Case studies
Real reviews, run through ClientClue AI
About these case studies
Whose reviews these are, and what they show
- These are public Google Play reviews. Anyone can read them on the Play Store.
- Each company is named so you can see where the reviews come from.
- None of these companies is a ClientClue customer, and none of them took part.
- The counts show what reviewers wrote. They are not ClientClue's view of the company's service.
- Reviewer names and personal details are hidden. So are other companies named inside a review.
How we measured
Counted by code, checked word for word
The cases are public Google Play reviews. They are not support tickets, and the people who wrote them are not our customers. Reviewer names, phone numbers and other personal details are hidden.
Every figure here was counted by ClientClue on those reviews, with the number of messages it comes from.
4,286 quotes shown
Every quote was found word for word in its message. 2 more were dropped because the words did not match.Counted across 5 cases
MTN Nigeria: 376 messages a minute
590 messages read in 94 seconds.Measured on the run shown above
Kuda: 207 messages a minute
1,000 messages read in 289 seconds.Measured on the run shown above
Jumia: 154 messages a minute
922 messages read in 360 seconds.Measured on the run shown above
OPay: 415 messages a minute
1,039 messages read in 150 seconds.Measured on the run shown above
inDrive: 365 messages a minute
866 messages read in 142 seconds.Measured on the run shown above
MTN Nigeria: 1% of 1 and 2 star reviews read as positive
8% of 4 and 5 star reviews read as negative. The star and the words do not always agree.Star against text, 590 messages
Kuda: 16% of 1 and 2 star reviews read as positive
2% of 4 and 5 star reviews read as negative. The star and the words do not always agree.Star against text, 1,000 messages
Jumia: 35% of 1 and 2 star reviews read as positive
1% of 4 and 5 star reviews read as negative. The star and the words do not always agree.Star against text, 922 messages
OPay: 26% of 1 and 2 star reviews read as positive
4% of 4 and 5 star reviews read as negative. The star and the words do not always agree.Star against text, 1,039 messages
inDrive: 6% of 1 and 2 star reviews read as positive
3% of 4 and 5 star reviews read as negative. The star and the words do not always agree.Star against text, 866 messages
ClientClue AI was tested on tweets in four Nigerian languages that people had labelled by hand. Each figure is the share it read the same way, with the number of tweets tested.
- Nigerian Pidgin400 tweets tested60%Between 55% and 65%
- Hausa200 tweets tested75%Between 69% and 80%
- Igbo200 tweets tested71%Between 64% and 76%
- Yoruba200 tweets tested82%Between 76% and 86%
The test uses NaijaSenti, tweets labelled by people, licensed CC BY 4.0. Credit: Muhammad et al., LREC 2022. See the dataset.
Tweets are not customer messages, so these figures show how well ClientClue AI reads each language, not how it will do on your file.