How to Prevent Your Analytics Solution from Falling Apart

How to Prevent Your Analytics Solution from Falling Apart

by Prasanna Amanan

How to Test Your Strategy

What does fail mean in the context of an Analytics solution?

?Is your analytics project, product or your solution, often finding itself in a situation where you are unable to cater to varied requests from your stakeholders? Do you often have discussions debating the integrity of the data or the solution itself? Does the debate often involve what can’t be accomplished with the current solution? Are you finding it difficult to convince your stakeholders to adopt the solution or do the stakeholders doubt the effectiveness of the solution, does the solution not achieve the intended purpose?

?Combination of the above situations or more, likely indicates inability of the analytics solution to meet the needs of the stakeholders or the end users.

?Let me discuss an example, in an organization which operated in 30+ countries and 70+ location with 10000+ people, having 10 different core products, 4 different business verticals, there was a drive to standardize metrics and KPI of product development, although the products themselves were different, each product solved different problems to different customers, every product at that time was at a different life stage of the product life cycle. Yet the objective was to compare the products, standardize engineering priorities, practices, processes, etc., make the metrics comparable between the products and utilize the metrics for making decisions on product investment and performance of product engineering teams.

?The first point I want to highlight is, using data to make sense is good, using data to make decisions is excellent but it does come with a caveat, data alone will not be your guardian angel, but other factors will play a significant part in decision making and knowing those factors is not only essential but critical to effectiveness of decisions, for all the noise internet makes about data driven businesses we need to accept that, every data driven decision comes with “terms & conditions apply” tag.

What does this mean? In the above case of standardization of metrics, it is no brainer, products with different purposes and traversing different life stages of product life cycle exhibit various characteristics, when viewed from a product engineering parlance it makes it very clear why the comparability between products is narrowed. Thus, decision making about products when using standardized metrics cannot be taken at face value but account for the products life stage, customers, etc. ?

?The second take away here, unless various stakeholders are consulted, heard out, actions taken to ease their inconveniences, particularly their fears, understand their challenges including agreeing on the limitations of the approach and transparently establish how the analytics solution will be used by various stakeholders including senior management, implementing such as system is going to be more discouraging than useful for making meaningful improvements across product engineering group. An important point to understand and accept here is, none of the above constraints discussed has to do with the effectiveness and the soundness of the technical solution however if unhandled effectively across the stakeholders the polarization towards the analytics solution begins at this stage, i.e., polarization is achieved even before the implementation.

?

So, what should you do? What can you learn from the above example and ensure you don’t have to revisit the problem in its same state or its manifestation to other forms?

?Understand the value chain of data science, i.e., using data to address a business problem, how data can be transformed to business value, typically we define a use case backwards.

As a hypothetical example, if you have a team of 5 people handling 80,000 emails in a year and often the team is overstretched, inconvenienced because the emails come 24/7 and they can’t work in shifts due to prevailing external conditions.

?Assumption here is each email requires 4 minutes to read and reply, that leaves with 320,000 minutes of work a year. We have anywhere 384 minutes a day per person to reply to the emails (20% of time 96 minutes reserved for other tasks). A person from the team can at a maximum do 96 emails.

Based on the annual volume 80,000 emails, in a day each member of the team needs to roughly reply to 80 emails a day given we have 250 working days. Implies anywhere between 64 - 80 emails are required to be addressed per person of the team on a daily basis, including sustain days when a team member may be off for the day. Clearly, this situation illustrates the pressure on the team, the challenge to keep up with the workload is significant.

Therefore, here a possible value perspective is to reduce the total time spent on emails (annually, per person, etc.,) which translates to considering possible actions that can make an impact to total emails read, received, including eliminating rework of any sorts. At this stage data insights matter, by grouping the emails received and classifying the repeating patterns let us assume one of the patterns leads us to determine 20% effort goes in replying to emails where the requested information is already available on the website but often unread by the sender. Possible solution requires behavioral change, possibly understanding the issue behind the behavior and designing intervention. Again, let us assume the following actions are taken,

?(a)????????????????we improve the readability of the website,

(b)????????????????add more languages options to the website,

(c)????????????????enable searching the website for answers,

(d)????????????????improve the FAQ section (make it searchable) ?and categorize

the sections to reflect the patterns observed in our insights,

(e)????????????????add keyword matches based on insight,

(f) ??????? make explainer videos, etc.,

?Hopefully these actions help reduce 30% of new inflow of emails, this would mean assuming 10% increase of annual emails leads to 88,000 emails i.e., 352,000 minutes and 30% reduction means the annual volume reduces to 57,200 emails.

