How can retailers get the most value out of their data.

Fahad Zaidi
Fahad Zaidi 02/03/2020·8 min read

This article was originally published in Data Driven Investor, by Fahad Zaidi, Quorso's Lead Data Architect.

Having spent the better part of the last decade helping organizations in the U.S. and the U.K. use data to drive profit, efficiency, and performance improvements, I thought it would be helpful to jot down best practices for organizations that are looking to get the most value out of their enterprise-level financial and operational data sets.

Every organization entering the roaring 20s knows that they must use their data as a strategic differentiator to gain a competitive advantage. In this article I examine the five things all companies, in all industry verticals, must do to maximize the value of their financial and operational data sets:
  1. Consolidate and map disparate data
  2. Identify, prioritize and personalize insights
  3. Initiate and track all actions
  4. Measure the specific impact of each action taken
  5. Share and scale the actions that work

1. Consolidation.

Common data hierarchy across the enterprise.

Data & Insights are no longer the cherries on top. They are the base upon which all successful organizations are built.
Generating valuable insights of course requires organized and structured data. However, I have been amazed over the years how many household name companies we work with don’t have a common data hierarchy or continuous data transformation pipeline.
Too many companies are being sold multi-year (and multi-million dollar) consulting efforts to clean and consolidate their data, but the reality is that with some focus and judicial deployment of basic ML, for most retailers, cleaning data can take weeks not months and remain valid on an ongoing basis.
Having structured as opposed to messy data will allow anyone to generate and action insights, without having to do laborious, custom data-wrangling each time they need to query their data sets.

2. Insights > Charts.

A focus on creating insights rather than fancy charts and visuals.

Business Intelligence, Enterprise Performance Management, and other reporting suites offer the promise of insights via visuals and dashboards but that’s it.
Visualizing data is a good step forward from manually cyphering through large amounts of data sets using complex queries, but dashboards and charts are not enough.
Rather than focusing on how best to visualize a trend, companies should focus on how best to generate relevant insights using their data sets. Visuals may assist in insight explanation from time to time, but the goal should be to generate actionable insights rather than a fancy chart here and there.
For example, using a line chart and time series to visualize “women’s T-shirt” sales across 300 apparel stores in a large chain, will show us how women’s T-shirts are trending up or down. This is a nice bit of information, but it doesn’t really tell us anything insightful. Overlaying last year’s dataset for the same time period is another nice bit of information, but again lacks actionable insight.
Instead, analyzing a particular store’s sales of “women’s white v-neck T-shirts” against a control group, (ie. stores with the same demographics and the same range, space and pricing for that product), is much more useful. Any variation below the control group median, highlights a potential problem that we know must have either a solution, or a valid explanation.
For example, if one store is selling 30% fewer “women’s white v-neck T-shirts”, it might be due to poor placement. Perhaps the other stores have this product in a more prominent location on the floor, or have placed it alongside a commonly paired item, (eg. “women’s denim shorts”).
On the other hand, if it is due to a factor outside the store’s immediate control, (eg. the merchandiser is not sending the correct ratio pack of “women’s white v-neck T-shirts” for this store’s demographic), then this explanation should be captured and quickly escalated, to ensure central teams are made aware and alerted to rectify such issues.
Wherever the source of the problem, comparing to a control group, at a granular product level, produces real insight — not just another chart. ­

3. Track Actions

Track every action taken on each insight (both good and bad).

Each time an operational action is taken on each insight, it should be tracked.
This is primarily where BI and reporting tools start to fall short.
No BI tool has robust write-back functionality that allows for operational users to track actions that they’re taking and how their actions impact the data.
For example, just having a fancy chart that shows “women’s T-shirt” sale trend lines, won’t allow us to see how sales of “women’s white v-neck T-shirts” have gone up or down at a particular store since the placement of that specific product was adjusted. Indirectly, we can do some manual analysis on standardized charts to find this information, but a fancy front end on its own is not enough to track 1,000s of operational actions being taken by the front line on the insights that the data is generating.
Enterprise-level communication tools (Slack, Teams, Email, etc.) allow users to indirectly share actions, but these actions don’t tie back to the data where the insights are coming from. Even enterprise giants like Microsoft have only gone as far as embedding dashboards into communication tools, but the back ends are not linked and tracking actions is still very much an ad-hoc exercise.

