Search volume and seasonality: don't be fooled by demand

Next to every keyword sits a frequency number, and it is easy to read it as demand. The temptation is simple: the bigger the figure, the more important the phrase, the more money behind it. In reality the number is a snapshot of one moment, taken without regard to season, and it says nothing about whether that demand will turn into inquiries. A large frequency can be empty. A modest one can pay the bills.

For a company selling into Russia from abroad the temptation is stronger. Where Google Keyword Planner hands you a rounded range, Yandex Wordstat hands you an exact count of impressions for the last month, for any region you pick. Precision reads as truth. It is still one month with all the noise inside it.

There are two ways to be fooled: overrating a bare figure, and believing in other people’s activity as a parsing service draws it for you. Below are the typical traps where visible demand diverges from real demand.

Frequency is a snapshot, not a flow

The figure next to a phrase was taken at one point in time. Demand breathes with the season: the same service collects a different frequency in summer and in winter, so a low-season measurement undersells the product and a peak one oversells it. Building campaign structure and budget on one month of frequency is fixing a random point on a graph and calling it the graph. What you need is the movement over a period.

The Russian calendar makes this sharper than most Western markets. The country effectively stops from the last days of December until around 10 January. Business demand sags again around the May holidays. Consumer categories follow the dacha season from late spring, and much household and service demand resets in September when the city fills up again. A snapshot taken in early January makes a healthy category look dead. One taken in the pre-New-Year rush makes a seasonal category look permanent.

Wordstat has a history view that shows the phrase month by month for two years. That is the number to put into the media plan: a budget set on the peak month is underspent for the rest of the year, and one set on the trough is starved when the season arrives.

The headline figure hides a second thing. Base frequency in Wordstat is inclusive: it counts every longer query that contains your phrase, so it absorbs all the synonym rubbish that broad match will later mix into your traffic. A large number is a ceiling from which the irrelevant part has yet to be subtracted, never a promise of quality clicks. Demand is a live and dirty quantity. Frequency shows its raw size and says nothing about how much will reach the till.

A long keyword is not hot demand

There is a standing myth: the more words in a keyword, the higher the conversion rate, so long multi-word phrases are the most valuable ones. Observation across several accounts in different niches does not confirm it. Conversion does rise when a phrase goes from one word to several. Beyond a certain length it falls more often than it rises.

There are two reasons. A long phrase does not fit into an ad headline whole, so you keep the main words, and the ad answers the detailed query worse than it could. And for a very specific query the site rarely has a dedicated landing page, so the visitor leaves: the offer was fine, the page was about something other than their case.

In Russian both problems are a little worse than in English: words are longer and inflect, so a five-word query eats the headline limit faster, and the exact phrasing a person typed rarely matches your keyword form for form. Yandex Direct matches word forms automatically, which helps with impressions and does nothing for the fit between the page and the query.

The conclusion is not that length is harmful. It is that length is not a linear law. Test the dependency on your own data and check whether the long queries have pages to land on. Where narrow queries genuinely have their own worked-out landing pages, the picture will be different. I treat Yandex Ads and the pages behind them as one job for exactly this reason.

Real demand lives in the exact match

The most valuable thing in an account is zero distance between the query and the keyword: the person typed exactly what you have as a key. Those segments carry the lowest cost per lead and the best click-through rate, conversion and position. The further the query sits from the keyword, through synonyms and nested variations, the worse every metric gets: dearer leads, lower conversion, lower position. So the value is not in the size of the frequency. It is in the core of exact matches inside it, which is the part to protect and to expand first.

This matters more in Yandex Direct than in Google Ads because there is no phrase match and no exact match as Google defines them. Every keyword is broad by default, narrowed only by operators: quotes fix the number of words, an exclamation mark fixes a word form, a plus sign keeps a stop word, square brackets fix word order. A team that ports its Google structure across, with its split into exact and phrase campaigns, arrives at an account where everything is broad and the exact-match core is not visible as a separate thing. It has to be carved out deliberately and managed as its own segment.

