How to build meaningful Queries in Kapiche

A collection of our top tips for creating high quality queries and themes in Kapiche!

Ryan Stuart avatar
Written by Ryan Stuart
Updated over a week ago

Queries are the building blocks for insights - you can't have a Theme to report on without first building it using Queries in Kapiche - and there are a lot of different ways you can approach this.

With this in mind, we've put together our top tips for building meaningful queries!

Shortcuts:

A couple of things worth knowing before you get started:

  • If you aren't sure what Queries are, check out this handbook first

  • A Query represents the elements that are used to create a Theme and once that Query has been saved, we'll then refer to it as a Theme

Let's dive in! 🏊


Know which mode of querying you're in

There's two primary modes of querying in Kapiche:

  1. Using Queries to build specific Themes to report on and/or track over time

  2. Using Queries to answer specific and/or ad-hoc questions (QA)

It can be helpful to know which mode of querying you're in before you get started as this can change the way you approach query building!

For example:

If you're wanting to build Themes that you can measure and report on on an ongoing basis, then you're going to want to start off at the Storyboard level which will help you identify the top driving Themes your customers are talking about which you can then save and report on using your Dashboard.

But if you're in QA Mode and are trying to answer a specific question (e.g. "How many people mentioned App Crashes in August 2022?") then you would start by running a custom query that's going to help you answer that question.

We find that the easiest way to work out which mode you're in is to ask yourself whether you'd like to come back and see how this topic is trending over time or if it's just an ad-hoc question that someone in your team has asked you!


Keep your audience & purpose in mind

If we're going to be using queries to build Themes for reporting purposes we'll usually start by answering the following questions:

  • Who are we building insights reports for?

  • What types of insights do they care about?

  • Who are our different departments?

  • What are the different touch points of the user experience?

The reason knowing who we're building insights for is valuable is because it

can often guide the way in which we approach building out our queries.

For example if we're building insights reports for our C-Suite who are only focused on how each department is performing at the very top level, then we're going to build a much broader set of queries like this:

  • Customer Service

  • Product Range

  • Product Quality

  • Pricing

But if we're building out reports for Department Managers who are looking for highly actionable insights, then we'll need to build a much more targeted set of queries for each Department that breaks down the key areas that they can influence.

For example if we have a Customer Service Team who is looking for specific insights about the different touch points they own, we'll likely build out a set of queries for them that looks like this:

  • Customer Service // Wait Times

  • Customer Service // Staff Knowledge

  • Customer Service // Tone of Voice

  • Customer Service // General

We'd then typically create what we call a Parent Theme where we combine the different Customer Service Queries together into one bigger Theme (Customer Service) so that we can see both how the overall Department is performing against other Departments as well as how each of the specific Sub themes are performing.

From there, we rinse and repeat this workflow, building out Parent and Sub Themes for each of the key user touch points / business areas as needed!


Let your Storyboard (aka your customers!) guide you

One of the learning curves you'll likely experience with Kapiche is learning how to let your Storyboard guide you towards identifying and building out your themes (e.g. it's super common for our users to start by looking for the specific things that they're used to reporting on).

The challenge with this is that you don't know what you don't know - which is exactly why Kapiche exists - to show you exactly what your customers are talking about in their own words without the human bias that comes with Themes that were originally built via manually coding.

With this in mind, when you're first getting started, we recommend using your Storyboard to identify your initial set of Themes as this is going to show you what your customers care about the most and what they're most vocal about.

There's a good chance that the themes you're already aware of will surface but it's also possible that there are new Themes that bubble up in Kapiche that you weren't aware of that might be having a bigger impact than expected!


Use your Context Network & Verbatims to vet your queries

Your Context Network is a SUPER handy tool that helps you to make sure that your query is successfully bringing in the verbatims you're intending.

We recommend leaning on this as much as possible as you're adding different things to your query to see how the different bubbles change and interact with one another (as the Context Network is going to show you what the key topics of the verbatims are that have been included in your query and will update in real time as new things get added in/removed).

If you find that the context network changes dramatically (e.g. a new set of terms appears in its own cluster in the network) after adding something to your query, we recommend taking a closer look at the surrounding terms in the Context Network to make sure they are still relevant to the Theme you're building.

In the example above you can see that there's two unique Context Networks that have appeared that don't seem overly related to one another - this is an indicator that we've included a term in our query that might be bringing in verbatims that aren't quite related to the Theme we're trying to build!

If this happens, you can tweak the query either by removing the term you previously included that resulted in the "false positives" or by adding in some "does include" or "does not include" parameters to filter out the verbatims you don't want to include.