4. Measuring the impact of each action.

How does each action affect EBIT and your strategic Key Performance Indicators (KPIs)?
Simply measuring performance before and after a certain date when an action plan was executed to increase sales, reduce cost, or optimize labor is not enough.
Tracking how actions affect EBIT (profit), operational metrics, and strategic KPIs allows us to truly use insights to improve business performance.
For example, it’s easy to indirectly look at basic trend lines to see if an action plan to drive sales of “women’s white v-neck T-shirts” is increasing or decreasing sales. However, this isn’t the most useful way of measuring success.
What’s harder and more worthwhile, is to track the gap to median of the control group of comparable stores, to truly see if the plan is working, as well as track how this action is affecting the organization’s KPIs and operational metrics. For example, while it is good that sales of “women’s white v-neck T-shirts” are going up, it is as important to ensure service isn’t suffering as a result by tracking other metrics, such as availability and customer satisfaction.

5. Sharing relevant peer actions.

Scale successful actions across the enterprise.

Operational problem solving is a human exercise. There’s lots of technology that can augment good decision making, but this will always be an exercise where human ingenuity is at the core.
People too often still solve problems in silos, especially in large, multi-site, distributed businesses. This problem becomes even more apparent when large, multi-site businesses operate in a franchise model.
For example, if Jane Doe operates apparel store X on Main Street, and has been crushing performance for “women’s white v-neck T-shirts” due to intelligent store placement, optimizing supplier ratio packs, and training her direct reports on how to cross-sell the product with “women’s denim shorts” — this is great; but Mary Doe, one mile down the street, who manages a very similar apparel store operated by a different franchisee, may never be able to learn from Superstar Jane Doe.
This is important because not scaling the most successful actions across the organization limits the organization’s full financial and operational potential.

How can organizations accomplish steps 1–5.

From my experience, there’s no shortage of tools that can help any organization accomplish each individual step listed above.
However, there’s only one tool I’ve come across thus far in my career that helps organizations accomplish all five of the above points in one single tool — Quorso.
Quorso was founded by two former McKinsey & Co. Partners, after they came to the realization that there are roughly 20–30 million people that control 60–70% of the global economy, specifically as it relates to profit. These 20–30 million people are folks that make decisions daily, weekly, monthly, quarterly, etc. that directly affect their organization’s P&L. The number of people and the percent of the global economy this will affect is only going to go up, as more and more people enter the workforce.
At Quorso, we’ve built the world’s first intelligent performance improvement platform that allows organizations of any kind to surface relevant insights, take and track action on each relevant insight, measure EBIT, KPI, and Op Metric performance for each action, and share best practices across the organization when actions are successful.
Since launching our product, the industries that have taken akin to our creation have been Retail, Hospitality, Food & Beverage and Transportation & Logistics.
I joined Quorso as their first Data Engineer and our Data Acquisition & Ingestion engine is capable of working with any type of data source. We then create a common data hierarchy and continuously load data into the platform based on the appropriate time series cadence.
The way we expose relevant opportunities is by using a recommendation algorithm analogous to Netflix, Spotify, or iTunes to ensure the opportunities are relevant to the user based on prior experience, the area of the P&L they control, strategic company initiatives, etc.
We only show controllable opportunities — this is what marketing experts call “actionable insights”.
Every action in Quorso is tracked, and the app is smart enough to share relevant peer actions to the right people to ensure problems are not being solved in silos.
Thus, organizations that use Quorso perform up to their full financial and operational potential.