This article is a compilation of frequently asked questions about our OpenAI integration and feedback topics management in general.
Start with checking the table of contents (left sidebar →) and click on any topic that interests you.
How to classify feedback into topics automatically?
To have your survey responses analyzed and automatically classified into topics, you will need to activate the “Automated topic & sentiment classification” option within a survey campaign’s Analytics section.
Choose a classifier that best reflects your business and save the campaign settings. From that moment on, all survey responses received in that particular campaign will be analyzed using OpenAI’s algorithm, which will identify all topics relevant to the response’s context, and will assign a sentiment (positive, neutral, negative) to each topic.
How does the OpenAI integration work?
When you receive a survey response with a text comment, we send it in a prompt to OpenAI for analysis. Along with the survey response, we’re also sending over up to 20 most used topics from your account, and we’re asking OpenAI to pick relevant topics from the list or suggest new ones.
OpenAI will then send us back a list with topics, the sentiment for each topic, and a list of other parameters per topic that participate in the calculation of a topic’s Impact Score.
What data are you sending to OpenAI, and is it safe?
We’re not sending any of your, as a Retently client, or survey respondents’ identifiable information. We’re sending the main open-ended text comment, which may or may not include personal data, depending on what the survey respondent has decided to share.
Also, per OpenAI’s “API data usage policies” from June 14, 2023, quote: “OpenAI will not use data submitted by customers via our API to train or improve our models, unless you explicitly decide to share your data with us for this purpose”.
Retently hasn’t opted in to share any of the submitted data.
Why are new topics created automatically in my account?
If you notice that your survey responses include topics that you haven’t created yourself, then there are a few possibilities:
1. The OpenAI integration creates them:
When OpenAI’s algorithm analyzes your survey responses, it may suggest new topics if any relevant alternative is missing from the 20 top topics that were sent for analysis. If OpenAI suggests a topic that you already have in your account (maybe it’s below the top 20 topics mark), our system will be able to identify it as a duplicate and apply the existing one to the survey response and dismiss the new topics suggested by OpenAI.
However, we will only be able to identify duplicates if the newly suggested topic by OpenAI is identical to the existing one (case insensitive).
If you need to have more control over your topics list, then you can deactivate the option that allows OpenAI to suggest new topics. To do so, access each survey campaign where you have activated the “Automated topic & sentiment classification” option, and in the Analytics section, remove the tick from the “Allow to create new topics automatically” option.
2. They’re created by an external integration via Zapier or API:
You can add topics to survey responses using our Zapier app or API endpoint. In the first case, you can check your Zapier account and see if you have any Zap automation that uses the “Apply topic to responses” action. In the case of API, reach out to our support team, and we will check if your account is making any requests to the “Add feedback topics” API endpoint.
3. They’re created from a Multiple Options survey question:
It is also possible to create and further add as topics the options that survey respondents select and submit in a Multiple Options question. To deactivate this option, access your survey campaigns and the templates assigned to them.
Next, in the "Survey questions" section, access a “Multiple Options” question (please keep in mind that the questions might have a different name), remove the tick from the “Apply as feedback topic” option, and save your changes.
4. Another user has created them:
Finally, the simplest explanation could be that another user in your account has created them. Currently, there is no possibility of knowing which user has created which topic, but you can preview the list of users that have access to your account and reach out to them for further clarification.
Why are topics created by me not added to the responses?
If you’ve selected the option to have your survey responses analyzed automatically by OpenAI, then what happens is that along with a survey response, we’re also sending OpenAI up to 20 of your most popular topics (the ones that have been added to a bigger number of responses).
If the new topic that you’ve added did not make it to the top 20 list, then it will not be sent to OpenAI for analysis. Here are a few actions that you can take in this situation:
Do nothing and wait until OpenAI suggests your topic enough times, and it will grow into the top 20 topics list.
Review your topics list and identify a top topic that would make more sense to be renamed as one of the less popular topics. For example, you might have a popular “Cost” topic that can easily be renamed to “Pricing” (assuming this is the name of the new topic you need to create).
Review your top topics and identify two or more that could be merged into a new one. For example, you could have “Support” and “Service”, and you can merge them into a new topic: “Customer Service”.
Delete top topics that you feel are redundant and bring no real value.
Lastly, you can manually go through your past survey responses and add the new topic to responses until it makes it to the top 20 list. However, this approach might become very challenging if the 20th topic was added to hundreds or thousands of responses.
Why does the Impact Score not reflect a topic's importance accurately?
The Impact Score’s main goal is to show you your most impactful topics, whether they are positive or negative drivers for your business.
There isn’t a maximum score, as it’s mainly a means to compare the score of all topics. The topic with the biggest score among others is the most impactful.
There are a few parameters that are taken into account when calculating the Impact Score of a particular topic. You can learn more about them here.
But there is one parameter in particular that makes the biggest difference: Topic Relevance Score, which helps our system understand which topic is the most relevant and important to your business.
The Relevance Score is a number between 1 and 5. Topics with a higher relevance score will have a higher Impact Score. By default, topics have a Relevance Score of 3, but you can manually adjust it on the Feedback Attributes page. Learn how to adjust the Relevance Score.