Data lake

Unstructured data is information with no predefined schema: call recordings, chat transcripts, email bodies, open-ended survey comments, app reviews, and social posts. It contrasts with structured fields like ticket category, CSAT rating, or handle time, which record that something happened but not why. In CX, the diagnostic detail lives in the free text. A ticket disposition coded “billing” only becomes useful once the transcript shows customers describing a confusing invoice layout. Voice of customer analytics depends on reading this text at scale. OnClarity’s Voice of Customer does this for Grubhub, unifying diner feedback into one view. Before analysis, unstructured text typically needs data cleansing to remove duplicates and standardize formatting.

Root cause analysis is the practice of tracing a symptom, such as a spike in complaints, back to the underlying process, policy, or product defect that produced it. The alternative, resolving each ticket in isolation, never stops the symptom from recurring. Established methods include 5 Whys (repeated questioning to reach the originating cause), fishbone (Ishikawa) diagrams (mapping possible causes by category), and Pareto analysis (isolating the small number of causes driving most of the volume). In CX, this starts with clustering tickets by theme, then tracing the largest theme to its source. OnClarity’s Voice of Customer automates this by surfacing root causes directly from feedback. Grubhub uses it to turn feedback into insights that shape its roadmap. The real test of success is recurrence: a fix that addresses the root cause should stop the issue from reappearing.

A vector database stores embeddings, which are numeric representations of text or audio. It retrieves results by semantic similarity rather than exact keyword match. It uses approximate nearest neighbor search to find the closest matches quickly, plus metadata filtering to apply permissions or source constraints before results are returned. In customer service, this is what grounds AI answers in a knowledge base. A customer asking why their card was declined retrieves an article titled “Payment Authorization Failures” even without using those exact words. OnClarity’s AI agents rely on this retrieval layer to keep responses accurate and sourced.

Data cleansing is the process of correcting or removing records that are duplicated, incomplete, malformed, or inconsistent before they enter analysis. Standard operations include deduplication, standardizing formats (dates, channels, agent names), validating entries against defined rules, handling missing values, and reconciling customer or ticket identifiers across source systems. Skipping this step lets duplicate tickets inflate complaint volumes and distort topic rankings. For example, a customer’s chat and email thread about the same issue should be merged into one case before counting contacts. Cleansing differs from data enrichment, which adds new attributes rather than fixing existing ones. It often happens as a first pass over unstructured data, before any topic or sentiment model runs.