AUTOMATIC MASTER DATA CLEANSING

We will perform fast analysis of the master data and as a result: define the project scope by excluding inactive or irrelevant items, identify potential duplicates, establish categories of cleansing based on identified classes, determine the necessary resources and timeline for project implementation. Typically, this activity does not take longer than 1 week.

We will apply the following 5 steps methodology to define templates

1. Identify a noun per each record (a defining word of an item, usually the first word in the description)
2. Attach templates per all nouns
3. Define attributes per each template (based on best practice, internal expert databases and current catalogue)
4. Validate mandatory attributes, that are required to be filled in in order for an item to be exactly identified
5. Confirm final templates as a result of cleansing and validation cycles

We will scan the dataset with our Synopps platform and correct more than 10 000 errors in a typical catalogue of 60 000 records, including replacing untypable characters, removing excess symbols, correcting syntax and spelling errors, replacing letters incorrectly used as numbers and many more. This activity will ensure better decomposition in the following step.

An adaptive system of algorithms recognizes more than 220 000 significant blocks such as item classes, characteristics, and their values, part numbers and variations, values via keywords, etc. in a typical catalogue. To normalize and clean these recognized blocks, we apply the following steps:

1. Confirm and validate data blocks
2. Standardize attribute values per templates & cleansing document
3. Redistribute the order of blocks according to the template
4. Classify each record per agreed classifier
5. Categorize each item per normalization type & data enrichment of mandatory attributes (optional)

Identify over 7 000 potential duplicates in a typical catalogue of 60 000 records with Synopps algorithms, that compare classes considering synonyms, compare part numbers and their variations, without considering formatting, and compare values of characteristics, rather than their textual representation. Then validate with our expert team and end-users.

We redesign Master Data Management processes, such as new record creation or editing of an existing record, based on the world's best practices and business specifics. The deliverables include gap analysis and to-be process design, suitable IT solution or excel-based tool implementation, teams training and documentation, hand-over and support to ensure sustainable data quality.

We calculate the forecast of warehouse balances and provide a list of positions that should not be purchased, thereby ensuring a multiple payback of the project. We apply a decision matrix based on a stock position, consumption and purchase requests, in order to identify an invisible stock that should be consumed, and excessing requests that should be blocked to deliver immediate cash savings.