Our
difference
You don’t need to be a data scientist to understand the common sense behind these differences—each small in its own right, accumulating to deliver significant value. But if you have some data scientists in your team, we’d be happy to dive into the weeds with them.
Recovering lost profits
A lot of retailers price their discount strategies using a simple matrix like this one.
Often adhering to a time-based, pre-fixed rule; such as 10% discount after two weeks, 20% after four weeks, and so on.
What we know is that profit lost trading margin for volume is not easily recovered. For example: a 30% discount on a 40% margin SKU requires as much as four times the sales volume to maintain profitability.
Self-configuration
Set models
Real-life events that change “what was expected” mess with forecasting models, because forecasting models exist to tell us “what is expected.”
One example of this upheaval was the COVID-19 pandemic. Almost overnight, the retail environment—from brick-and-mortar stores or e-commerce—was fundamentally changed.
Forecasting models that relied on “what was expected’” data were rendered useless because they couldn’t adjust to, or account for ”what was not expected.” In the end, these models had to be manually reconfigured, once the chaos had settled, and new stable patterns began to emerge.
Real-time models
Tymestack didn’t have this problem. It recognized the concept drift and adjusted to the new behaviour automatically.
Recent events, like the pandemic, have made the world feel like a less predictable place. We predict that this unpredictability will only increase, so we’ve been developing Tymestack to operate in times of great uncertainty.
And we’ve delivered. Tymestack now has the unique ability to incorporate brand new data streams that have no history, no patterns, and no predictability. We’re building Tymestack to be the go-to model for uncertain times.
Normal/expected behavior
Concept drift
Discounting
Too late, and too hard
Due to the sheer volume of SKUs large retailers have to manage, underperformance of individual SKU sales is often identified too late - frequently leading to large, rounded-up discounts being applied to shift stock rapidly.
Early diagnosis and action
Tymestack is able to identify sales underperformance early on.
Small, early price changes keep sales on track and avoids panicked over-discounting later on.
End-of-life/seasonal discounting
Opportunity lost -$320
A heavy discount applied late in the season produces results, at a great cost to margin and too late to reach the sales target
Dynamic discounting
Opportunities recovered +$240
Underperformance is identified early in the season and a small discount applied. A small mid-season discount maintains this momentum
20 units are sold at full price. 30 units are sold with a 10% discount.
The remaining 50 units are sold at a 30% discount. All 100 units were sold.
If each unit cost $5 and should have sold for $10 at full price, the profit
is $320, the opportunity lost is $180—making the GMROI 64%
50 units are sold at full price. The remaining 50 units are discounted by 40%.
At the end of the season 30 discounted units sold, and 20 units remain unsold.
If each item cost $5 and should have sold for $10 at full price the profit is $180,
with $100 worth of stock on hand—making the GMROI 36%
SKU relationships
Dynamic neural relationships
Humans intuitively understand that Flippers and a Mask & Snorkel Set are connected. A database lists these items as two distinct SKUs - so the connection needs to be manually labelled for a computer to "understand" a connection.
AI offers a new paradigm - the ability to analyze data and automatically identify relationships between different SKUs.
A complex web of weights and impact
Even if the connections are made, the neural network presents complex problems. First, how much will the price of one SKU affect its own performance. Secondly, how will it affect other SKUs? Thirdly, which impacts are one-way?
These relationships influence which SKUs we discount, by how much, and when. When this is done across every SKU in every store, we significantly improve margin and stock sell-through rates.
Price elasticity
Static price elasticity
Most price optimisation
models calculate price elasticity separately from demand. If the model calculates a 10% discount will result in a 25% sales lift, it locks that result in.
This fixed ratio then gets applied every time, no matter the timing or environment.
Combined dynamic price:demand elasticity
Tymestack does not lock in price elasticity. Instead it uses POS data to model price dynamically. If available, it can also directly factor in hundreds of secondary inputs such as time, weather, and market trends. The result? A precise model of consumers responding to price changes in real-time.
Static price elasticity
Assumption:
That the same discount always results in the same sales effect. Even if that discount is applied three months apart
That the same discount always results in the same sales effect. Even if that discount is applied three months apart
Dynamic price elasticity
Reality:
Understanding that the same discount can have a different sales effect
Understanding that the same discount can have a different sales effect
Granular discounting
Blanket discounting
Blanket discounting is a practical compromise heavily used by retailers who discount to drive volume. It exists because calculating unique discounts for individual SKUs is too complex to do efficiently.
Combined dynamic price:demand elasticity
Tymestack has the computation power and algorithm complexity to calculate unique discounts at an individual SKU level.
Small margin recoveries accumulate to significant gains.
A fixed 20% discount across all SKUs
Optimal discounts calculated for each SKU
How we work
The start of every relationship is us proving our ROI with a simple A/B test against your current methodology.
Understanding your business. Scoping A/B test and deployment. Calculating potential return
Data pipelining. Minimal client involvement and seamless system integration
No client-side IT upgrades, process, or
infrastructure changes
Interfaces customized for business objectives and critical roles
A/B test. Confirm competitive advantage and significance of ROI
Scaling