A Look Back at 2019…
Happy New Year everyone! As we kick off the product activities for 2020 and beyond, I thought it would be a good idea to review the progress we made with Cuddle in the past year. 2019 was a momentous year for us and in terms of our product evolution. Here’s my take on some of the most interesting progress we made on Cuddle.ai in our endeavor to make data and analytics more accessible for the business user.
1. We Made it Easier for you to Find Data You Need
One of the key tenets at Cuddle is giving users the ability to self-reliantly query, analyze and decipher insights in their data. Current analytics platforms — built from the ground up for the technical data analyst — stymie and overwhelm the business user. Continuously challenging ourselves to make it easier for a non-technical user to discover data insights, with minimal effort is what sets Cuddle apart from the crowd and brings much love from our customers. We doubled-down on making it easier for users to search through their data by:
- Enhancements to Our Natural Language Engine (ASK): We performed extensive user research to thoroughly understand how our users ask questions to our natural-language question-and-answer engine. We discovered that sales managers typically want to find their top/bottom performing reps without the fuss of pulling out a list and then sorting from highest to lowest. We also found that CXO users on Cuddle ask pretty complicated questions that include multiple measures and time periods. We took their feedback to enhance our NLP engine and add in support for questions with complicated time periods (Sales from Monday to Wednesday), multiple measures (show me sales, discount, distribution and margin by brand for this quarter) as well as address “top/bottom” questions (bottom 3 sales reps by dollar sales this week)
- Auto Analysis Summary: User research strikes again! One of the crucial questions asked by one of our users was — if I had a business analyst, she would not just answer the question I asked, but would go further and curate the additional data that gives greater context. So immediately in the next version of Cuddle, we prioritized building an auto-analysis summary under each answer that provides additional context to the answer (trend, drivers, key attributes and related measures). Doing so meant that users did not have to constantly keep asking questions each time, plus found it easier to discover driver measures that they had not previously considered!
Check out more information on our Natural Language ASK engine here.
2. We Made it Easier for Data You Need to Find You
The typical business user that works with Cuddle is exceedingly busy. She has a business to run and people to manage. Expecting her to constantly do a deep dive into her data every day is simply not the right expectation — and one of the constraints that current BI and Analytics tools don’t solve for. The way we addressed this bottleneck was through an innovative feature — called Nudges — that proactively alerts business executives when there are any anomalous outlying conditions within the business areas that they manage. Nudges helped us flip the old paradigm of users discovering data to data discovering users — and has been hugely well received by our clients. However, innovation at Cuddle didn’t stop there and we took it up a notch last year.
- Trend Consistency and Unexpected Events: Anomalies that need attention come in all shapes and quantums. While Cuddle could earlier identify a significant anomalous change in data vs period ago, using machine-learning algorithms, we knew that our users would also need to observe trend consistency and trend reversals in a shorter, continuous period. To that end, we significantly expanded our definition of ML-based anomaly detection to include trends over a shorter period of time that require attention.
- Drivers and Drill Downs: A tip of the hat to one of the sales conversations I had with a large Fortune 500 company. I was demonstrating a Proof-of-Concept of Cuddle using his sales data where I showed him a Nudge highlighting an unexpected trend reversal in the average selling price for one of his brands. His immediate next question was — Why? Why did the sales drop? We knew at that moment, that standalone Nudges are insufficient and we need to use AI to give more context to why the anomalous condition occurred. Our response to this was to create an automated ML-model that explores driver measures to look for correlation (to answer the ‘why’) as well as searching down the attribute hierarchy to surface hotspots where a trend is more prominent (to answer the ‘where’).
Get more information on how Cuddle Nudges you on the information you need to know here.
3. We Made it Easier for You to Analyze Information
Cuddle is first and foremost, an analytics platform. So deploying major upgrades that help our users better analyze their data was a crucial part of what we need to do.
- Improved Visualization Library: At Cuddle, we had a huge incentive to disrupt the existing analytics software industry by creating appealing, intuitive visualizations on the mobile. We enhanced Cuddle’s ability to visualize data by introducing multi-series line and bar charts, a unique way to visualize achievement against goals among others.
- Support for Non-Aggregatable Measures: Non-aggregatable measures i.e. KPIs that cannot be aggregated over certain dimensions or time periods — are extremely poorly dealt with in most analytics software out there today. An example of a non-aggregatable measure is probability — which cannot be simply added up across dimensions or time-periods and needs to be managed differently. Similarly, we have measures such as headcount which cannot be aggregated in certain cases. For example, if your department headcount for January is 200, February is 250 and March is 300, then your headcount for the quarter is 300, and not an aggregation of the past 3 months. A simple expectation from an analytics platform that Cuddle is solving for. More details on this here.
4. We Made Huge Strides Towards Platformizing Cuddle
Our vision for Cuddle is and always has been to graduate towards a completely self-service model. One that saves millions of dollars in maintenance contracts to implementation / services vendors, while not requiring an army of internal analysts to set up and run. In 2020, we will be launching the command center on Cuddle (essentially an admin module) for clients’ internal stakeholders to manage usage, user provisioning, activity and data setup — while running the fine balance between expectations of high sophistication and low complexity. The strides we made last year included:
- Multi-Cloud Strategy: In the end of 2018, Cuddle was only available as a SaaS-based, vendor managed platform on Azure. It was essentially us paraphrasing Henry Ford who is said to have quoted that his customers can have their car in any color, so long as it is black — essentially an unscalable way to grow any business. Azure was our platform of choice, but it was always the step one towards a multi-cloud strategy. The services we use are essentially replicable and available across all major IaaS vendors. Today, Cuddle is available as both a vendor-managed and a client-managed service, on Azure, AWS, and GCP.
- Multi-Tenancy: The second major change we made was to introduce multi-tenancy for Cuddle-managed customer deployments. Multi-tenancy allows us to keep our costs low and in turn keep our prices low — and makes it easier for each customer to receive all of our upgrades simultaneously while providing increased scalability.
There’s still so much to do as we set out to build the future of BI and Analytics, one that is powered by AI and really furthers the cause of data-driven decision-making without all of the bottlenecks that we see today. I’m immensely excited about the year ahead and what the future holds for us and our clients.
Onwards and Upwards!