Bringing the promise of ML to your MDM : Part III
We had studied the broad feature set of AI Data Cleanser in the last article and saw how it addressed key challenges to help enterprises manage their data better.
If you’ve been following along so far, congrats!
This could be an interesting read for you – as we look some of the interesting use cases our customers have challenged us with.
We will cover 3 implementations addressing very different industries and customer needs – the situation that motivated the problem, the gap that prevented a straightforward resolution, AIDC implementation and the end result/impact.
As mentioned earlier, we generally deploy a subset of the different components of the product, as per the requirement.
With that said, let’s dive in!
Use case 1: Helping a large industrial firm validate and cleanse 500 thousand customer addresses across 25 countries, thereby helping them improve last mile delivery to customers.
|The client leveraged a product licensed on an annual basis, to validate its customer and vendor addresses||1. The product functioned in a black-box manner, without a clear definition of validation sources or criteria for validity of a record||1. We built a custom framework to help the customer leverage industry standard address verification sources by country|
|The product was unable to validate addresses in key growing markets, which led to operational impact for sales teams||2. There was no conclusion drawn on the business name located at an address||2. Incorporation of confidence metrics and validation categories helped end users understand the output better|
|3. User feedback for address validity/categories could not be provided||3. Coupled with the customer mastering solution, a validated set of records were created|
Use case 2: Deploying a contact cleansing and enrichment solution for a technology firm deployed on salesforce, thereby helping them map validated contacts to internal sales teams
|The client procured leads for marketing initiatives from multiple internal and syndicated sources.||1. Different sources presented different data formats, naming conventions, and standards.||1. We built a data discovery layer to ingest, unify and standardize data from over 12 sources.||1. The client was able to now leverage the solution to access a centralised, mastered list of contacts|
|The challenge lay in tying these siloed sources together, validating, cleansing and mapping leads to teams in order to plan, measure and track campaigns.||2. Due to manual inputs in the process, there were multiple duplicate and redundant records.||2. Deployed AIDC’s contact cleansing module helped identify and master a large number of duplicates, errors and exceptions from the dataset||2. Campaign tracking and reporting was made available to business leadership|
|3. A large number of leads were unverified and had insufficient attributes to execute a personalised campaign.||3. The cleansed dataset was then mapped with D&B to verify and enrich contact attributes||3. Subsequent pilot campaigns saw an average lift of 5% in click through rates, measured through multiple A/B tests|
Use case 3: An interesting implementation for a large retailer; an automated product cleansing and hierarchy solution which transitioned the customer from an internal system to the GS1 standard
|The client is a large retailer based in the EU, that wished to shift its product hierarchy master data to the GS1 standard||1. Over ~28K unique SKUs had to be mapped from an existing hierarchy to an equivalent GS1 hierarchy||1. We identified alternate sources to verify and enrich the existing product data (product catalogs, e-commerce portals etc.)|
|This would help the retailer improve data accuracy and integrity, speed up supply chain responsiveness and simplify reporting across product categories||2. In many cases, product attributes were missing/inaccurate/insufficient||2. We deployed a custom ML-based mapping algorithm that matched products to GS1 with 3 levels – targeted match, synonym-based match, augmented match|
|3. There was inherent variation between the client’s internal hierarchy and the GS1 standard||3. User feedback was incorporated in sample outputs to re-train the model|
These are just some of the types of problems that have been solved with AI Data Cleanser.
We hope you have enjoyed this series of articles detailing our offering to help your business improve the quality and reliability of data.
For any queries or a free demo, drop us at note – [email protected].