Don’t Pay More For Data Analysis
When running an efficient team or business, constant improvement is paramount. This of course involves reviewing activities, testing activities and taking hard decisions. And while there certainly is a place for intuition, you also want to make sure to make an informed decision based on your past learnings. After all, this is how grow happens.
These learnings are manifested in data, and the easier it is to digest, and the more specific it is to answer the right question, the more value you get from it. This match of data and decision is not something that just happens. It needs to be well planned from the onset. In the most basic sense you want to have a measure of success or no success to decide on whether to continue a specific activity or not. But there are also questions of: What should be tweaked? Where does failure or success come from? What do we need to do more of? The good news is that recent technological development have made a range of ready-made tools and custom capabilities possible. The bad news is that it’s not often used to the max.
Most businesses don’t maximize their tools to support analytics and decision making
We often see clients approach data the wrong way around. The questions they are asking are: What data do we have available? And what system should we use to collect and store data? A few months later, the question of reporting then becomes relevant with questions like: What views can we pull, how regularly can we do this? At this point, the capabilities are often very restricted and additional efforts are necessary for retrofitting. This can quickly become resource intensive and cost you a lot of money.
Once a system is implemented and a data structure is set up, any modification will be resource intensive. The right approach therefore is to first ask the question: What decisions need to be supported? Only once you have an answer to this question you can start addressing questions like ‘What data needs to be captured when?’, ‘How does data need to be stored?’, Hence database design, supporting system, set up and reporting mechanism and format will be determined and will now be thought through and mapped out before implementation with the end goal in mind.
Now in reality, most companies are already working with multiple legacy systems. Nevertheless, the right starting point remains the question of what decisions need to be supported. You can THEN start working through what you can do within your current setup and whether your current setup is appropriate or not within the context of the decision goal. In order to save money in the long run, data usage must be crystal clear. The paradigm needs to shift from ‘What do we have’ to ‘What do we need’. Otherwise you will continue going from interim solution to interim solution which might be a good short-term fix but will cost you long-term.
Ask yourself a few questions:
- Do you have enterprise wide consensus on big data usage?
- Was the business working together with the IT team to own and manage the data design process?
- Is the analytics team well integrated within the strategic leadership?
If your answer to any of these is ‘No’, you may want to start looking a bit more closely as this could be a symptom of an non integrated approach that might bubble up soon.
How do you get from your strategic questions to big data design?
Thinking through big data from end to beginning can be an overwhelming tasks, that’s why you will usually work with a consultant or data specialist to work through all details before implementation.
At a minimum you want a vendor to set up the technical side for you. This means that it’s up to you to define and lay out the design. This might work if you have a strong analytic team and they are deeply connected to the business. Otherwise working with a strictly technical vendor really isn’t enough. You need a partner who can guide you in translating decision needs into reporting formats and data setup requirements, meaning somebody who can bridge the gap between business and data into a connecting story. If these resources aren’t available in-house, you need to make sure the vendor you’re working with is bringing this capability.
Why? Because you are maximizing your learnings. Because you are keeping knowledge recorded and within the enterprise. Because leadership can confidently take decisions quickly and move the company forward. Because you are not wasting time producing reports manually trying to compare apples to apples. Pretty powerful stuff.
The data triangle
There are three components to big data: A) Reporting B) Data capture which mostly also means a user interface and C) Database structure.
These elements all work together. You can’t report on anything that you don’t capture at some point – be it along a workflow process, be it via manual entry, be it by connecting to other data sources. All data you capture (which ever method) needs to be held in your central database, the central source of truth, in a format that you can report on.
For extra complexity and a reality check – add on the matter of data quality if there is manual input at any stage along the process.
Let’s look at the example of an international marketing team for example and go through the three elements of this triangle.
The most basic report that is relevant here is campaign success. How do you measure success? Depending on the nature and objective of the campaign it could be total or incremental revenue generated, clicks and views, unprompted recall, etc. You will want to look at consumer behavior by channels and by customer segment to enable tweaking and optimization of campaigns moving forward. Maybe there are even other inter-dependencies like campaigns that are running in parallel or macroeconomic impact of different geographic regions.
Now you will be working with multiple data sources here. Most likely it’s a combination of new data capture, integration with other internal systems and then external data sources.
New data capture can be fed by a workflow management system with which different stakeholders along the process interact. Data will manually be input along the process – requests will be submitted, approvals will be given. Note here that user interface design is paramount in determining the quality of data you will get in this process.
Then the marketing team might want to integrate with the email system for example, in order to match any specific campaign to different email sends. This is important to correctly attribute opening rates, click rates and other engagement reports. The same is true for external data sources. Most likely any marketing team will be working agencies (for example PR, paid search) to activate and execute a campaign.
It is important to work with your agency and internal teams to discuss requirements and make adjustments on both ends to reports created. The alternative to this is having to spend a full day of manually pulling reports and reformatting in MS excel and ppt for a presentable report that tells the whole story and thus allows decision making.
This is your single source of truth. A database needs to be sustainable, meaning it needs to be able to grow with you. This is in terms of amount of data captured and stored, this is in terms of speed of returning and capturing data to sustain your workflow management system or other user interface, this is in terms of allowing new kind of data captured as reporting requirements change, and of course in terms of enabling reporting.
Requirements here are individual depending on your reporting and process needs. But these are the generic attributes you want to be looking for.
- mandated formats to capture text, date, numbers, currencies
- as little duplication as possible
- connection to third party data sources
- security, maintenance, backup
You can see that reporting and the underlying data availability is a complex topic bearing unlocked opportunities. Make sure you think about reporting needs proactively from the outside view and you’ll be able to A) drive your company forward and B) make savings in the long run.