Search

This report shows the most frequent searches and enables a quick review of them and a fast way to add semantic filterings.

Element	Use
	Allows you to select the amount of results you would like to view.
	Allows you to select two dates to form a range of time in which you would like to get results for.
Search Query	This column lists search queries, with the total number being in the last row. You can also click on any search query to view a test request of said query. You can filter the listed queries by typing into the input box on the first row.
Results	This column displays the number of results. They may take some time to load though. You can enter two numbers in the first row (top=min, bottom=max) to filter the listed results by range.
Visits	This column displays the number of visits. You can enter two numbers in the first row (top=min, bottom=max) to filter the listed results by range.

All columns (except actions) can be sorted ascending or descending by clicking on the

1.2.1 Semantic Filtering

Semantic Filtering basically does what the name already says. It takes the semantic meaning of the search query and if there is a corresponding filtering option, is uses that. For example: Usually when the customer types ‘shampoo’ in the search bar, he expects to see just shampoos, but as the word ‘shampoo’ is in the description of lots of other products, these other products also appear in the product listing. If semantic filters are defined, the search term ‘shampoo’ will only show products in the category ‘shampoo’ or whatever you have defined.

To define the rules, follow this process.

1.2.1.1 Quick add

First of all, you should add all the semantic filters, that correspond one to one with a category, a brand or something similar. For this go to the general view and click on the ‘search’ – widget.

Then in the ‘actions’ – menu, select ‘Semantic Filterings’

Right at the bottom you will find a button called ‘quick add’ and a dropdown menu labelled ‘Select Field’. Select a Field, that makes sense for your shop, for example Category Id or brand and then click the button. In this example, I’ll use Category Id.

Here you have to click ‘search’ without a search query to get all of the field values, in this case, the categories. Then you go through them and select the ones, that make sense.

If the category is for example Köder, you expect, that when the searches for ‘Köder, he will only see products, that are in the category Köder. All the other products (which may have ‘Köder’ in the description), get filtered out. Based on that logic you can tick quite a lot of the cases off.

The categories, that you shouldn’t set semantic filters for, are as follows:

• Shop

• Hot topics (blog)

• Deutsch, Français, English, … (Languages)

• Accessories, Diverses (too broad)

• Hot topics (blog articles)

• Topical, Sale, Current, Christmas, Specials (seasonal/temporary)

• Brands, which are also set as categories

• …

When you have selected all of the values you want, click ‘add’.

You can check, edit, test and remove the filters, you have set, here.

Edit

Here you can edit the search query, the field and its value. You can click on the magnifying glass to search for category ids.

Test

Here you can test the effect of the semantic filtering. It automatically makes a search with the query and you can check the difference between the production version and the test version with the filter set. Normally the hitcount of the test version should be lower as you’re setting a filter, which limits the values.

Remove

This is pretty self-explanatory.

When you are done, click save and then back in the widget, click save & test.

1.2.1.2 Search report

When you have done as much as you can with the quick add, you can open the search cockpit under Reports -> Search.

Normally when analysing the searches, you want to sort by visits and descending, so you can see what most people are searching for, so you can improve the quality for the vast majority of the people.

Now it comes to filling out the queries, which haven’t got a semantic filter from the quick add. There are various scenarios which you have to consider, when doing this.

Scenario 1: Typos

A few customers don’t know how some products are spelt, or simply mistyped it when searching.

When this is the case, you can simply add the same category of the correctly spelled query.

Scenario 2: Clear matches

Some customers know exactly what they are looking for, so they type in a specific type of product which doesn’t correspond word for word with a category. For example, ‘rute’ has a group id ‘1593’ and feederrute doesn’t have one. So, to make sure the person searching for ‘feederrute’, doesn’t get just any kind of product with feederrute in its description, but actual fishing rods, you should define a semantic filter for it.

To do this you have to click on the blue link ‘none’. That opens the semantic filter menu.

Click on the pen icon to edit the filter for that query.

Fill in the relevant data (e.g. Field = Category Id, Value = 1593)

Then go to the show columns menu to show the ‘skip keyword’ column.

You have to set that to false, otherwise when the customer searches for ‘feederrute’, he won’t only find those, but also all the other products from the category ‘rute’ and the goal is to limit the values, not extend them.

When you have done and saved this, you can click on the tick in the action column and test the changes, that you made. Like before, the goal is normally to limit the results.

Scenario 3: Multiple categories with the same name

If a customer searches for a very generic term (e.g. Stiefel) and there are many sub-categories called Stiefel (e.g. Reiter/Stiefel, Regen/Stiefel, Jagt/Stiefel), you can add all of these ‘filters’ to the same search term.

Make sure that the ‘or’ clause is enabled.

Scenario 4: Search Quality

If none of the previous scenarios applied, you can filter the products by checking the quality of the correlation of the search query and the products. To do that you would first have to check in the test view, if the filter would make sense. In this example I will use the search query ‘wecker’.

If the customer would search for wecker. The fuzzy search would realize it can split the word after weck and therefore would show you products called weckglas, which you don’t want. To filter out these products you have to follow these steps:

First open the test view

When you have done this refresh the results by clicking on the button at the top

Then you have to scroll down until you see a quite drastic jump in the score like as follows

You will notice that the second product isn’t an alarm clock and has more values for the fields we added above.

The more products that have these values, the clearer it should be, that the filter should be applied.

The first value means that the search term doesn’t appear in the product title and the second one means, that that the product doesn’t correlate with the products that should be appearing. It happens quite often, that there are products, which don’t have the search term in the title, so if you’re unsure whether to apply the filter for the search term or not, you can check by simply searching for the product in the store and deciding if the product fits the search term or not.

If you decide to apply the filter you can follow the procedure of adding a semantic filter (Scenario 2: Clear matches). There are some important differences, you have to take note of:

The symbol should be not equals as those are the product that should not appear. The keyword should not be skipped and the or clause should be set to false if you chose to filter by both filters. Then you can save and test the filter(s) like before.