We evaluate the effectiveness of advertising sites on the Internet. Easy!

  • Tutorial
Everyone who has ever launched media or contextual advertising campaigns was faced with questions: what sites to choose and how to evaluate their effectiveness in order to carry out reorganization after a test campaign?

As a rule, before the start of the campaign, the advertiser collects a ton of data about the site, and after the end of the campaign he gets another ton of statistics. And then the fun begins - what to do with it?

Of course, you can figure out which site is better, but how to really calculate how much it is better? 2 times or 3.5?

In fact, this can easily be estimated by rating the sites using an applied methodology based on a simple mathematical model.

So, initially, for the correct assessment of the effectiveness of an advertising campaign, its goals are determined. They may sound different:
  • Introduce potential buyers with a new promotion / product.
  • To increase the recognition of the company on the Internet.
  • Increase sales.
  • Increase traffic to the site as a whole.
  • And so on and so forth ...

Of course, the whole set of goals can be reduced to two - “Sales” and “Recognition”, but to select sites, the goal should be formulated more precisely - this will help to choose the data necessary for analysis. Of course, for different purposes of the advertising campaign, the weight (value) of the criteria for selecting sites will be different.

The selection criteria for ranking can only be those indicators that can be measured. Moreover, for the initial selection of sites, the list of such criteria is shorter, and the data for them is provided by the advertising sites themselves. Typically, these are:
  • The number of impressions.
  • Cost per impression or cost per click.
  • CTR
  • The number of clicks.

The last two indicators should be treated with caution: firstly, they are predictive, and secondly, they largely depend on the quality of your ad.

After conducting a test advertising campaign, we will have much more criteria for the effectiveness of sites for their subsequent rehabilitation:
  • Actual ad impressions.
  • The number of clicks from an ad.
  • The total cost of an advertising campaign.
  • The number of actions we want (views of landing pages, purchases, the number of visitors with the time we need spent on the site) is the number of conversions.
  • CTR of a specific ad, banner.
  • Percentage of conversions (note that there can be several goals, respectively, and there are also several conversions).

Now we proceed directly to the methodology:

Stage 1. Normalization of parameters.

At the first stage, all the data (the number of impressions in units, cost in rubles, CTR as a percentage, etc.) must be reduced to a single value that can later be added up.

Adding rubles, interest, and thousands of impressions is likely to be a bad idea. :)

You need to bring them to a single value by normalizing each parameter on each site.

We will normalize either to the maximum or to the minimum, depending on the parameter. If the parameter “the more the better” (for example, this is the number of impressions of an advertisement), then we will normalize to the maximum. In the opposite case (most often these are cost indicators: the lower the cost per click, the better in 99% of cases) we will carry out normalization to a minimum.

Normalization to the maximumFor example, for the number of impressions.

Initially, we have the following data:
image

In this parameter, Site No. 3 showed its best. We take it as the basis for normalization.

Further, we divide all the data on the basis: the estimate for site No. 1 is 10 000/12 500 = 0.8, site No. 2 is 5,000/12 500 = 0.4, site No. 3 is 12 500/12 500 = 1. B a cell with a normalization base always turns 1.

So, we have a formula for normalizing to the maximum: IndNorm = Ind / MaxInd , where
  • Indnorm - the normalized value of the criterion
  • Indus - the initial value of the criterion
  • MaxInd - the maximum value of the criterion for all sites

As a result, we get the following table:
image

Normalization to a minimum is identical.
image

At the cost of the click, the most profitable was Site No. 2. We take it as the basis.

It is important to note that when normalizing to a minimum, we divide the basis on the indicator values ​​for each site according to the formula IndNorm = MinInd / Ind , where MinInd is the minimum value of the criterion for all sites.

Evaluation for site number 1 - 7/10 = 0.7, site number 2 - 7/7 = 1, site number 3 - 7/8 = 0.875. In the cell with the base - again 1.

Summary:
image

2 stage. The choice of parameter weight.

Weight is determined based on the goals of the advertising campaign. The more important the parameter, the greater the weight. The sum of the weights should be equal to 1.

For example, we have the cost per click, the number of transitions, the number of impressions and conversions.

The goal of the campaign is to increase conversion and minimize its cost. In this case, the highest weight is assigned for the number of conversions, for example, 0.4. The rest is distributed according to other parameters, for example: cost per click - 0.3, number of clicks - 0.2, number of impressions - 0.1. In total - 1.

If the goal of the campaign is to catch traffic to the site, then the number of clicks will be at the forefront, if recognition and branding will be the number of hits, etc.

3 stage. Summarizing.

We multiply the normalized scores for each criterion by the weight of this criterion and summarize the scores for all the criteria / indicators. If the weight of the criterion “Number of impressions per period” was 0.2, and the criterion “Cost per click” was 0.8, then in the end we will get the following rating:
image

For convenience, you can multiply the rating by 100 (or by 1000) to get “ beautiful ”rating values:
image

According to the results of rating calculation, we can clearly see that Site No. 3 was the best, Site No. 2 was very slightly worse (only 2 points), and Site No. 1 was 20% worse.

In the example, we compared the sites by two parameters, in practice, we recommend taking at least four parameters.

What does it look like in practice?

Let's say we are faced with the task of increasing the number of calls to an offline store. We conducted a test advertising campaign and now we must compare advertising platforms and, having excluded the ineffective ones, continue the campaign with the effective ones.

We have collected the following data:
image

1. From the available data we select the criteria that are most relevant to the goals of our advertising campaign:

  • Clicks - clicks on advertisements. The simplest criterion: the more clicks from this site, the more attractive it is for us. Data is provided by advertising sites.
  • Cost per visitor . The budget of an advertising campaign is always an important criterion.
  • The number of visitor actions we need (conversions). We summarized all the visits with the necessary actions - from viewing a page with a coupon to viewing 3 or more pages from a catalog.
  • The cost of one action . Criterion for the cost of a quality audience.

image

2. We normalize the data:
image

3. Based on the goal of the advertising campaign, we set the weights of the available criteria:
image

4. We superimpose the weights of the criteria on the normalized data and get the following:
image

5. Summarize the estimates according to the criteria for each site, multiply by 100 and get the final site rating:
image

Thus Thus, to solve the task set for the advertising campaign (to attract customers to an offline store), the VKontakte site is best suited with a significant margin, in second place is a banner on the women's portal, in third place is Odnoklas snikes. And it is precisely on VKontakte advertising that the advertising budget should be directed first.

Of course, even a priority site can run out of resources on impressions / clicks. Therefore, the rest of the budget must be allocated in accordance with other tasks of reaching the target audience, taking into account its socio-demographic factors, solvency and other things. You can simply add the necessary (depending on the goals of the campaign) criteria for evaluation and create a new rating for secondary tasks to determine the sites of the "second wave".

Also popular now: