All posts by Marilyn Bernardo

Measuring What Matters

Retailers should focus on supply chain outcomes, not just forecast accuracy metrics, when gauging success of inventory optimisation efforts.

Retailers commonly track forecast accuracy improvement to measure performance of their demand forecasts. Whilst forecast accuracy is certainly a metric that should be tracked, retailers need to be wary of drawing incomplete conclusions if this is where their analysis begins and ends. Why? Biased forecasts can obscure significant gains of key performance measures such as out-of-stock rates or inventory turns… in other words, metrics that directly impact turnover and profitability.

Mean absolute percentage error (MAPE) or root mean square error (RMSE) are often used by retailers to judge the accuracy of forecasts. Though widely used, these measures can be flawed when evaluating the efficacy of point solution overrides, especially in the presence of biased forecasts.

In our work with retailers, Planalytics has typically found some level of positive bias in forecasts, often in the +5% to +20% range. Whether this is intentional to ensure high service levels or due to challenges from in intermittent demand, this bias has proven to be persistent across all types of retailers and demand forecasting applications.

At a macro forecasting level, this bias may not pose obvious problems. For example, when choosing the optimal time series forecasting methodology (e.g. AVS Graves vs. Exponential Smoothing), some combination of error and bias factors will allow a retailer to choose which method they believe will maximise supply chain outcomes. However, biased forecasts will mask the effectiveness of point solution overrides.

Consider the following hypothetical scenario in which a demand forecast has an average bias of 10%. Figure 1 illustrates how the override – although “perfect” in that the base forecast with the adjustment exactly match observed demand – will appear to have a negative effect if only measured by forecast accuracy.

This hypothetical forecast assumes a perfect unbiased demand forecast of 100 units, a perfect forecast override of 10% and a known +10% bias to the forecast. When calculating the error rates from this “perfect” forecast, we see that the biased forecast appears to have an error rate of 0% while the biased overridden forecast has a MAPE of 10%. This result is despite the fact that the override exactly captures the increased demand. It is easy to see how a retailer can come to incorrect conclusions about the improvements that are being provided by the override and the accuracy of biased baseline forecast.

To more accurately gauge the effectiveness of forecast overrides, retailers utilise a before and after A/B testing approach such as difference-in-differences.

The difference-in-differences approach compares performance of supply chain metrics like out-of-stock rate or turn rate before and after implementation of a new forecasting technique. The test can be set up either as a comparison between stores or between products. The economists who popularized the difference-in-differences methodology recently won the Nobel Prize in Economics, further validating the above measurement methodology (The Royal Swedish Academy of Sciences 2021).

An increasingly common demand forecast adjustment or override that retailers look to address is the weather’s impact on consumer purchasing. This is no surprise as the weather is constantly changing and it continually and significantly alters consumer demand. Projecting how much (% or unit volume) demand will increase or decrease due to the weather across products and stores via overrides to the base forecast is a proven way to limit both understocking and overstocking specific products and categories in different locations (that will experience different weather-influenced demand trends).

Using an example in which Planalytics’ Weather-Driven Demand (WDD) outputs are used as forecast overrides will show how real financial gains may get overlooked when forecast accuracy improvement is the only metric evaluated. In a comparison (Figure 2) of the baseline difference in out-of-stock rate pre- and post-WDD overrides over the course of several weeks one can see how this disconnect can occur. The difference between the baseline and post-implementation out-of-stock rates is the treatment effect from the methodology change.

Instead of looking at changes in MAPE which (as was shown earlier) can produce spurious conclusions, the difference-in-differences approach on out-of-stock (OSS) rate illustrates the exact effect WDD forecast overrides have on the relevant supply chain outcome. In the end, it is this measurement of benefit and its financial value that ultimately matters to a business.

The example above using Weather-Driven Demand forecasts generate a measurable 100-200 bps improvement in out-of-stock rates, 5-10% reductions in on-hand inventory, and 7-15% reductions in shrink for perishable items. In calculating bottom line financial return that results from improvements in these key business metrics, retailers typically capture EBITDA gains between 2-5 million for every 1 billion in revenue exactly captures the increased demand. It is easy to see how a retailer can come to incorrect conclusions about the improvements that are being provided by the override and the accuracy of biased baseline forecast.

David Frieberg, VP Marketing

Read the original article here.

view details »

Why the Same Temperature Can Feel Different Somewhere Else

FiveThirtyEight published an interesting article (“Why The Same Temperature Can Feel Different Somewhere Else”) discussing how reactions to the same weather conditions are all “relative” based on a person’s location and related cultural norms.

