“Consolidation of customer data requires great methodological accuracy”

Whether it is for creating customer warehouses or using them directly in marketing campaigns, data standardization is the absolute pursuit of any company that wants to develop knowledge of its customers or communicate coherently with them. Consistency that must be expressed over time, in the brand’s product or service space, and across the communication channels used.

This goal, “data collection,” is expressed quite simply. And since the human brain knows how to do this effortlessly, automating it seems easy. However there is nothing. In fact, our brain is an expert at recognizing similarities that allow everyday recognition of objects and people, and applies them naturally and effortlessly to the data presented. But the diversity of the data and above all the heterogeneity of the data makes it difficult to automate the subject. So it is necessary to put a little order in it in order to succeed.

Standardizing customer data requires great methodological accuracy. Three basic steps in this process.

1) data normalization

The same information can be represented in multiple ways. For example, a date can be stored in several ways: in a text format, with many possible representations (often language dependent), in a numeric format, with a number representing the seconds or any other unit of time since a reference date. After today, it is also necessary to encode the local time used or its equivalent in UTC. Lots of variations and complications!

So this date field must be normalized, so that later one customer’s anniversary date can be compared to another. This is what the first stage of standardization will do, from scripts or using specialized tools like CDPs.

The same applies to postal addresses, used case or character encoding, etc. This step aims to ensure that the same object has only one possible representation and is therefore comparable.

“The idea of ​​depositing data in raw form in huge repositories, in the hope that the miracle AI will pull the big marrow, has now persisted”

2) Level one leveling

Once the data has been normalized, similar records must be reconciled to eliminate redundancy and then standardized. Starting with the raw records and applying priority rule sets, parameterized algorithms will decide merging opportunities, prioritize, and thus the outcome (also called a “master record” or “golden record”). Applicable rules may depend on the use case provided. These choices also depend on the industries they apply to and the availability of specific settlement keys, or combinations of keys (order number, vehicle registration number, bank contract number, email, loyalty, etc.).

The result is a set of unique customer records, which are a good representation of key data.

3) Level two level

Since the client cannot be reduced to this single “fixed” record, it is then important to attach it, and thus standardize its information (transactions, web logs, etc.), i.e. to attach data sets to it. This step is very necessary because it alone makes it possible to reconstruct the aggregate information: average customer basket, cross-channels, identify the most purchased products, RFM segments (for recency, frequency, quantity), etc.

By keeping track of the primary merged records (“lineage”), it is actually possible to link all records associated with the master record (web records, transactions, orders, shipments, etc.) and thus get the full picture. , equivalent to a cleanly reconfigured client file, with all its attributes. This form then becomes the basis for any new action with the client.

This basic data

If data is a company’s treasure, then it constitutes a perishable asset that in practice requires real care to preserve, and even develop, its intrinsic value. The idea of ​​depositing data in raw form in huge repositories, in the hope that a miracle AI will pull the big marrow, has now taken off. The implementation and use of customer data in the company will pass through the implementation of value-added operations, the standardization of which is the cornerstone.

This standardization will require all departments (business, law, data, and IT) to participate in a project that will generally be led by IT departments on the basis of common goals.


Written by Stephane Deusch,
CEO and Co-founder of imagino

Do you need additional information?


Mind study service upon request

Leave a Comment