Cleansing your data for upload

1. Cleansed data

2. CRM solution

1. Cleansed data

To get the most value out of the data you send to the Wordnerds Platform, it is best to clean it up before sending it. Ideally, ensure only the data you want to analyse is contained in the 'content' cells.

Email data:

• Make sure you filter the raw email content for the useful part, you want the plain text version of the email (not HTML and not a raw email meant for your email client)
• You normally won’t want a full email thread every time. Ask your team to filter for just the most recent message in the thread. Libraries like this can be helpful.

Call transcripts:

• Are you interested in both sides of the conversation? Normally the useful part is what your customer is saying.
• Are you interested in the full call transcript? The useful information is normally at the start, this can clear up noise in your data and reduce your costs.

CRM:

• Often, structured data (standard questions, account numbers, names) are brought across as text content, this will give worse results. This kind of data can be brought in as metadata or themes. Text content should normally be words directly from your customer. If uploading this data manually, we have suggested some formatting tips below. 

2. Manual CRM solution

The issue we're trying to solve is that any cell in a spreadsheet that is uploaded as 'content' will be taken at face value. What content more typically looks like is an open text response to a survey question, where the actual question would be in a separate cell. What CRM content currently looks like is a lot of different fields merged into one. 

We've highlighted this with the example below. This is one cell that has been uploaded as 'content'.  Which means the platform understands this as one open text response. If this was a survey, it would be the response to one question. It takes everything in this cell at face value. The highlighted sections show where the data may belong with the yellow sections showing text which is more typically classed as open text 'content'.

What impact does this have?

• Your overall sentiment for that response is based on ALL language in that cell, so even those that are part of the basic format such as  'sent', 'subject'.

• It means that your unsupervised topics will be heavily impacted. Their job is to surface the most used phrases and words in your content. That means you are likely to have a topic called 'email' or 'sent'. 

• If you want to drill into what people are saying you probably just want to isolate the open-end text. 

Below is a suggestion for how to split out that one cell for upload to Wordnerds:

written by Lyndsay, Senior Customer Success Manager