Keyphrase detection & clustering

One of the most crucial steps towards in-depth insights into your textual data is the design of the codebook. At the same time, it is also one of the most challenging ones: Creating a meaningful, well distinguishable set of codes without knowing the data can be tedious.

We already partly addressed this challenge through our high-quality codebook templates. But we wanted to be even more helpful:

That's why we added clustered keyphrase extraction

Screen Shot 2019-06-19 at 08.35.59.png

The keyphrases help you to design your codebook even more seamlessly.

Keyphrase detection will automatically run within the codebook assistant, whenever you upload a new dataset without codebook.

Give it a try and let us know what you think!

Fancy, customizable visualizations 📈🌠

Tired of the same old bar-chart?

Design eye-catching diagrams right from the app.


GraphLine & Pietile-treemap.jpgBar

Choose between bar, line, graph, treemap and graph charts, customize the look and feel and compare series over time or additional variables in your dataset. You'll find the new advanced visualizations where you found our standard bar-chart.

Missing your favorite visualization? Let us know what you would like to see next - we want to build it!

Codebook Templates

Autocode responses without any manual effort

Don't have fixed codebooks you want to apply to your open-ends and don't want to start from scratch?

You can now choose from a variety of Codebook Templates and directly apply them to your data. Watch your responses get tagged fully automatically with industry-specific topics! You can even add custom codes and see how they are applied instantly. Screen Shot 2019-04-01 at 15.31.50.png

This mode helps you to either get a head-start for your in-depth analysis or lets you tag datasets that you don't want to code manually without any effort.

Just upload your data as open answers and you will be guided through the interactive assistant.

Drastically improved list Autocoder

The automatic coding for list questions now handles inconsistently delimited mentions and can be reapplied consistently to new waves of a study.

To try it yourself, download this demo dataset, then upload it as list question.

Check out this Tutorial to learn how a brand awareness study is analyzed within minutes.

Mentions with different or no delimiter are categorized correctly

Improvements in detail

When collecting responses for brand awareness-like questions, you often provide multiple fields for participants to enter their answers. However, many users will enter multiple mentions in the same field, either separating them by commas or dashes, or not at all (see screenshot above). Our algorithm can now identify the relevant concepts in these cases as well!

Inheritance can now also be applied to list questions: If you already have a previous iteration of your survey and link the data during upload, the AI will group new responses into the codes defined in the previous survey. It will also suggest new codes whenever it makes sense, which you can accept or not.

Presenting @ Quirks Chicago 2019

We're excited to be speaking at the Quirks event in Chicago this Wednesday, 3rd April!

Together with Frank Buckler from Success Drivers, we will show you how to find hidden Loyality drivers from simple NPS questions, using

If you are there as well, let's grab a coffee!

Drop me a note by Email or through the Event's klik App.

Support for 36 languages

Codit now supports most latin languages as well as some non-latin languages

Newly added languages include:

  • French
  • Dutch
  • Italian
  • Portuguese
  • Spanish
  • Indonesian
  • and many more

Screen Shot 2018-12-19 at 15.51.56.png

For a complete list, see the Language drop-down on the import page.

Please note that the automatic sentiment is currently only available for English and German.

Automatically code very short verbatims

Tired of coding different spellings of brand names in brand trackers?

In the past weeks we have built the new "Autocoder" which can code very short list-type verbatims - for example occuring in brand trackers - fully automatically. Read on to discover when and how to use it. Screenshot 2018-11-05 at 11.11.48.png

How to code list answers

Upon opening your project for the first time, you will be presented with the view above. Gain a quick overview over what codes we've discovered and the different spellings associated with it. You will be also able to directly see the resulting code-distribution. Adapt the settings on "Minimum count" and "Term similarity" if necessary to get more or less codes or to group them more or less aggressively.

Screenshot 2018-11-05 at 10.00.47.png

How to import list answer data

During import you will now be asked which type of data you want to code. Click on "List question" if your verbatims consist of very few words. For example, answers to the question "What is the most important online service for you?" would fall in this category. Make sure that every row only contains one term. If respondents entered multiple answers, e.g. separated by commas, split them into individual rows first.

If you would like to code open-ends with full sentences, click on "Open question" and you will see the flow you're used to.

Book a demo if you'd like to get a more in depth introduction into the Autocoder

Change code order

Lots of people requested the ability to change the code order. You can now do so in two ways:

  • Drag and drop using the mouse
  • Press Shift and left or right arrow on the keyboard when a code is selected

Screen Shot 2018-09-18 at 14.24.35.png

Updated Privacy Policy

We've updated our privacy policy. Here are some key changes:

Explicit GDPR compliance statement

We've always been compliant with the General Data Protection Regulation (GDPR). We've updated our privacy policy to make this more clear and outline the key components on how we comply.

Clarification on how we use your data

We're using your uploaded text data to improve the engine behind We've outlined this explicitly in the new privacy policy. You are still the sole owner of your data and can remove it from the platform at any time.

To read the new Privacy Policy, go to the registration on

Engine improvements

We've been working hard to improve the engine behind over the last couple of weeks.

Specifically, we've added better support for listing questions such as in brand perception questions. is now able to pick up different spellings much quicker than before.

This works by comparing the code name to the answers in a clever way. To maximize effectiveness of this we depend on you choosing the code name correctly.

Some examples of code names:

  • Bad: "Pepsi/Pepsi Cola/Pepsi Beverages", Good: "Pepsi"
  • Bad: "Bang&Olufsen/B&O", Good: "Bang&Olufsen"

If you have any questions regarding this improvement or are unsure how to name your codes, contact us via the chat.

Please note that this improvement only applies if the answers are short (less than 8 words per answer on average).

No published changelogs yet.

Surely Codit will start publishing changelogs very soon.

Check out our other public changelogs: Buffer, Mention, Respond by Buffer, JSFiddle, Olark, Droplr, Piwik Pro, Prott, Ustream, ViralSweep, StartupThreads, Userlike, Unixstickers, Survicate, Envoy, Gmelius, CodeTree