Hot keys in rya. How many keywords are optimal for YAN? Mobile ID targeting

Hello friends!

I decided to enlighten you a little and tell you about how YAN works in Yandex.Direct at the moment and what phrases are needed for effective work campaigns. In general, the theme of the advertising network is very popular and is overgrown with a large number of myths, fairy tales and tales. Hence, respectively, different approaches to setting up a campaign: some do it this way, others do it differently. It is quite difficult for a beginner in such a fuss to find a specific solution.

How does RSS work?

Yandex.Direct is a system with its own specific algorithms, formulas, and other “tricky things”. There is a place for modern developments in the field of IT, which are actively used by the system. We can safely say that the YAN uses machine learning technology.

From here came the method of setting up a campaign based on high-frequency keywords Oh. In Lesson 11, I talk about this:

Keywords must have at least 200 impressions per month! This is necessary for a larger audience coverage.

Thanks to machine learning, the system understands the subject of the advertised product based on key phrases and selects for display the audience that would be interested in it. Moreover, not only the target audience is covered, but also “similar” (who else would be interested in our product).

However, it should be understood that not only our “keys” are used for this, but also data from other sources. Now I will explain.

There is such a term - Big DATA - unstructured data of huge volumes. Data about everything in the world: about you and me, about our interests, about our field of activity and about everything else that surrounds us. Large corporations (including Yandex) have learned how to collect Big DATA, how to analyze and use it to achieve business goals.

This data is collected from various sources. As an example, the Yandex.Metrica counter that everyone installs for free on their sites, they get good functionality of the analytics system (which, by the way, could be paid), in return they transfer data about visitors to Yandex: geography, demographics, behavior, interests, and so on. Then all this is used to develop advertising technologies, create new products, improve the algorithms of the search engine and the operation of advertising services, which results in a very good profit. Naturally, one metric is not enough here, but that's a completely different story.

In general, everything that we do on the Internet, everything that we are interested in is used. Used to improve efficiency advertising systems- the main source of income for search engines. You probably noticed this garbage - you went to a site about building houses, then you went to Gismeteo and you are shown ads for house building companies, or some related product. Noticed? I see this often.

Do not discount thematic targeting - this is when ads for bricks are shown on a construction site. This was once the main method of working with the audience in ad networks, now it is fading into the background. Still, the behavioral shows itself much better.

Conclusion: as soon as the picture about you is formed, the system will start to catch up with advertising that is potentially interesting to you. You will naturally click on it and possibly purchase the product. It works like this.

More about keys

If you have ever assembled a semantic core for contextual advertising, then you know that out of the total number of high-frequency requests, 10-20 percent remain. Somewhere more, somewhere less, and somewhere there are 1-2 of them at all. And these 10-20% are enough for the YAN campaign to work effectively. The guys from Yandex confirm this.

And the thing is that high-frequency queries help the system reach as many audiences as possible for displaying ads, the main thing is that they be targeted (this is the most important thing). Any refinement of the list, the addition of more precise phrases may cause a failure in the campaign. Here is an example — this is how the campaign with high-frequency users worked before adding new phrases:

Everything is stable: impressions, clicks, cost per click, and expenses. And then I decided to add a dozen phrases, and the phrases were hot. I thought that it would give an effect in the form of an increase in conversions, but it turned out to be quite the opposite:

As you can see, the campaign started to work unstable (yes, there were weekends), but on weekdays the picture didn’t get any better. Plus, the applications have practically stopped. I had to, as they say, “back up” back to the initial version and increase the bid so that the campaign gained momentum faster.

As a result, the rollback gave its result: the campaign began to work stably, bringing applications / calls.

• Tweets in YAN work fine;
• Any clarification, addition to the list of phrases gives either the opposite effect, or none at all;
• In which case, roll back to get the campaign back on track.

That's all for me, dear friends!

Goodbye!

The question is certainly ambiguous, because opinions on this issue are quite polar: someone says that a small number of high-frequency (HF) and mid-frequency (MF) target words are enough, others argue that it is necessary to collect fairly accurate semantics.

To express our opinion on this matter, let's first recall what is Advertising Network Yandex.

As the name implies, YAN is a huge network of sites that place on their pages advertisements Yandex. It can be both sites of your subject, and pages with non-targeted content.

Therefore, do not forget that in the settings of the campaign that you are doing for YAN, there is a checkmark to take into account user preferences or not.

What does it affect?

By default, it is not set in the campaign settings; user preferences are taken into account. Thus, your ads are shown to all visitors on sites whose topics are determined by your key phrases and to those visitors who entered these key phrases into the Yandex search line.

When might you want to disable user preferences? For example, you sell sleeping bags and want to increase the volume of visitors to your site through advertising on related topics. Then you create a campaign with key phrases, for example, "tourist tent" and check the box "Ignore user preferences." Thus, your ads will be shown on sites dedicated to tour tents. But those people who are currently viewing materials about tents, theoretically, may be interested in buying a sleeping bag. And as practice shows, they are interested.

Important reminder!

If in a regular YAN campaign (taking into account preferences) you can make 1 ad = several keys, then in a campaign without preferences, we recommend making 1 ad = 1 key! The key phrase in this case is a description of the theme of the site, if there are a lot of them, Yandex will try to select sites suitable for all the keys of this ad, which will significantly reduce the number of sites.

And now back to phrases, which ones to take - common masks (HF) or phrases of 7-8 words (LF)?

First, let's make a reservation right away that there is no generally accepted exact number of requests at which a key is considered high-frequency, medium-frequency or low-frequency. The number of requests is different for different niches and even for different regions in the same niche.

It's just that the entire pool of requests is conditionally divided into parts in accordance with the frequency of requests.

For example, we selected keys with a frequency of up to 1000 as high-frequency

With a frequency less than 200, like woofers

Accordingly, those keywords that are between 1000 and 200 are taken as midrange (MF).

In most niches, especially if you need a quick start, high-frequency and mid-frequency target phrases work better. But of course, here you also need to be smart about choosing keywords, because if you launch a YAN advertising campaign with one high-frequency sleeping bag, imagine how much junk traffic will get to your site and even, as a rule, low price clicking in YAN will not save you from "draining" the budget.

Therefore, it is often enough to take a small amount of HF and MF, but targeted, key phrases consisting of 2-3 words. The number again depends on the niche, 30-50 keys are enough for someone, and somewhere you need to take 100-300 to cover most of your topic queries in networks. We take the request “sleeping bags” as a wide mask, but we don’t use this mask itself in advertising, but we launch high-frequency speakers embedded in it in YAN.

Naturally, this strategy implies that you understand what your cost per click (CPC) is for you and select high-frequency and mid-range keywords based on this data.

As for low-frequency speakers (LF), they are added already in the next turn, with a more in-depth study of the campaign, or, for example, to appear to those users who introduced a particular product model. It turns out a kind of autoretargeting. Again, very narrow target phrases like “where can I buy a sleeping bag with insulation near the Butovo metro station” should not be taken, because. According to the recently introduced Yandex rules, keywords with a display frequency of 1-5 per month receive the "Few Requests" status and are not shown. Therefore, there is no point in messing with them.

Summing up, we recall that if you want to launch quickly and as efficiently as possible, then at the first stage you should not be smart and fill in tens of thousands of keywords in YAN. Test a niche or destination with 200-300 phrases - often enough to understand if the game is worth the candle. And if so, the further creative process is largely dependent on many variables.

1. Separate the campaigns and the semantic core into search and YAN.
2. Add "non-obvious semantics".
3. Target impressions through Yandex.Audience.
4. Optimize campaigns through Yandex. Audiences.
5. Set up analytics with the Callibri service and check how your hypotheses work.

Let's analyze each step in detail.

Different campaigns and semantic core on search and YAN

The worst mistake is to connect the same campaign for search and YAN.

Ads and targeting work on search and in YAN according to completely different principles. Let's first understand the types of targeting in Yandex.Direct.

So why can't you just take and copy the campaign from the search and show it on YAN?

Let's remember how we behave on the Internet

Differences between ads on search and in YAN

The use of different targeting causes different contexts for the visitor. During the search, he himself is interested in clicking on an ad that matches his request. When a visitor simply “surfs” on the Internet or sees an ad in his mail that we show him with the help of YAN, he may have already forgotten his request. And being, for example, in the mail solves some other problem. Therefore, in order to attract his attention, we need to build communication in a completely different way.