The same place hides a reporting distortion. A broad single-word category keyword pulls in impressions that by meaning belong to narrow subcategory keywords in other groups. The subcategory looks weaker than it is, because part of its demand has been credited to the category, and any cluster-level analysis lies. Do not believe that a subcategory “has little demand” until you have checked whether the broad category keyword is stealing it. Either remove such single-word keywords where overlapping subcategories exist, or separate the levels with cross-negatives, meaning each group’s core words added as negatives to the groups above it.

Someone else’s demand, drawn for you, is fake

Services promise to show which phrases a competitor is running on, and it is tempting to build a demand map on that. It is a mirage. When you parse a competitor’s ads against your own keyword set, what you see are your search queries, not their keywords. The fact that somebody’s ads surfaced on your ten thousand queries does not mean they have ten thousand keywords.

The Russian market has its own parsing tools, local equivalents of the Western spy services, with the same flaw amplified by how Yandex Direct works. Autotargeting and broad match show one ad on a heap of random queries and send them to random pages. The results are personalised on top of that and change almost immediately. A single-word keyword in broad match triggers impressions on nearly any query, so a domain collects the same ads over and over, most likely from one group. You cannot reconstruct a competitor’s semantics from that, nor their group structure and landing pages. The only source of certainty is direct access to their account, which you do not have.

This does not make the exercise useless. It moves the focus from stealing keywords to studying the landing pages and the claims of the category leaders. For a company entering Russia that is often the more valuable output anyway: how local leaders phrase the offer, what they promise on delivery and payment, what they put above the fold. Just do not pass off the parsed list as a map of demand, and never build a promise to a client on it.

The demand you lose in silence

The reverse trap: you think you are covering a query, and there are no impressions on it. This is the classic Yandex Direct square-bracket problem. It has no equivalent in Google Ads, so teams coming from Google walk straight into it.

If every word in a phrase is bracketed, a preposition inside the phrase is fixed as a required part of the query structure, whether or not the preposition itself sits in its own brackets. The phrase [окна] в [москве], windows in Moscow, shows only on queries where “in Moscow” is present with the preposition. On the query окна москва, windows Moscow with no preposition, which is how a large share of Russian users actually type, it never shows at all.

Across an account this is a noticeable share of lost impressions. On the highest-frequency phrases the losses are multiples of what you see, and the reports do not show them until you open the search queries deliberately. The check is simple: in the search query report, find the bracketed phrases and ask whether the matched queries ever omit the fixed preposition. If they never do, that is missed demand you closed off to yourself with punctuation.

The wider lesson for a foreign advertiser: Russian prepositions are stop words for Yandex, dropped from matching unless pinned with a plus sign, and the operators interact in ways the help pages describe one at a time. Before launch, someone who reads Russian has to check operator use in every group.

What to check in your demand this week

• Look at the movement, not the point. Rate frequency over a period and by season, using the two-year history in Wordstat, and check which month your snapshot was taken in.
• Do not pray to keyword length. Check on your own data where conversion stops rising, and whether the long queries have pages to land on.
• Protect the exact-match core. Separate the zero-distance segment and manage it apart from the synonym tail.
• Catch the cannibalisation. Check whether a broad category keyword is stealing impressions from the subcategories, and add cross-negatives where it is.
• Open the brackets. Find the phrases in the query report where a fixed preposition is cutting impressions.

Demand is not one number. It is a live quantity that moves with the season, gets distorted by match type and hides behind other people’s ads. Bare frequency shows the size of the shadow rather than the object. Whoever reads the figure literally pays for rubbish and loses what they thought was covered. Whoever looks at the movement, the exactness of the match and the origin of the data sees demand as it is.

If it looks as though there is plenty of demand and few inquiries, bring the account to a free review: I will show where the frequency is lying and what is really behind it.

Frequently asked questions

Sources

• Russian version of this article on belokrylovo.ru: Частотность и сезонность: как не обмануться спросом