Basics of advertising campaign media planning and analysis of the semantic core in Yandex.Direct, or how not to drain half of the advertising budget into the toilet

Recently, another advertiser came to us. The field of activity is the sale of LED equipment - large spotlights and LED strips for street lighting and illumination of buildings and shop windows. Accordingly, we are talking about B2B, the client is interested only in large orders in its subject matter from construction companies, management companies, the housing and utilities sector, and “not at all interested in those who want to buy a couple of lamps in order to make beautiful lighting on their balcony.”

The topic is quite narrow, the market is mature, medium competitive, with steady and formed demand. The client has successfully and successfully traded offline and decided to try the Internet, since there is no problem with money to expand the spheres of influence. The goal is to cover the entire target audience and convert the maximum available number of clients. He brought with him a media plan calculated with the help of the Yandex Direct predictor by his marketing department.
')


Here is this media plan.



Everything is simple and clear. You pay 350 thousand per month - you get 6,700 visitors to the site. The client is quite satisfied with this situation, he is satisfied with the work of his marketing department. That is, we simply need to turn the media plan into reality.

But we have to tell the customer the shocking truth of life. 60% of the advertising budget for this media plan will be spent inefficiently, and simply “merge”

"Where? How? Why? This is my target audience! Who else will look for such requests, if not my potential clients? ”

So, we look at the order of the semantic query core. We agree to mark the target requests with a tick, cross the - non-target.

Request "LED backlight"


Immediately striking a huge amount of untargeted audience in the semantic core. Out of the total volume of requests for 4790 hits, only 5 requests are targeted for us with a total of 917 hits per month. Everything else is rubbish. Trash audience is diverse. These are motorists who are looking for an opportunity to tune cars (LED car lights, LED interior lights, VAZ LED lights, LED instrument lighting), and those who are looking for an opportunity to make beautiful lighting at home (LED ceiling lights, LED interior lighting) light and who already have the lamps and they don’t know what to do with them (installation of the LED backlight), and technomaniacs looking for a new device (laptop with LED backlight, LED TV). None of them will ever make a multi-thousandth order for a couple of thousand LED lamps for street lighting, advertising displays of this audience are absolutely harmful for the client’s RC.

Request "LED lamps"


Here the situation is complicated by various abbreviations. Quick googling helps to understand what is Central Asia, and what is Central Asia. H4 - lamps in car headlights (obviously garbage), e27 - type of lamp socket of our client (CA). Do not save time on googling of this request and analysis of issue. Do not save time on full immersion in the subject. Often, even a guru in any business topic cannot, by the appearance of a request, determine what exactly a potential client wanted to find with this phrase.

Another small remark about the query "buy (buy) LED lamps." The query contains the word. In 99% of cases, the real buyer is looking for the product he needs with the initial form of the verb - to buy. In order to play it safe and prevent advertising from showing itself to the sales managers of competitors, we use the word form fixation operator (!). So, we take a request in this form in the Republic of Kazakhstan “buy LED equipment!”.

We carry out the same analysis on other requests.
Request "led strip"


Request "LED lights"


etc

So, this is how our refreshed, and optimized media plan looks like.



This media plan is not the final version (many requests need to be re-crushed, you need to pick up many more low-frequency queries, add common typos, professional slang, customer product brands and competing brands), but you can get a general idea after garbage collection. The forecast budget has lost 62.5%. But in terms of the number of requests, it has repeatedly grown fat — 270 requests instead of 6 initial ones. These 270 low-frequency queries describe our Central Asia where the initial ones are more accurate, and they require almost three times less money.

In other words, the guys from our client’s marketing department overlooked an elephant - a huge layer of untargeted audience. In total, such a mistake would lead to the misuse of the extra 220 tr monthly. This happens very often. The client is sitting on a narrow topic, and doesn’t compete with the segment in real life. And when the turn comes to advertise on the Internet, the RA will say (or will not say) that in the semantic core, according to their chosen request, in addition to its target audience, there is also a neighbor’s one and in order not to merge the budget, it’s necessary to prevent the target audience from showing audience.

Axiom 1. Displays of a non-target audience merge the budget.
Conclusion from the axiom 1. When media planning, as well as directly in the Republic of Kazakhstan, one should never allow the possibility of showing the Republic of Kazakhstan to a non-target audience.

Axiom 2. Any even the narrowest subject contains some amount of garbage.
Conclusion from axiom 2. For each query in the Republic of Kazakhstan you need to look at the semantic core in WordState.

Axiom 3. High-frequency queries always contain more garbage than low-frequency ones.
Conclusion from axiom 3. High-frequency queries can never be taken in its original form. We split the query into a set of derivatives, quoting the original query (maybe even in several iterations). This approach is described in detail here.
habrahabr.ru/blogs/context/107330

Successful sales!