Analytics without Business knowledge?

I came across a situation while working on a data analytics project. I was asked to come up with recommendations from the data so that the objective is fulfilled.

I saw the previous analysis that was done by others from the team. There they actually came up with insights from the data.(What the data is telling). But that was not actually a recommendations. I mean to say that most of the people I see actually come up with the insights from the data that is pretty straight forward.

But the actually analytics is all about solving business problems right. The insights they gave was so and so entity is doing so and so (from graphs and charts etc…). But when I presented this to the higher official. The response I got was . “Yes my friend even we know that the entity is doing so and so” What next?

Well my question is people nowadays do some courses and get into analytics. But actually, I think they fail to understand the business which is very important. So if we know we can actually tell from the data " Where to focus on to achieve the objectives" and next comes “how to focus on that” Again here one needs to know the business knowledge. So I feel that people who are working in analytics mostly do reporting job and are under the impression that they can completely into analytics.

Hi folks can you share your comments on this please?


What is your model / data telling you?
And how is the quality of the data / measurements?

You can of cause not get a recommendation without perspective!

Even with “business knowledge”, the strategy is on maximizing the impact of the objective which, if successfull, generates more value in form of sales / positive income / return on investment.

To be very confident about the strategy, one should run a model, which learns from a whole bunch of data given the perspective as “goodness factor”. It shall learn form the analytical model environments and the records with reinforcement learning (for example) how to invest to succeed, given a certain probability! - It is a problem, which existed for a very long time in many economical fields, that people do more belief in a single (ofter low positioned) persons optinion, rather than taking responsibility about their given data and given models handling the data.

Many models in classical machine learning do profit from the models “awareness” in terms of confidences or probabilities. They would not give any recommendation if they are not sure enough about their answer.

Anyway those confidences (even if not calculated with a probabilistic model - which would be the best solution) would help giving a “computer guided” recommendation without loading all the weight on your desk just over your own shoulders.

1 Like

Thanks for the comment. I am not very clear in what context you are telling. Are you stating, building models are important than domain knowledge