#13 - Get more values by finding the constraints in your data orgranisation
And some thoughts on being customer-focused in strategy, doing machine learning and utilising big data
This Week’s Principle
Finding the Constraint in your operation (link)
Constraint is a leverage point that will help you unlock massive value in your operation. Many operational structures are complex, but there will only be a few key constraints. If there are changes to these constraints, the values generated will be significantly impacted. Vice versa, if changes happen to non-constrained areas, the impact will be moderate.
To find the constraint, analysis of the Metrics Tree is crucial. Coined by
in this article, this term refers to a visual representation of all metrics in an operation (either in a business function, or entire business). Ergest also provided us with an example of a Metrics Tree:The yellow boxes represent conversion rate type metrics. The blue boxes represent absolute numbers. The product of the two feeds into the next level ultimately ending up in the revenue box. The red boxes represent “levers” or “input metrics” which are things you can manipulate to directly impact the conversion rates. - Ergest’s narrative of the Metrics Tree diagram.
There is a lot to be said on how to analyse the Metrics Tree to find the constraints, as Ergest pointed out. Some practical tips I can take away are:
Benchmark and analysis historical patterns of a metric
Look out for metrics that are both upstream and massively off target.
This Week’s Strategy
“When your companies are in an early stage, instead of focusing on the competitors, focus on the customers.” - Jeff Bezos
Perhaps this is not a specific insight for data leaders, but I do love the mindset conveyed here by Mr Bezos. Being the data leader in a customer-focused company, for me, is much more attractive than being in a competitor-focused one. The outlook is more optimistic, and your imagination is not shaped by what your competitors are doing. So as a data leader in an early stage company, what can you do to influence your business partners to become more customer-focused?
Enjoying these bite-size content? Subscribe to Data & Beyond Dispatch to receive notifications whenever a new issue is released.
This Week’s Operation
Start with end user’s need instead of model accuracy
Sometimes a vivid example of a failure in the past can be more effective than a list of 5 best success stories. That was my thought when I came across
’s story featured in this edition of Data Analysis Journal. The story went like this:At DoorDash, Daniel was tasked with developing a predictive model to estimate the value of every restaurant in the world. The model’s estimate would be used by the sales team as leads. The model was developed from a database of restaurants, with the target variable being the restaurant’s performance on DoorDash. (I suppose this would be bookings per month.)
When the initial predictions were made and the first list of potentially high-performing restaurants were passed to sales, everyone was disappointed. Near 20% of the list were unworkable leads, some of them were not even restaurants. The team discovered that the low quality dataset caused the issue, which cannot be compensated through model accuracy like Daniel initially thought. He soon realised he was mistaken, and for the next year, the team worked on nothing but database quality.
“My major learning was to consider the end user of a model or system when launching a project” - Daniel said, which I couldn’t agree more. Throughout my career in the tech team, I couldn’t help noticing that engineering or data science teams often get bogged down on the technical details, without a clear sight of the end user of their products. Reading Daniel’s story, my inner marketer is clapping, cheering and nodding - yes, focus on the end user first.
---
I also pleasantly discovered Daniels’ data story-telling newsletter, which used data brilliantly to tell stories on pop culture. It’s a fun and stimulating read, check it out here:
This Week’s Impact
A Rib Special for you - delivered by Barbecue chain’s Big Data [1]
Dickey’s Barbecue Pit, which operates 514 restaurants across the US, have gathered a large amount of data from various sources and brought that data together to maintain a competitive advantage. The data comes from point-of-sales systems, inventory systems, loyalty programmes, as well as customer surveys and marketing promotions.
The restaurant chain built a proprietary big data analytics system - Smoke Stack - to crunch the numbers from various sources, and provide near real-time feedback on sales and other key performance indicators. The feedback is broadcasted to users across the companies, from boardroom members to restaurant floor, whose access to data would help them do a better job. This process is enhanced by a flexible, user-friendly platform, making it especially easy for non-data savvy people to understand the insights.
These feedback are translated into instant actions to course correct, such as:
When sales is lower-than-expected one lunchtime, and there are an amount of ribs available at the restaurant, the team send out text inviting people in the local area to come for rib specials.
Where sales at a certain store falls short of expectation, the team send out training and operational support.
User’s feedbacks is used to select menu items, based on simplicity of preparation, profitability, quality and brand.
Course-correct action is often taken every 12 - 24 hours, instead of weekly (or monthly, like in many businesses). The chain’s management is able to make quick decisions, leading to increased savings and revenue.
[1] Insights summarised from Chapter 27, Big Data in Practice, Bernard Marr.
That’s it for this week! If you enjoy or get puzzled by the content, please leave a comment so we can continue the discussion. Throw in a like as or share as well if you know of someone who may enjoy this newsletter. Thanks!