Why Accuracy SLAs can create or destroy the value of your service

SLA literature in the marketplace waxes eloquent on topics like Availability and Performance. However one of the most ignored topics in an increasingly data-driven world are service levels that deal with Accuracy. Not paying attention to demonstrating accuracy can poke large holes in the value of your cloud and big data solutions. Here is how you can address such gaps.

Connecting to the Interweb Tubes

At first glance, Accuracy sounds soft and qualitative. A recent deep dive into this topic forced me to look the dimensions of Accuracy and I emerged with two aspects: Data Accuracy and Process Accuracy.

How Accurate is Your Data?

Accurate data is the basis of decision-making. In today’s world of big data and cloud enabled applications, where data resides physically in multiple locations, data accuracy is of prime importance. Lets look at two aspects of measuring data accuracy integrity and recency.

Data Integrity

  • This is a measure of how data is protected against corruption through logical errors, user input errors or hardware errors.
  • If data integrity cannot be ensured, this has a severe backlash on the quality of service that your application is providing.
  • A system which cannot guarantee certain levels of data integrity is of not much use even though it might satisfy high performance and availability SLAs.
  • So while ensuring that your application performance and availability, also ensure the same for your data.
  • So how do you measure data integrity? Data Profiling is a common approach towards measuring data integrity.
  • There are multiple technical solutions (as a Google Search on “measure data integrity” will reveal) which I will not cover in this blog post.
  • Focus on how to demonstrate measures for Data Integrity with your SLA Definition. 

Data Currency:

  • In an information-hungry world that relies on big data and predictive analytics to solve problems, the rate of data gathering and capture is increasing exponentially.
  • Data in such real-life databases can become obsolete rapidly.
  • Capturing data across various dimensions can sometimes led to multiple values of the same entity sitting in a database.
  • What is worse: some of these values would have been one correct – but most may have lost their recency and turned stale.
  • This can skew data-driven decisions badly especially when layers like predictive analytics pre-process data and you rely on the interpretation.
  • Sometimes such interpretations cause automatic algorithms to take actions which worsens the problem.
  • With distributed databases and data-warehouses spanning across different locations, latencies can introduce data currency errors too.
  • Especially in a high volume transaction system, such measures are critical.
  • If this is your world – then your SLA Management should demonstrate how good your application or your service is able to correctly identify the current value of an entity and answer queries with these current values, in the absence of timestamps? 

The Human Side: Process Accuracy

We should not forget the human side of data handling – this is where the second aspect of Accuracy comes in. And this is process accuracy. How accurate is your data assimilation process?

  • A typical data-warehouse system relies on multiple data feeds.
  • The number of such feeds continuously increases as the complexity of the application and data landscape increases.
  • Most organizations have very complex Extract-Transform-Load stages that make logical sense of the conglomerate data out of such feeds.
  • These are often very complex job control algorithms that are built in the form of workflows.
  • As the number of feeds increases, the complexity of such algorithms exponentially rises.
  • This reaches a point that logical errors creep in due to human design. This article talking about ETL architecture will give you a feel of how human intervention and decision making can impact otherwise sound data.

The human impact of your data

  • Performance data is an excellent example to explore the human impact of data.
  • Such data is the basis for financial rewards and career-making decisions.
  • To demonstrate value in such an environment,  you have to be able to demonstrate the accuracy of:
    • people filling forms or data in a database,
    • whether the right and complete data is being extracted for analysis,
    • whether all data is being used for analysis? what analysis algorithms are being used? Are they applied uniformly?
    • How is this analysis being interpreted? How are conclusions being drawn?
  • If your service is a Human Resources Platform as a Service offering, Accuracy measurements and SLAs for each of the above questions is critical to the value that you are able to offer
  • Sometimes this can be more important than the performance and availability of the system that you are running. Stacey Barr in this article raises some important aspects of the human side of data.

Are you creating value with Accuracy?

Depending on how data intensive your service is (large volumes, transactions, data-warehouses etc.), the concept of Accuracy will play a large role in how your service is being perceived.

Formulating an Accuracy SLA Definition is very situation-based. There is no industry standard. The environment that your service serves will show whether you should you be looking at duplication? or consistency and synchronisation? or data coverage?

Just like Performance SLAs, you are on the right track when you study the needs of the business that you are serving, and then look at how these needs depend on the different quality dimensions of data in your service that you offer. Here is your opportunity to demonstrate the value you are creating in numbers.


How data intensive is your service? Have you explored how Accuracy based SLAs can create or destroy the value of your service?

photo credit: nickwheeleroz via photopin cc

Speak Your Mind


Wordpress SEO Plugin by Wordpress SEO Plugin
%d bloggers like this: