Importyour file
- CSV
- Excel
- Paste
- Four file kinds
Drop in a CSV or Excel file, or paste messages. ClientClue matches the columns and shows what it will drop and hide before it reads anything.
Every customer who leftcomplained first.
The problem
Complaints pile up faster than anyone can read them.
413,587 complaints against four phone networks in six months. 23,396 were about data running out.
NCC data, reported by LeadershipMoney problems end up at the regulator.
23,129 complaints about banks and fintechs reached the Central Bank in a year. ₦19.12bn and $329.3m were refunded.
CBN Annual Report, reported by Prime Business AfricaSlow handling gets fined.
₦430m in fines across 21 penalties for slow or poor complaint handling.
CBN Annual Report, reported by NairametricsComplaints arrive in chat and app reviews, not one inbox.
95% of smartphones in Nigeria have WhatsApp installed, so complaints arrive in chat as well as in ticket forms.
Orange Group and KPMG Nigeria Smartphone Study, reported by BrandsPurNG
How ClientClue solves it
The four questions your team asks, with numbers for answers.
What are customers complaining about most?
Top issues, counted.
56 messages
"App lagging and slow response" is the biggest issue in the public Google Play reviews of MTN Nigeria: 9% of the 590 messages.
Is anything new going wrong?
Rising and New, the week it starts.
3 Rising, 0 New
In the last full week, "App crashes or closes unexpectedly" was Rising: 9 this week, 5 the week before.
Why are people leaving?
Cancellation reasons, in their own words.
14 leaving
Messages across 8 weeks said the customer was leaving or had left. In the last full week the top reason was a move to a rival.
What do I show the team each week?
The weekly report.
5 issues
The weekly report for Week 7 lists the top 5 issues, with 14 quotes checked word for word, and what is Rising.
Core features
From a pile of messages to a plan for the week.
From file to findings
The dashboard, live
Watch a real week of complaints get sorted.
- Read
- Group
- Spot
- Report
MTN Nigeria, 590 messages
Messages arriving, 0 of 40 shown
Issues, counted as they land
- App lagging and slow response0
- Airtime and data balance not updating0
- App crashes or closes unexpectedly0
- Poor internet connection speed0
- App login authentication error0
- Other issues0
Play
Week
How it works
From a file to a report in four steps.
Readevery message
- English
- Pidgin
- Hausa
- Yoruba
- Igbo
ClientClue AI reads every message and notes its language, feeling and kind: problem, question, request, praise or leaving.
Groupwhat repeats
- Areas
- Issues
- Rising
- New this week
Messages that mean the same thing become one issue with a plain name. Problems that grow, or appear for the first time, get a flag with their numbers.
Reportevery week
- Checked quotes
- PDF
- Share link
- WhatsApp summary
Your team gets one weekly report, on the day it picks: top issues, what is rising, why customers leave, and quotes checked word for word.
How ClientClue AI works
It reads Pidgin, slang and one-word reviews.
- OTP not arriving for login
- Unable to log in
- Delivery fee too high
- App slow to open
- Fares too low for fuel costs
- Excessive data depletion
- Account restricted or locked
Code does the counting. ClientClue AI does the reading. Every number comes from code, and every quote is checked by code.
Hide personal details
- Code does
- Finds names, phone numbers, emails, account numbers and card numbers and replaces them with labels like [phone].
- ClientClue AI does
- Only ever sees the hidden version.
Find the issues
- Code does
- Picks a spread of up to 400 of your messages.
- ClientClue AI does
- Proposes areas and issues in plain names, such as "Data deducted but not delivered".
Read each message
- Code does
- Sends messages in batches of about 50 and checks that every message comes back.
- ClientClue AI does
- Notes the language, the feeling, the kind, the issue, and an English line when the message is not English.
Check every quote
- Code does
- Looks for each quote word for word in its message. A quote that is not found is dropped.
- ClientClue AI does
- Picks the short part of a message that says it best.
Count
- Code does
- Counts each issue by week and works out Rising and New with fixed rules.
- ClientClue AI does
- Does no counting.
Write the report
- Code does
- Picks the issues, the numbers and the quotes, then checks that every number in the written lines matches.
- ClientClue AI does
- Writes one plain line for each issue and a "check this first" line.
Before and after
One file in. One report out.
