Yandex.Direct Quotes against negative keywords or how to separate pearls from dung

The goal of setting up any advertising campaign comes down to the advertiser’s two main aspirations.
- Clearly aim at the target audience, squeeze maximum visitors out of it to your site (if someone from the target audience requests a keyword - we must do everything to get it to our site)
- In no case do not allow impressions to a non-target audience (shows of this audience will lower the precious CTR and increase the cost of a click, and possible random clicks of this audience will eat a piece of the budget)
To accomplish these goals, Direct provides us with 2 tools: Operator "quotes" and negative words. About them and speech in this article.

First, a little theory.
Extract their Yandex.Direct help

" " (). () , , .
«-» (-). , . «-» , . .

Example:
Construction: pink elephant-repair-maintenance will give impressions on requests for the sale of a pink elephant , rent a pink elephant and all possible requests with this phrase except for the repair of a pink elephant and service of a pink elephant
The design "pink elephant" will give impressions only for two of these words in all word forms ( "pink elephant" , "pink elephants" )

Negative words

')
The logic of a novice advertiser when creating an ad in Yandex.Direct is simple and insanely erroneous. Here is an approximate course of his reflections. “I sell cars (Ford Focus, for example). Contextual advertising shows ads only my target audience. My target audience is those who are looking for a Ford Focus. I take this keyword, make an ad - and get coverage of all who are interested.
However, immediately after the start of such a ROK, the necessary rates in any competitive topic creep up sharply, the CTR floundering around zero.

The reason is simple. If we look at WordStat - we see a huge amount of thematic garbage in the Ford Focus request (a request in conjunction with the words "spare parts", "service", "repair", "forum", "club", etc. - absolutely untargeted audience)

The logic of a more experienced advertiser. “Yeah, I looked at WordStat. I see a lot of garbage there. Yeah, we don’t deal with the service - minus, we don’t have spare parts ... the forum is not that, the photo is not that ... this too, this too, this too ... this too ... phew, tired ... oh, the limit on the number of minus words is reached ... ok , I'll go smoke"
Minimizing all garbage on any request is simply impossible. This is an axiom, accept it.

Also, do not forget that in each, even the narrowest request, there is a lot of thematic noise not detected by statistics (WordStat shows requests that are regularly requested from about 5 people per month)

Even if you sell some exclusive Ferrari California and otminusovali long footcloth from any junk words - there are always a dozen idiots who ask for something ridiculous and not at all obvious at first glance. ( “Repairing a Ferrari California at home , ” “where they give a new Ferrari California for free” - yes, yes, there are such requests too). Due to the rarity of such requests, they are not detected by Wordstat, it seems to the advertiser that the request is free from thematic noise. However, it is not.

As a result, any such garbage request will underestimate your CTR and overestimate the cost of a click, which in highly competitive commercial topics is simply fatally overwhelming a substantial part of the budget.

How to avoid it?

Quotes

Below is a discussion of the strategy that we widely use for clients. It is not suitable for campaigns whose goal is to collect the cheapest possible warm traffic and for campaigns in low-competitive topics (labor costs are not comparable with the profit received). In these cases, it is sufficient to use a long wrapper on the obvious campaign negative keywords. But in cases where your competitors don’t give rest, and the price per click is over $ 5-10, it will be irreplaceable. It will be about the correct use of quotes.
Take the request "Ford Focus"
This is an HF request. It is uniquely quotes and added to the Republic of Kazakhstan. However, if we add only it in exact correspondence, in this case many target audiences are left overboard. It is necessary to break it into target mid-range requests. We go to WordStat and choose what we need.
ford focus buy
ford focus sale
Ford Focus authorized dealer
etc. throughout the semantic core.

We download these requests and add them to the Republic of Kazakhstan. We take the original unquoted requests and again we torture WordStat. We pass through the semantic core of each of these queries and crush again.

Ford Focus sale turns to

Ford Focus 2 sale
Ford Focus 3 sale
ford focus car sales
etc

We quote these requests, we add them to the Republic of Kazakhstan, we divide them further.

Ford Focus car sales turns into

Ford Focus Car Dealer
Ford Focus Sale of new cars
ford focus hatchback car sales
etc

At this stage of fragmentation (the average frequency of requests is less than 30 according to WordStat), we stop. There is nowhere to split further, and Direct will not allow to create a keyword longer than 5 words. If we have overdone it - and the forecast of hits for some query was 0 - then we remove the quotes, we don’t need a dead load in the Republic of Kazakhstan.
The resulting footcloth through the AutoCorrect is adapted to other sold Ford models and also added to the Republic of Kazakhstan. Do not forget about the English writing model.
It is possible that as a result, the number of keev for an ordinary RK ordinary car dealership will exceed a couple thousand. This is normal. It is better to spend a day or two on setting up the Republic of Kazakhstan than to spend the lion's share of the budget for nothing every month and not even suspect it, as it happens in most cases.
Under each request from the resulting long footwoman you need to make a unique announcement. In most cases, it is sufficient to uniquely identify the headline and ensure that the request is included in the ad header. This significantly increases the CTR, plus it reduces the risks that our Central Bank will pass by this request. (especially in the case of low-frequency queries and indistinct search results), since Yandex allocates bold text and the title of the ad when it coincides with the query text.
Also, the creation of a unique ad for each request is an ideal opportunity for audience segmentation (the authors of the requests “sale of refrigerators in Moscow” and “Samsung RL39THCTS online store” have varying degrees of readiness to purchase, and they need to be processed on the website differently)

Only such an approach is guaranteed to solve 2 tasks indicated in the introduction to the article: it does not allow for any one demonstration of knowingly untargeted, as well as muddy and undetectable statistics of the audience and squeezes the maximum clicks out of the target audience.

Good sales!