Therefore, each member of the team now has to read 46 - 57 emails while the capacity is about 96 emails per individual. Thus, even if not a radical impact to the situation, the current solution should at least provide respite to the team to sustain absence of a fellow team member enabling better workload management.

?Assuming, the organization repeats the learning using data and drives intervention we can hope the situation will be under control in the short-term i.e., 8 - 12 months.

u/t = fn (d, i, a, r) = F(O1)

where n = 1, where O is Observation U = ?F(On) where t > 0, n >1

?However, if an organization wants to benefit beyond an optimization activity, i.e., embrace solutions beyond short-term perspective which considers significant impact to current ways of operating, they may have to consider radical, innovative, out-of-box approach that enables them to outsmart competition. Often, such attempts can be classified as transforming their operational model or rewiring their business model. The Use Case Framework described above is to some extent useful in its potential to unlock value as it depends on the iterations, stakeholder perspectives, etc., above framework is likely the most often employed approach whether a multimillion-dollar enterprise operating in a small region or a billion-dollar enterprise operating worldwide. Use Case Framework embraces a data perspective and applies a data lens to address the business problem or challenge.

?

How to use a stakeholder perspective to unlock value from data for other than optimization purposes, say innovating or redefining problem to transform business?

?Begin with a stakeholder map and understand respective pain points. Capture the decisions the users/stakeholders are allowed to make. Stakeholder interviews are a fantastic opportunity to capture not only pain points but the thought process of respective stakeholders, fears, concern including culture, workability and evaluate feasibility of the possible solution. Define and validate what according to the respective stakeholders is the critical objective, what information would enable users to establish a metric that assists in tracking the progress towards attaining objective.

?If you want to go one step ahead, you may want to develop Data Value Proposition, this would be similar to Customer Value Proposition (i.e., establish the value proposition for each customer persona; what pains and challenges the product solves), the difference here, (i.e., establish the value proposition of the data for each stakeholder, how does the data assist in enabling the desired outcome) map the data types, data sources, value of the data perceived by the stakeholder, success criteria identified by the stakeholder for desired outcome, etc.,. See below example using Data Value Proposition which acts as an input for Analytics Use Case Canvas.

Data Value Proposition Canvas

Recollecting our objective, to navigate the process of innovation using data i.e., redefining problem to transform business. The Use Case Framework is an iterative approach to view the business problem from a data lens or data perspective. The Data Value Proposition Canvas enables validating priorities or objectives of the stakeholders underpinned by value of data. The Analytics Use Case Canvas enables validating analytical applications or solutions from the stakeholders/users’ point of view. The Analytics Use Case Canvas helps to conceptualize and develop?data-driven analytics solutions?which have a?high user acceptance?including enabling more value-add to respective stakeholders. DVP facilitates crafting AUCC.

The elements of AUCC, includes mapping per stakeholder the following: (a) solution vs benefit, (b) objective vs results, (c) action vs decisions. The Solution part of the map tries to capture the analytical solution most appropriate to the business problem assimilating users’ perspective. The Benefit part aims to enable value-add to respective stakeholders. The Objective captures what is absolutely critical to the users’ and what is considered useful for the desired result, it also considers a pain-gain approach (data wise), understanding actions and decisions that may be derived viewing the business problem from a holistic perspective. The canvas captures and facilitates a stakeholder approach to the problem and lays emphasis on data most relevant and preferred by respective stakeholder which in turn leads to insights when designing interventions or choosing possible solutions from a list of possibilities.

?Let us see, how using the Analytics Use Case Canvas, we can identify an ideal solution for the hoteliers’ challenge to keep up with email inquiries that we discussed earlier. Focusing on the Solution, the attempt here is to radically eliminate the present situation of overstretched staff yet be responsive to customer inquiries in a personalized way and bring down time spent on emails leading to cost savings. Observations and analysis captured in Data Value Proposition Canvas feeds into the critical objectives the Solution must meet at a minimum, as this ensures value-add to each stakeholder is explicit which drastically enhances acceptance of the solution.

