guides

Each row of your file is one customer message. A ticket, a review, a cancellation note and a survey answer are each one row.

If your export has the customer's words in several columns, keep the one that holds the message and leave the rest. ClientClue asks you to confirm which column that is.

You do not need a spreadsheet at all. Paste messages one on each line, or save a plain text file with one message on each line.

Only the message column is needed. Every other column makes the report better, and ClientClue matches them by their header names.

Message: the customer's words. Headers such as message, comment, review, feedback or description are matched.

Date: when the customer wrote. This puts each message in its week, which Rising and New this week need.

Rating: stars or a score from 1 to 5. It sits beside the text, so you can see when the two disagree.

Channel: where it came from, such as WhatsApp, email or the app store.

Customer: a customer reference or ID. It lets ClientClue count people instead of messages. The original is never stored; only a salted hash is kept.

Amount: the money at stake in naira, mostly for cancellation files. Amounts like ₦5,000 are read as 5000.

Reason: the free-text reason a customer gave for leaving.

Status, product and subject columns are read when they are there.

ClientClue reads the day first, which is how most Nigerian exports write it: in 05/11/24 the 5 is the day. It looks at your whole date column and switches to month first only when the dates clearly are written that way.

It also reads dates with the year first, dates with the month in words, a time after the date, and the day counters Excel keeps inside date cells.

A date with no time zone is read as Lagos time. Weeks are counted in Lagos time too, and each week starts on your report day, which is Monday unless you pick another.

A message with no readable date is still read and shows in your Messages list, but it is left out of the week-by-week counts and the weekly report. A file with no date column at all cannot show Rising or New this week.

Both work. A CSV can use commas, semicolons or tabs, and ClientClue finds out which. Files saved with a byte order mark at the start, which Excel adds, are read correctly.

An Excel file is read from the first sheet that has data in it. If your messages are on a later sheet, move that sheet to the front or save it as a CSV.

A file can be up to 15 MB and 60,000 rows. For a bigger file, split it by month and add the parts one after another.

Before ClientClue AI reads a word, code replaces personal details with labels: names from name columns become [name], phone numbers [phone], email addresses [email], 10-digit account numbers [account], 11-digit BVN and NIN numbers [id number], card numbers [card] and social handles @[handle].

BVN is the bank identity number every Nigerian bank account has. NIN is the national identity number.

Columns that hold names, phones or emails are never shown to ClientClue AI as examples. After you add the file, ClientClue shows how many of each kind it hid, so you can see it working.

The file you upload is not kept. Only the hidden version of each message is stored.

ClientClue drops empty rows, exact duplicates, replies written by your own agents, automatic replies and greetings with nothing else in them. It shows you the count of each before you start.

One-word messages such as Slow are kept, because short complaints are still complaints.

You can add your own words to skip, for example the line your agents always begin with.