“In much of the United States, the high 80s in Fahrenheit is hot, but it’s not hot-hot. It could even be a day of sweet relief in the South, maybe time for a family picnic. But last month, across the United Kingdom, headlines warned of temperatures that could hit 31 degrees Celsius. When Americans found out that translated to 88 degrees Fahrenheit, they quickly concluded: “Europeans are weak.” And while temperatures in Europe kept going up, eventually hitting levels even Texans would find daunting and killing thousands of people across the continent, the question of how a temperature could mean serious danger in one place while being an average summer Saturday in another remained.

The temperature may be an objective number, but how we experience it is not. Culture influences the biology and psychology of thermal comfort, shaping what our bodies are used to dealing with and how our homes and businesses are set up to adapt.”

At Planalytics, where we help retailers and other consumer-based businesses quantify and forecast how much demand for a product or service will increase / decrease due to the weather, the points made in this article are certainly confirmed by the metrics we produce for clients daily.

The below example for a very basic commodity product – bottled water – illustrates the subjective nature of temperature on different populaces.  Weather-Driven Demand (WDD) analytics isolate and measure the impact that the weather alone has on demand. The graphic below shows the large variance in WDDs for different U.S. cities that all experienced high temperatures in the low 90s during a mid-July period. The WDDs indicated the increase / decrease this temperature level had on bottled water demand compared to each location’s normal for that time of year.


In addition to looking at different responses to identical absolute metrics such as a temperature or a heat index, the changes in weather measures for people in a particular location and a particular time of year has a significant impact on the activities people pursue and the products they purchase.  Whether it is a change from what’s normal or what the recent weather situation has been, consumer reactions are again localized in nature. The below example, shows how differently demand for three products will change with an 8 degree drop in temperature in Mid-April:


And while the FiveThirtyEight article highlights temperature, other weather components like rain or snow and the combination of those factors add further complexity in terms of behavior and consumer purchasing.  Planalytics applies advanced statistical methods and machine learning, mountains of sales data, years of unmatched retail demand expertise to the most omnipresent external influencer of purchasing — the weather — to help businesses understand, quantify, and proactively manage the never-ending impacts.  Learn more at

view details »

Week Ending August 13, 2022 / Retail Week 28 / August Week 2

Cooler in the East and Along the West Coast, Summer Heat in the Middle. Showers for Many. West-East Split in Canada. Showery West and East.


Weather-Driven Demand (WDD) Category Notables

Weather-Driven Demand LegendWDDs represent the estimated % change in demand of the product / category based purely on the year-over-year changes in weather. 


Weekly Summary


A Welcome Cool-Down. Cooler air provided welcome relief from the recent sizzling conditions in the central and eastern U.S. The West Coast had cool onshore flow and the interior West stayed hotter than LY. The Eastern North Central region was coolest since 2014, the Southeast since 2015. New Orleans was coolest since 2008, Las Vegas since 2010, Atlanta and Chicago since 2014. Contrarily, the Northwest region was hottest since 1971. Denver was hottest in the past 60+ years, Salt Lake City since 2013, and Oklahoma City & Philadelphia since 2016.

Summertime Showers. Showers and storms were scattered across the central and eastern states while monsoon rains persisted in the West. The Northwest was wettest since 2019. Nashville was wettest in the past 60+ years, Las Vegas since 1984, Memphis since 2002, Atlanta since 2005, and Houston since 2007. There were around 360 reports of severe weather but only 2 tornadoes.

A Shift in Canada. Temperatures slipped below LY for much of eastern Canada while the West (apart from the immediate coast) turned hotter. Montreal and Ottawa were coolest since 2014, Toronto and Winnipeg since 2019. Conversely, Calgary was hottest in the past 60+ years, Edmonton since 2014. Rain was most prevalent in the East and West. Ottawa was wettest in the past 60+ years, Montreal, Saint John, and Quebec City since 2015.

Last Year (week-ending 08/14/21), the U.S. was hottest in the past 60+ years and wettest since 2018. Canada was hottest in the past 60+ years and wettest since 2017.


Weekend Review

The weekend trended hotter than LY for much of the central U.S. but the West and East trended cooler. Rain and storms moved from the North Central region to the Ohio Valley. Showers and storms were also strung along the Gulf and Southeast Coasts and the interior West. Western and eastern Canadian markets were cooler than LY while the Prairies remained hotter. Showers and storms crossed the Prairies and light rain fell in the Maritimes.


Contact us to learn how predictive weather-driven demand analytics can help your organization.


view details »

Predictive Weather-Driven Demand Analytics Is the Secret Ingredient Enabling Retailers to Succeed in Uncertain Times


There are many external variables influencing businesses and consumers today.

Recently, we’ve heard a lot about volatile supply chains, delayed product shipments, fluctuating inventory levels, inflationary pressures, and rising prices. In addition, retailers continue to be understaffed. Each of these individual variables denote business disruptions, distractions, and challenges for businesses who are looking to meet customer demand and achieve service level targets.