Differences between search queries and YAN

The semantic core is the key queries by which Yandex understands who to show the ad to.

Non-obvious semantics

Yandex Research

In 2013, Yandex conducted a study during which it found that more than half of search queries are unique. Different people queries are worded differently. By using long tail queries in YAN, you are losing a huge amount of your audience.
If you use request masks in YAN, then you get access to the entire audience, no matter how uniquely it formulates its request.

So what is a Query Mask?

Query mask - the widest, high-frequency query, consisting of two words. These two words may not coincide in meaning, but when they intersect, they give us access to the desired audience, since they are query markers.
991 bytes

"Burn the grass" - what is this query about? Who is recruiting him? Gardeners, pyromaniacs or pyromaniac gardeners? In fact, this is the topic of weight loss: the query "Fat Burning Herbs"

How to find hundreds of the most effective queries if you are not an expert in the subject

For example, we want to select a semantic core in YAN for a company engaged in the wholesale of bags. Without being an expert in a niche, it is very difficult to pick up a semantic core. The first thing that comes to your mind is high-frequency commercial queries. That is, those that all your competitors use. Therefore, the cost per click will be very high. What happens if we use request masks?
991 bytes

In the wholesale of any product, markers are: wholesale, price, distributor, dealer, supplier, customs, customs clearance, markup, retail. These words are the first part of the markers. The second is the names of brands of bags: Bvlgary, A. Valentino. By multiplying these queries among ourselves, we reach the audience of visitors who are interested in selling bags in bulk.
991 bytes

Yandex.Audience: even more from YAN

Segment from Yandex. Metrics

Internet shop of women's bags. We segment bags by color, set up retargeting so that every buyer of a brown bag, for example, 20 days after the purchase, sees an ad “Brown gloves will match your bag.”

• You can find an audience similar to your customer base. This targeting is powered by Crypto Technology. The technology analyzes each of the 1000 clients and finds patterns, patterns in behavior: what sites they visit, what their interests are. And then, of all Yandex users, it tries to find the same ones.
• Or find an audience similar to your website visitors

Mobile ID targeting

This is suitable for those who have mobile app, for example, for regular guests. Your IT specialists can find out the ID of the mobile devices from which your application was installed. You upload a list of these IDs to Yandex.Audience, and Yandex can:

• show ads to these users
• find similar users and show ads to them

Hyperlocal targeting

• Located in this location now
• Stay at this location regularly.
• Were here N days for a certain period

We can both turn off impressions for unwanted audiences, and raise rates for the audience that is most interesting to us.

Case.

The clinic found that the visitors to their site are both male and female. However, women make an appointment in 90% of cases. After all, it is women who most often take care of the health of the family and make such decisions. That is, men clicked on ads, spent the budget, but did not buy. And the female audience converted many times better. This is exactly the case when you need to completely exclude men from showing ads using bid adjustments. Or lower rates on them, reducing budget spending.

Banks have access to information about people who have bad credit. In order not to show loan advertisements to such people, the database of “bad” borrowers was uploaded to Yandex. Audiences and set the bid "0" in front of this audience.

Using Callibri to set up YAN

Case.

Tour operator "Lucky case" ordered the setting of contextual advertising. There were few sales. We started to figure out why. It turned out that there were a lot of advertising calls. Several calls from the office were listened to with the head of the company together.

Client: Hello, we have a company of 12 people, we are interested in a New Year's trip to Deer Streams.

Manager: Have you already decided on the dates and type of rooms?

Client: No

Manager: Call when you decide.

And this was not an isolated case. We calculated: as a result, managers "leaked" sales for a total amount of 400 thousand rubles.

Also, using Callibri, you can analyze search queries for which there were conversions. In you see the request for which there was a conversion

The approach to choosing key phrases for YAN is fundamentally different from that for Yandex.Direct search. It is extremely important to start collecting keys with preliminary analysis target audience and drawing up a portrait of a potential buyer. Why and why?

