Solutions · Touchpoints Where do customers run into problems?
Compare ratings and comments on your touchpoints – from booking to support. See where criticism clusters and read the individual comments in the original.
Where things go wrong is in the comments – spread across the whole journey.
Booking, check-in, support: each stage comes with its own questions and its own comments. A complete picture doesn’t emerge on its own.
The challenge: Each stage on its own
Every touchpoint is analyzed separately. Topics that run through the entire journey barely stand out.
With deepsight: One topic system for all touchpoints
Analyze several open-ended questions from one study together. That way you spot topics such as waiting times that come up at several stages.
Go to topic analysisThe challenge: A share without substance
Negative comments pile up for support. The number alone doesn’t show what exactly is being criticized.
With deepsight: From share to comment
For every touchpoint, you see which topics are criticized and read the comments behind them verbatim.
Go to sentiment analysisThe challenge: Complaints, questions and praise mixed up
Complaints sit next to questions and praise. If you want to report to a specific team, you first sort by hand.
With deepsight: Intents analyzed separately
deepsight sorts comments by intent. You report clusters of complaints about a touchpoint to the team responsible in one go.
Go to intent analysis
Criticism per touchpoint, with the comment behind it
The touchpoint is a field in your data, for example from the question a comment refers to. Select a touchpoint.
Illustrative figures. A high share shows where criticism clusters. It does not tell you on its own how strongly a touchpoint affects the overall rating. Read small bases, such as support, with caution.
“Booking on the website was quick, it’s just that the room categories were hard to compare.”
Select a category to see the matching passage.
What a touchpoint analysis answers
- At which touchpoints do critical comments cluster, and on which topics?
- Which topics run across several touchpoints, such as waiting times?
- Which topics do customers who rate a touchpoint poorly mention more often than satisfied customers?
- Is a touchpoint rated and commented on differently depending on channel, location or customer group?
- Has criticism of a touchpoint changed between two waves?
- Which comments are complaints, which are questions, and which are praise?
What you can start with
What matters is that each comment can be linked to a touchpoint.
From comments to comparison
- Bring in your data
Upload a file or set up a connector. Personal information is masked before the analyses run; the anonymization does not detect every piece of information.
Anonymization - Check data quality
If you wish, spot empty or content-free answers and filter them out. The check also filters out answers such as “all good”. If that is a statement about a touchpoint, weigh up whether the check is the right fit.
Data quality check - Identify topics
Have topics suggested or apply your own codebook. Sentiment is classified for each topic.
Topic analysis - Compare touchpoints
If the touchpoint is a column, choose it as an attribute and compare topics and ratings, with significance indicators if you want them. With one open-ended question per touchpoint, you read the topics for each question.
Group comparison - Ask and check
Ask questions about your analysis and check the answers against the data and evidence.
Assistant
What you end up with
Good to know: Wave comparison is included in every paid plan – see the plan comparison. If your rating tables by touchpoint come from a different analysis tool, slide factory turns them into slides in your own template.
The share shows where criticism clusters. The comments show what is being criticized.
Comparing touchpoints fairly
Base and multiple mentions
The base consists of the analyzed comments on a touchpoint, not every customer who used it; whatever the data quality check filters out doesn’t count. The base is shown next to each share. If mostly dissatisfied customers comment, the share of negative comments will be higher than among all users. Multiple mentions are possible; values with a base below 10 are flagged or hidden.
Same respondents, several touchpoints
Significance indicators compare a group with the rest of the sample, from 30 cases per group, and treat cases as independent. If the same people comment on several touchpoints, that is not the case – treat the indicators as a rough guide.
Frequency is not impact
Many negative comments show where criticism clusters – not how strongly a touchpoint affects the overall rating. People who contact support have often had a problem already.
Review and adjust
For every topic you see the assigned comments. Changes are saved as a version, and a new AI run can produce different results. More in our methodology.
Good to know: More on significance indicators and minimum bases in group comparison. You can weight results by a weight column or to target distributions (rim weighting).
Questions about touchpoint analysis
If there is only one general open-ended question, set up the touchpoints as parent categories in your codebook, with the topics below them. This only works if a comment clearly names the touchpoint – “took forever” on its own cannot be assigned to one. Check the assignments against the original responses.
No. But you can compare which topics customers with a low overall rating mention more often than satisfied customers – in a group comparison with the rating as the attribute, with significance indicators if you want them. For your own statistical models, export the data to Excel or CSV.
Combine the exports into one file and include the source as a column; you can then select the source in a group comparison like any other attribute. To do this, export reviews from the connector to Excel or CSV. Since scales and audiences differ from source to source, compare sources side by side rather than merging them into a single figure.
This often happens at touchpoints such as reception or advisory services. Our own anonymization masks names before the analyses run; you decide which other types of information are masked. It does not detect every piece of information: indirect references such as “the young man at reception on the night shift” are not picked up automatically, so you mask passages like these by hand. Check results before sharing them.
deepsight cloud runs on Microsoft Azure in the EU. Find out more under Security.
Related pages
Analyze your touchpoints
In a personal demo, we show you the touchpoint comparison using an example from your field. Afterwards, you receive test access – a completed wave of your journey survey works well for this.