Repost from:


Organisations are already using live web chat – and increasingly chatbots – as a critical sales and customer support channel. Gartner says over 80% of companies offer some form of live chat on their website or mobile applications in order to interact with customers, and the chatbot market is likely to grow to $994 million by 2024 according to other market research. Since web chat is both 17-30% cheaper to address questions than a phone call and has the ability to increase conversion by around 20%, this explosion in growth isn’t surprising.

Because web chats and chatbot data can include rich detail about customers – product preferences, complaints or changes to their account – businesses are seeking out more advanced ways to leverage the data in these conversations. The most popular initiative is utilising the data to develop a 360 degree customer profile, which are of the basis of personalised marketing. With 41% of consumers saying they will ditch a company that doesn’t personalise their product/service effectively, bettering understanding customers is no longer just a nice-to-have.

Often the biggest barrier to leveraging web chat and chatbot data, however, is preparing it. Cleansing and preparing this unstructured data and incorporating it with structured customer information has typically required highly skilled data scientists spending enormous amounts of time to bring these diverse sources of data together. This challenge is only multiplied by the rate at which chats are generated, which, for most organisations, can reach the hundreds of thousands each month.

As marketers begin preparing this data, there are several considerations to keep in mind in order to extract the most value from it, ensure its accuracy, and utilise solutions to  accelerate the notoriously time-consuming process. At Trifacta, we call these considerations our tenets of clean data. For those working with web chat or chatbot data – or any diverse  data – we’ve listed them below.

Understand the context to set appropriate targets

Involving individuals with the appropriate context for the data they are working with at the preparation stage is essential to understanding how it should be transformed and what “good enough” looks like from a data quality perspective. For web chat and chatbot data, those who are closely aligned to customers and associated business objectives can better understand the nuances of it, and  focus their preparation efforts and targets accordingly.

Identify issues early and often

When preparing any data – unstructured or otherwise – it’s essential to ensure that the data is consistent, complete ,current and conforms to known standards or patterns. Marketers and their team of experts should be checking every data set against these four Cs of data quality and identify any issues early, and often. Web chat and chatbot data, which is often riddled with typos or mistakes as customers type quickly, is likely to skew the results of any initiative if not remediated. It’s far more efficient to spot these issues early on in the process, rather than when an analysis isn’t delivering quality results.

Allow others across the business to collaborate and contribute additional data

In today’s customer-centric world, there are numerous departments around the business interested in mining web chat data for new insights and contributing their own data to add value to the organisation. For example, combining individual spending patterns from the finance team with positive web chat mentions around “credit cards” or negative mentions about “debt” could help financial services marketers promote certain financial products more accurately. For this kind of data sharing to happen, experts across the business should have the ability to access and transform web chat and chatbot data with data preparation solutions that encourage openness, collaboration and enable easy information sharing. Restricting access to non-PII customer data only restricts the benefits it can generate.

Once preparation is set in motion, constantly monitor

Once organisations automate data preparation workflows to transform this data, they must still continually validate the results of these workflows. They should always look to answer questions like, “is the data that showed up today what we expected? How is it different than what we have historically seen?” and “are variances meaningful?”. This is an ongoing effort that requires automation to ensure data pipelines and the resulting analysis doesn’t degrade unexpectedly over time. It’s vital to keep checking back.

Ensure total transparency throughout the entire process

With increasingly strict and evolving legislation around data privacy, being able to ensure  transparency throughout the data preparation process is essential. It’s not enough to communicate results; you need to show your work, whether to meet external compliance requirements, or for your own internal credibility. This is particularly important for organisations dealing with highly sensitive information like bank details that may be shared over web chat. To ensure your results are secure and can be reproduced, understood and trusted, you have to be able to audit how and when the data was transformed, as well as who transformed it.

With web chats increasingly becoming the customer service channel of choice, the value that they provide personalised marketing campaigns only continues to grow. This data can help power an organisation, but only with proper data preparation practices, as detailed in the clean data tenets.