If some request seems controversial to you, check it through Wordstat - how much the interests of users in the issue coincide with the offer.

After that, work with the keys to add negative words, but you need to do this very carefully, since the wrong actions can cut off most of the target audience. To do this, we will use Direct's Budget Forecast (https://direct.yandex.ru/registered/main.pl?cmd=advancedForecast).

In the window, enter the first key phrase from the HF list and click "calculate".

In the pop-up window, we look through the nested phrases and explicitly (!) mark non-target phrases in the checkbox. As a result, a minus word is added to the original key and it will not be shown for this query.

We copy the received from the line and return it back to Excel. Do not add more than 2-3 negative words in one key, otherwise you will significantly reduce the audience potential buyers by making the wrong decision.

Next, you need to remove the “+” sign before the stop words in all keywords where it is present. To do this, in the upper right corner of the Excel file, click “find and select” - “replace”, write “+” in the upper field, and leave the bottom blank, “replace all” and the pluses will be deleted.

On this HF page is ready. I don’t know how many keywords you can collect for YAN, the more directions you find, the wider the list will be and, accordingly, the reach of the audience, I did everything just for an example, without working through the phrases properly.

We do the whole operation with a group of the remaining mid-frequency and low-frequency queries up to 50 per month, moving the data to a new page. When viewing the list and deleting non-target keys, transfer related keys that come across to the appropriate page.

After sorting the column by the number of words in the phrase, we leave only those that have no more than 4 words, and delete the rest, as they will only slow down the campaign due to poor response.

For MF and LF keys, you should not add negative words at all, but they must be grouped according to semantic or some other criteria for individual audience segments.

This can be, depending on the offer, the word “male” for targeting by gender, “rubber” for the material of manufacture, “Moscow” for geographic affiliation, etc.

For example, I see the word "buy" in the list - these are hot requests and it is worth combining them into one group. I won’t describe how to do this in Excel, if you don’t know, look for the information yourself.

Thus, we distribute all the MF and LF keys on the page as separate ready-made bundles for setting up ads for them.

YAN target keywords are collected, repeat process #1 with near-topic (related) queries on a separate page, without grouping them into groups yet.

In the next article, we will look at what goals are needed for and group them by type of conversion.

Typically, search advertising campaigns use a one-keyword-one-ad approach. Thus, it turns out to create ad texts that best match the query potential client. What should be done with campaigns created for impressions in YAN, where there are no search queries? Let's clarify that we are talking about YAN campaigns based on thematic (rather than behavioral) targeting.

Initial data:

You can often hear the opinion that an advertising campaign for YAN should contain only high-frequency queries, and it is not recommended to use the syntax (quotes, exclamation mark, + sign) in phrases so as not to limit the number of suitable sites. However, the use of special characters will be justified in some cases. For example, if you sell sightseeing tours to St. Petersburg, then you just need to fix the preposition "in" so that your ads are not shown on sites where it is about searching for tours from St. Petersburg.

As we can see, there is a single rule by which the semantic core for YAN is formed - requests (key phrases) for setting advertising campaign, - does not exist. Each specific case, each individual topic will dictate its own conditions, so you should not blindly rely on random recommendations.

Decision

What queries will we include in the semantic core of YAN?

When ads are shown in YAN, the content of the site is analyzed for compliance with the given keywords, so the use of low-frequency queries will not play a big role in the formation of the semantic core. However, high-frequency (HF) and mid-frequency (MF) requests will fit perfectly. To collect HF and MF requests, it is very convenient to use the Yandex.Wordstat service, and be sure to pay attention to the right column as well. If before creating a campaign for YAN you managed to prepare a campaign for search, then you can immediately take the keys from the already collected database.

As for negative keywords, oddly enough, even in the field of bathroom renovation, they will be useful to us:

When forming a group of keywords, we often use the following approach: we divide the semantic core into several thematic groups. For example, bathroom refurbishment services may include plumbing, pipe replacement, tiling, and more. Therefore, in addition to a group of general queries like " bathroom renovation», « renovate the bathroom», « bathroom renovation” and others close to them, it makes sense to single out several small groups of key phrases by the name of individual services and write different ads for them that are as relevant as possible to the topic.