Text Analytics: The Digital Conversations You’re Not Reading
A retail bank we worked with had asophisticated speech analytics program. Every call analyzed, sentiment tracked,compliance screened, topics categorized. They were genuinely ahead of theirpeers on voice. Then we asked a simple question: what percentage of yourcustomer interactions happen on calls? The answer was 38%. The other 62% —chat, email, secure messaging, social — flowed through channels withessentially no analytics at all.
They had built an excellent analyticscapability over a minority of their customer conversations and left themajority unread. The chat transcripts were stored and never analyzed. The emailthreads were handled and forgotten. The messaging conversations were resolvedcase by case with no aggregate view. The bank could tell you the sentimenttrajectory of its phone customers in real time and had no idea what its chatcustomers were experiencing.
This is the current state of most contactcenter analytics. Voice got the investment because voice was where analyticsstarted. The shift of customer interaction toward text-based channels happenedfaster than the analytics followed, and most contact centers are now analyzingthe shrinking channel while ignoring the growing ones.
The analytics gap between voice and text isn’t because text isharder to analyze. It’s largely historical and organizational.
Speech analytics was a harder technical problem, so it attractedmore dedicated investment and tooling. Text, being natively machine-readable,was assumed to be easy — which paradoxically meant it got no dedicatedanalytics attention at all. The hard problem got solved; the easy problem gotignored.
Text channels grew up separately. Chat, email, and messaging wereoften deployed by different teams, on different platforms, at different times,with no analytics layer built in. Each channel became its own silo, and no oneowned the aggregate analytical view across them.
The volume crept up gradually. No single quarter saw text overtakevoice. The shift happened slowly enough that no one had a moment of recognizingthat the analytics investment was now pointed at the minority channel.
Conversation analytics appliedto text channels reveals the same categories of insight as speech analytics, acrossthe channels where most interaction now happens.
Issue patterns across channels. The same customer problem oftenappears differently in chat than on calls. Text analytics surfaces the issuescustomers prefer to raise in writing — frequently the more complex or sensitiveones — which are systematically absent from a voice-only analytics view.
Sentiment in writing. Customers express frustration differently intext than in speech, but they express it clearly. Sentiment analysis on chatand email captures dissatisfaction that never reaches the phone, including fromthe large population of customers who will never call but will abandon a chatin frustration.
Compliance in text. Regulatory requirements apply to writteninteractions as much as spoken ones, but most compliance monitoring is builtfor voice. Disclosures, consents, and prohibited statements in chat and emailfrequently go unmonitored entirely.
The cross-channel journey. Most importantly, text analytics combinedwith speech analytics reveals the full customer journey across channels. Wecovered the cost of channel-switching in our piece on the omnichannel tax. Youcan’t see that journey at all if you’re only analyzing one channel.
The case for text analytics is ultimately a coverage argument. Ifthe goal of conversation analytics is to understand what customers areexperiencing and to improve it, then analyzing 38% of interactions leaves themajority of customer experience invisible.
The 62% in text channels isn’t lower-value. In many businesses it’shigher-value — text channels skew toward younger, more digitally engagedcustomers, often with longer relationships and higher lifetime value. Theanalytics blind spot is concentrated exactly on the customer segment thatmatters most for future revenue.
And the channels keep shifting. Every quarter, more interactionmoves to text. An analytics strategy that doesn’t cover text is becoming lesscomplete over time, not more. The voice-only program that covered mostinteractions a few years ago covers a minority now and will cover less nextyear.
1. Calculate your actual channelmix. What percentage of customer interactionshappen on calls versus chat, email, and messaging? The voice share is probablylower than your analytics investment assumes.
2. Check whether your text channelsare analyzed at all. Chat and email transcripts areusually stored. Are they analyzed in aggregate, or just handled case by caseand forgotten?
3. Compare issue patterns betweenvoice and text. Pull the top issues from calls andfrom chat separately. The differences reveal what your voice-only view has beenmissing.
4. Audit compliance coverage on text. Are regulatory requirements monitored on chat and email the waythey are on calls? In most centers, the answer is no, and that’s an exposure.
5. Map one customer journey acrosschannels. Pick a customer who used multiplechannels for one issue. Trace it. The journey is invisible to single-channelanalytics and it’s where the real experience lives.
The conversation analytics revolutionstarted with voice because voice was the hard problem. While the industry wassolving it, customers moved most of their interactions to text, and theanalytics didn’t follow. The result is a generation of contact centersanalyzing the channel they used to live in while the majority of customerexperience flows, unread, through the channels they live in now.
The AI terminology chaos is real. Your "divide and conquer" framework is the clarity we needed.
Finally, a clear way to cut through the AI hype. It's not about the name, but the problem it solves.
