides look. One will propose a technical rebuild, one will propose content, one will propose links, and the reasons they give will tell you how they diagnose.
If your requirement stretches beyond search into build work, the market also has development teams worth reviewing on the local software company listing, and the wider search marketing category is useful for seeing how firms position themselves across other markets before you judge a local shortlist. When you have three you trust, request comparable proposals so you are weighing scope and staffing rather than presentation.