Avoid biased, broad and generic terms

A common mistake our users make when learning how to build queries is including biased, broad and/or generic terms when trying to create specific themes which can result in Themes that are either too biased or are too broad and aren't quite capturing what they intended.

Some examples of these terms are:

  • Good, Great, Love

  • Friendly

  • Long

  • Issue

  • Company Names (e.g. Kapiche)

Using positively geared terms like Good, Great and Love can unintentionally bias the metrics attached to your data (like Sentiment and NPS or CSAT) unless you round them out by also including negatively geared terms like Bad, Rude, Awful (which can become quite tedious and time consuming!)

Instead, we recommend looking at the Storyboard and Context Network to see what other terms are appearing along side these ones - e.g. what is happening that is Good or Great - is it the Service? The Product Quality? The Wait Times?

From there you can then use those more specific terms (Service, Quality, Wait) to build out a more balanced query that brings in all of the verbatims related to that specific theme regardless of their sentiment.

The same logic should also be applied when it comes to looking at broad terms that can be used in multiple contexts like "Issue" (which can relate to a number of different themes such as Issues with Lost Bags, Issues with Wait Times, Issues with Staff, etc).

In this example we recommend leaning on your Storyboard and Context Networks to help you identify whether the term you've used in your Query is being used in multiple contexts and if it is, try switching it out for something that's a bit more specific to the theme you're wanting to build.


Avoid using structured data

We often see users adding structured data (Fields) to their queries to help focus their Themes on specific segments like Detractors or customers with low CSAT scores (which makes sense as most folks are looking for the areas of the business they can improve on!).

The downside to doing this at the Query level instead of at the Dashboard level is that you then lose the ability to see how that theme is actually performing against other Themes (e.g. if you add a Detractor Segment to your query, you then lose sight of how many Promoters are talking about that theme, which may make that Theme look more negative than it actually is).

Instead, we recommend building what we call 360 Themes - Themes that aren't biased either way and instead look at how customers are talking about that Theme in any capacity.

This allows you to use your Dashboard to benchmark each theme against your core business metrics to see which Theme is having a positive or negative impact on your business (which won't be possible if you are using Segment filters inside your queries).

Check out this guide on Finding High Impact Themes using your Kapiche Dashboard to learn more!


Use Synonym Suggestions to build unbiased Queries

Synonym Suggestions is a super handy tool that helps you make sure that you're bringing in all of the possibly related terms to the ones you've included in your query - ultimately helping you improve the quality and accuracy of your query.

For example, when you add a Concept to your Query and click the 'OR' button, Kapiche will show you the Top 12 most-related Synonyms that we can see in your analysis as a way to help improve your query strength (as well as save you time!)

As you add a Synonym Suggestion to your Query, when you click OR again, we'll automatically refresh the list with a new set of suggestions that takes into account all of the concepts on that line which allows you to make sure that the query you're building captures all related terms that we can see in the data!

We'll even show you related terms that have been misspelled by multiple customers!


Sanity check you've captured everything

Once you dive in and start building out each of your Themes, it's a good idea to use the Dimming Switch on the Storyboard and the Unmapped Verbatims tool to sanity check that you've captured all of your core themes before you dive in and start analyzing them / reporting on them.

Storyboard Dimmer

On your Storyboard, you'll find a little toggle towards the top-left of the board that says "Dim concepts added to a Theme" - if you toggle this on, the Concepts that belong to one of your Themes will dim, helping you to spot whether there are some Themes that still need to be created (or whether there are some extra terms that could be added to existing Queries).

Unmapped Verbatims

Similar to the Storyboard Dimmer, the Unmapped Verbatims tool is a dedicated page that shows you the Top 100 Concepts and Terms that have not been captured by one of your saved queries and is a great way to spot unique Themes that have been missed or which Concepts/Terms need to be added to one of your existing Themes.

As a very high level guide (and definitely not a hard and fast rule!!) we recommend trying to get your total percent of unmapped verbatims down below the 30% mark before you start reporting on your Themes as this will ensure you've captured the most prominent Themes in your data.

Tip: You can also use Unmapped Verbatims to spot any new and emerging Themes that pop up in your data over time!

Learn more about how to use Unmapped Verbatims here.


There you have it! Our top tips for building meaningful queries 🙌

Questions? 🤔

If you have any questions about building Queries (or you just need some help!) you can get in touch with us any time by hitting the blue chat button to your right 👉

Did this answer your question?