For years I am trying to build a single source of the truth (SSOT). Now I read in "What's your data strategy" by Dallemulle and Davenport about MVOTs architecture: multiple versions of the truth.
But wait a minute, wasn't this exactly what we were trying to root out? Let's have a look.
Data vs information
First Dallemulle and Davenport distinguish between data and information. According to Peter Drucker, information is 'data endowed with relevance and purpose'. It's only when raw data is put in context it is transformed into information that can support decisions.
For example sales figures divided by product group compared by region and put next to last year's sales contains a lot of valuable information that makes it possible to follow up on your business strategy, while the raw data in this may be millions of cash register tickets that are hard to interprete without parsing, sorting or enriching it.
Many organizations start with a top-down, highly centralized, control-oriented architecture. This approach is effective for standardizing enterprise data and is necessary to keep the data safe and compliant. But this kind of data management can inhibit flexibility, making it harder to transform it into information that can be applied strategically.
Linked to this SSOT a more flexible way to use data is needed: MVOTs. They are the result of business-specific transformations of the data into relevant and purposefull information tailored to the different contexts that can exist in the different parts of the business.
For example a customer can mean very different things for different departments in your organization:
- For finance it is someone to send an invoice to,
- for sales the different business units at the customer can need a different sales approach,
- for after sales the different sites of the customer where repairs need to be done are relevant.
All the data related to the customer lives in the SSOT, but for the different departments MVOTs are created to suit their specific needs.
But why then, are we trying to root out multiple versions of the truth in the first place?
The multiple versions of the truth we sometimes find at our clients are not exactly based on a SSOT. The MVOTs at our clients exist not only on the information level, but also on the datalevel.
An example: projectmanagement
Every projectmanager keeps her own excel with her interpretation of timesheets, milestones, deliverables, hours to be invoiced on the project. So there is no governed SSOT on the data level that contains data in line with company wide definitions. This limits the ability to compare projects with each other, the client gets invoiced more or less depending on the interpretation of the timesheets, data exists on different levels of detail between projects, and so on.
How to solve this?
- A first step to solve this, is to model the processes that are followed on projects.
- Then you define a uniform way to perform projects within the organisation and agree upon definitions of the different elements in this proces.
- Next you implement templates, for instance of timesheets or projectcharters, to support the projectmanagement proces.
- After that you can start looking for a software to implement these templates in a more scalable and maintainable manner.
The database of this software contains the SSOT about projects on the datalevel. Now you can start building MVOTs based on this data for the different roles in the organisation that need information on the projects to make decisions. The projectmanager needs all the details to manage the day to day activities. But management only needs a couple of key figures about the project, maybe sorted by projectmanager, customer segment or number of billable hours. The data will be the same, but the context in which the data is presented will be tailored to the needs of the enduser.
In short: SSOT works at the data level, MVOTs use the data in the SSOT to support the management of information.