?Recollecting, insights from data analyzed on emails, we are aware that the patterns analyzed from the emails received gave way to take decisions and design interventions. However, we (humans) had to manually analyze the patterns and act (employ interventions) suitably, of course the dependency highlights, we will be always limited by our time, ability to detect patterns and intervene with probable course corrections as it might be impossible to address all the detected patterns at any given instant of time. In order to radically change the circumstances, we may need to eliminate dependency on humans as much as possible.

?If we want to unlock enormous benefits to all the stakeholders, we want to enable pattern identification as well as intervention are no longer dependent on human capabilities and time, rather employ an automated solution. Holding on to such a line of thought, understanding the current environment, i.e., text scan, patterns, decisions, non-human capabilities, automation, self-learning systems, we can relate these can be achieved by

?Analytics Use Case Canvas, Fig - 3

?computer software through advancements in natural language processing, machine learning, etc., it is important to understand how we accomplish producing such a software is completely different topic. However, what we have achieved at this stage is the use of the Use Case Framework, Data Value Proposition Canvas and Analytics Use Case Canvas we have concretized a possible solution that strongly meets the needs of stakeholder and ensure each stakeholder receives a unique value-add from the solution and above all ensure high acceptance of the solution by design.

?Bonus Story

?Going back to the hoteliers’ example, they went ahead with the machine learning solution and uncovered a new product altogether. A virtual receptionist for the hotel powered by ML, wherein the AI (Artificial Intelligence) enabled hyper-personalized automated enquiries & bookings, instant message to queries received via chatbot for the website and auto configure personalized newsletter based on recipient information obtained from Customer Relationship Management

?(CRM) and other relevant data platform or data sources to understand preferences of the recipient. The recipient received valid, relevant, hyper-personalized, specific communication including not losing out on the human aspects leading to higher success of engaging with the newsletter. You can find out more by visiting www.eecho.ai/de

?I genuinely believe the methods and tools discussed will help you to navigate the analytics solution fit, guide you through problem-solving using a data lens and developing use case that enables disruptive innovation unlocking business value.

EECHO in action, find more about the solution at www.eecho.ai, Fig - 4

Definitions:

Analytics Solution - Refers to a data analytics program used to organize vast amounts of information and change it into practical understanding.

?Business Analytics - Business analytics is a set of automated data analysis practices, tools and services that help you understand both what is happening in your business and why, to improve decision-making and help you plan for the future. The term “business analytics” is often used in association with business intelligence (BI) and big data analytics.

?Learn more about the various types of analytics used in business today here

?

References:

Schep, M., Tonk, S. and Griffioen, H. (2020). Defining Analytics Use-Cases. [online] GoDataDriven. Available at: https://godatadriven.com/blog/defining-analytics-use-cases/ [Accessed 9 Jan. 2022].

Szugat, M. (2021). Analytics Use Case Canvas. [online] Datentreiber. Available at:

https://www.datentreiber.de/en/method/analytics-use-case-canvas [Accessed 9 Jan. 2022].

Dataswati (n.d.). DUCC? - Data Use-Case Canvas (by Dataswati). [online] https://www.data-use-case-canvas.org/. Available at: https://www.data-use-case-canvas.org/ [Accessed 9 Jan. 2022].

White, A. (2019). Design a Data and Analytics Strategy. [online] Gartner. Available at: https://emtemp.gcom.cloud/ngw/globalassets/en/publications/documents/data-analytics-strategy-ebook.pdf.

Sundaram, M.P.J., Ning Su, Robert D. Austin, and Anand K. (2021). Why So Many Data Science Projects Fail to Deliver. [online] MIT Sloan Management Review. Available at: https://sloanreview.mit.edu/article/why-so-many-data-science-projects-fail-to-deliver/.

Acosta, J.M.R. (2021). The Analytics Lifecycle, Analytics Maturity, Analytics Canvas, and Absorptive Capacity. [online] Medium. Available at: https://juanmruiza.medium.com/the-analytics-lifecycle-analytics-maturity-analytics-canvas-and-absorptive-capacity-e0961946bb93 [Accessed 9 Jan. 2022].

For any clarifications the author can be reached on LinkedIn,

https://www.dhirubhai.net/in/prasanna-amanan/?????

?

??????????????????

?

?

?

?

?

?

?