FeedbackHub
Four hundred comments, one list of what to fix
Collecting feedback is the easy half. The reporting layer scores every response against the standard customer experience metrics, reads the free text for you, and ranks what customers keep bringing up so the next decision has something behind it.
Everything for the period, on one screen
Choose 7, 30, 90 or 365 days. The overview recalculates against that window and reports total responses, average rating, average satisfaction score, average effort score and Net Promoter Score, along with promoter and detractor counts.
A daily trend line shows response volume and average rating together. That pairing is more useful than either on its own, because it separates "we got fewer responses" from "we got worse", which are entirely different problems with entirely different fixes.
Days with no responses are plotted as zeros rather than skipped, so a gap in collection looks like a gap rather than a smooth line. If the QR cards fall off the tables for a fortnight, you will see it.
Three standard scores, each answering a different question
FeedbackHub uses the established customer experience metrics rather than a proprietary score, so your numbers can be compared against industry benchmarks and against whatever you measured before.
Net Promoter Score
Question: how likely are you to recommend us, from 0 to 10?
Calculation: the percentage of promoters, who answer 9 or 10, minus the percentage of detractors, who answer 6 or below. Passives at 7 and 8 count towards the total but not the score.
Range: −100 to +100. It measures loyalty and word of mouth, which is why it tracks repeat business better than satisfaction does.
Customer Satisfaction
Question: how would you rate your experience, usually one to five stars or faces?
Calculation: the average, plus the full distribution across each point on the scale.
Use it for: a specific interaction. CSAT is immediate and transactional, which makes it the right metric for "was that visit good", and the wrong one for "will they come back".
Customer Effort Score
Question: how easy was it to get what you needed, from 1 to 7?
Calculation: the average, with the distribution and the count of responses at 4 or below flagged as friction.
Use it for: finding the steps that wear people down. High effort predicts churn even when satisfaction still looks acceptable, so a falling CES is an early warning.
All three can live on the same form, or you can run them separately per form. Each response records whichever were answered, and the reporting keeps them apart rather than blending them into one average that means nothing.
Averages hide things. Distributions do not.
Rating distribution
Star ratings are broken out across all five points. An average of 3.5 built from fives and twos is a polarised business with a specific problem affecting a specific group of customers. An average of 3.5 built entirely from threes and fours is a mediocre business with no single fault. The average alone cannot tell them apart.
Effort distribution
Effort scores are broken out across all seven points, with the lowest scores surfaced first so friction is visible rather than buried. The platform also lists the individual responses behind the lowest effort scores, so you can read what those customers actually said.
Channel performance
Responses are counted per channel: email, SMS, WhatsApp, QR, extension and direct link. This is how you find out that the QR cards you were sceptical about are producing half your feedback, or that the SMS budget is buying very little.
Per-form summary
Each form gets its own row with response count, average rating, promoter and detractor counts. Run a form per branch or per service line and the comparison is immediate.
Responses needing attention
The platform assembles a shortlist of responses most likely to need a human, ordered by negative sentiment first, then detractor scores, then high effort and low ratings. It is the queue to work through on a Monday morning.
Recommended actions
Based on what your current data actually shows, the reports page suggests next steps: collect more responses if the sample is too small to trust, follow up with detractors, look at low-effort cases, or ask promoters for public reviews.
The part that reads the comments
Scores tell you that something changed. Only the free text tells you what. The problem is that free text does not scale: nobody is going to read four hundred comments a month, and if they did, they would not reliably notice that the same complaint appeared thirty-one times in slightly different words.
FeedbackHub passes each comment to a language model that returns three things: a sentiment score, a sentiment label of positive, neutral or negative, and one or more themes describing what the comment is about. Themes are stored per account with a running count, so "queue was mad", "waited ages to pay" and "till was slow" all accumulate against one entry rather than sitting as three separate observations.
The output is a ranked list of what customers praise and what they complain about, with the strongest negative theme for the period surfaced along with supporting quotes. That is as close to a prioritised to-do list as customer feedback gets.
Analysis runs in the background
Sentiment and theme extraction happen on a schedule rather than during submission, so the customer never waits for it. They see the thank-you screen immediately and the insight appears in your dashboard shortly afterwards.
Sentiment is a trigger, not just a label
A politely worded four star response can still describe a serious problem. Because negative sentiment can fire a workflow on its own, those cases raise an alert even though the score looked acceptable.
Themes accumulate over time
Theme counts are kept per account and build month over month, so you can watch an issue climb the list before it becomes the thing every review mentions.
How to read your first month of data
The first thing to check is not the score. It is the count. Twenty responses is directional at best, and drawing conclusions from a handful of answers is how businesses end up rearranging a whole service around one loud opinion. Keep collection running until patterns repeat before you act on anything except individual complaints, which are always worth answering regardless of sample size.
Once volume is there, read in this order. Start with the distribution rather than the average, because that tells you whether you have a broad problem or a specific one. Then look at the negative themes, because those name the problem. Then read the individual responses behind the worst effort scores, because those describe the mechanism. By that point you usually have a fix rather than a feeling.
Compare periods only when the underlying form has not changed materially, or accept that you are comparing two slightly different questions. Version history means you can always check which wording produced which set of answers.
Finally, watch the gap between your internal average and your public review average. When the internal number is meaningfully higher, your happy customers are not making it to the review sites, and review routing is the fix. When the two are close and both low, the feedback is telling you something about the business rather than about the collection method.
See the reporting with your own data in it
We can walk you through the dashboard using a sample of your own responses, so you can judge whether the reporting answers the questions you actually have.