Before
590 app reviews in a spreadsheet.
Public Google Play reviews of MTN Nigeria, over 8 weeks, in one file.
After
- 49
- issues found
- 3
- Rising or New problems in the last full week
- 582
- quotes checked word for word
- 94
- seconds to read every message
Case studies
Real reviews, run through ClientClue AI.
Public Google Play reviews, read by ClientClue. Each company is named for transparency; none is a customer or took part. Reviewer names and personal details are hidden. About these case studies
- 4,417messages read in the case studies
- 4,286quotes checked word for word
- 60%sentiment accuracy on 400 labelled Nigerian Pidgin tweets
- 376messages read a minute
What you can bring
Bring the file you already have.
Support tickets
Message, subject, date, channel, status, customer, product
Cancellations
Cancellation note or reason, date, customer, amount, product
App-store reviews
Review text, star rating, date
Survey answers and other feedback
Any text column, plus a date and a rating if you have them
CSV or Excel, or paste. Name, phone and email columns are hidden before reading.
Who reads it
The person who opens it every week.
Every number traces back
- Every count clicks through to the messages behind it
- Every quote is checked word for word against its message, and dropped if it does not match
- Every number in a report line is checked against the counts
Your data
- Names, phone numbers and account numbers are hidden before ClientClue AI reads anything
- Your original file is not kept after import
- Your messages are read in the EU
- You can delete your data any time
Pricing
Pay for the messages you read.
- Free300 messages, once₦0
- Starter3,000 messages a month₦30,000 a monthabout $21
- Pro15,000 messages a month₦100,000 a monthabout $71
- Business60,000 or more messages a monthfrom ₦350,000 a monthabout $248
100100,000
1,100messages a month
Starter
₦30,000 a month
about $21
Start with Free: 300 messages, once. You pay month to month and can stop any time. See every plan
Our story
Made for teams that hear the complaint too late.
Customers complain before they leave. In Nigeria they complain in app reviews, in chat, on social media and in support tickets. Often it is in Pidgin or a mix of languages. Most teams read a few of them and miss the pattern.
ClientClue reads all of them. It groups what repeats and flags a problem in the week it starts. Each week your team gets your customers' own words.
We built it in Nigeria, for Nigerian and African teams, and we publish how well it reads each language.
Questions
Plain answers to plain questions.
A CSV or Excel file of your customer messages, or messages pasted in one per line. Support tickets, cancellation notes, app-store reviews and other feedback all work. ClientClue matches the columns for you and you confirm them.
English, Pidgin, Hausa, Yoruba, Igbo and mixed messages. ClientClue AI keeps the original words and puts a one-line English version beside them. How well it reads each language is measured and published on the case studies page.
Names, phone numbers, emails, account numbers, card numbers and ID numbers such as BVN (the bank identity number every Nigerian account has) and NIN (the national identity number). They become labels like [phone]. You see what was hidden before you start.
In the EU. Your original file is not kept after import, only the hidden version of each message. The Your data page has the details and how to delete everything.
Free reads 300 messages, once. Starter is ₦30,000 a month (about $21), Pro is ₦100,000 a month (about $71) and Business is from ₦350,000 a month (about $248). You pay month to month and can stop any time.
A problem is Rising when it has at least 5 messages this week and at least 1.5 times the average of the 4 weeks before. It is New this week when it has at least 5 messages and none in the 4 weeks before. Every flag shows its numbers. Flags need messages from at least 2 earlier weeks.
The day your team picks, called the report day. The default is Monday. A week then runs for 7 days from that day, in Lagos time, and its report is ready on the next report day. If you pick Friday, weeks run Friday to Thursday and the report arrives every Friday. You can change the day later, and your weeks are counted again from the new day.
No. ClientClue reads, groups and reports. It never sends a message to a customer. Your team decides what to fix and who to answer.
No. Code hides names it can find before reading: names from your name columns, and names after a greeting or after "my name is". A name written in another way can slip through, so keep names and phone numbers in their own columns.
Guides
Short guides for your first file.
- 01Prepare your fileOne message on each row, the columns ClientClue reads, the date formats it understands and what it hides.
- 02Read your weekly reportEvery part of the report, what each number means, and what to do with it.
- 03Add a cancellations fileThe columns a cancellations file needs and how ClientClue counts the reasons customers give for leaving.