These are just some of the macro-environmental ‘unknowns’ that are impacting retailers and consumers.

Yet one of the most volatile external variables that impacts consumer purchasing decisions and overall retail sales is also one of the least analyzed: the weather.

No other external variable can influence your sales as frequently, immediately, or meaningfully as the weather. Because weather is always changing, retailers are tasked with addressing constant shifts in shopping patterns, as well as demand for specific products and services. Fortunately, for retailers, the impact of the weather on consumer purchasing is NOT an unknown — it is known, and can be unlocked through predictive demand analytics. Understanding and measuring the impact of weather is the first step. Once measured, it can quickly be implemented at scale across the entire business to drive improved results across systems integration, people, and processes.

SGG Associates LogoIn today’s environment, successful retailers are consistently evaluating activities that can improve customer loyalty and enhance the shopping experience. Predictive weather driven demand analytics are an enabler to quickly realize these benefits at scale. Planalytics and SGG + Associates partner together to leverage the power of predictive demand analytics to help retailers and brands proactively manage weather volatility and improve profitability.

view details »

Unilever Greece Expands Use of Planalytics Predictive Weather-Driven Demand Analytics


BERWYN, PA, June 6, 2022: Planalytics is pleased to announce that Unilever Greece has upgraded its agreement to receive enhanced Weather-Driven Demand (WDD) analytics. The expanded service adds future demand impact projections to the historical analytics Unilever Greece has been receiving. Unilever Greece now has access to predictive WDD values provided at the weekly level for an expanded set of product categories and regions.  Planalytics will also provide Unilever Greece with additional reports and live, on-demand insights via its client portal.


view details »

You can mark this down. Retailers preserve margin with predictive demand analytics.

Content by Planalytics

As the global supply chain issues subside and retail buying departments revert to more normal purchasing practices, markdowns will reemerge as a profit-busting challenge to manage and mitigate.

For retailers, the misalignment of inventories with consumer demand often leads to lost sales due to out-of-stocks or higher markdowns when they have overestimated demand. Revenue, profit, and shopper experience all take a hit when an item is not available when a customer wants it. On the other hand, going heavy on inventory as a way to avoid lost sales comes with huge costs.

A survey conducted by Coresight Research and Celect estimated that non-grocery retailers in the U.S. absorb markdown costs of about $300 billion annually, or about 12% of overall sales.

Markdowns are a costly expense but also an area of opportunity for retailers. So how do retailers use analytics to proactively address markdowns and enhance profitability?

Using Predictive Demand Analytics to Optimize Inventories

The above mentioned research found that “misjudged inventory decisions—including overbuying, buying the wrong type of products and misallocating inventory—account for an estimated 53% of unplanned markdown costs for retailers”. Diving into root causes, the report identified that the largest factor leading to unplanned markdowns was “reduced demand due to external factors such as unseasonable weather, sudden changes in consumer behavior and competitors’ unplanned promotional activities”.

Traffic levels (store and online) and the sales of particular products in specific locations do vary significantly from day-to-day and week-to-week due to changes in the weather. In fact, no other external variable influences demand shifts as frequently, meaningfully, and directly as the weather.

Ignoring the influence of the weather can increase markdowns and negatively affect profitability. Planalytics’ experience working with retailers has shown that leveraging predictive weather-driven demand analytics, both ahead of the selling season and during the end of the season when markdowns come into play, can preserve margin in several ways.

Weather-driven demand metrics isolate and quantify – as a percentage or in units – the influence that the weather has on sales. These metrics can be calculated months ahead for planning and merchandise allocation purposes (e.g. -15% weather-driven demand for Sweaters in Boston in October compared to the prior year) and recalculated in-season to factor in forecasted weather conditions (e.g. +28% weather-driven demand for Barbeque Grills in the southeast this weekend).

3 Ways to Limit Markdowns with Weather-Driven Demand Metrics

Markdown optimization can be a big contributor to profit enhancement. Retailers can expect to improve their gross margins by anticipating demand better and ultimately reducing markdown costs. Here are three ways to apply the analytics and profit:

• Pre-season planning & allocation: When retailers plan the next season or year they must correct the weather bias embedded in past sales performance. This “deweatherization” process improves accuracy by accounting for when favorable weather conditions exaggerated sales or unfavorable conditions deflated sales. In situations where prior sales are inflated, the positive weather environment rarely materializes to the same degree again the next year. The result: retailers often end up with excess inventories that need to be marked down to clear stocks. So instead of unintentionally chasing weather-biased sales from the prior year, which are statistically unlikely to repeat, retailers can use weather-driven demand to improve plan accuracy and adjust inventories on a market-by-market basis. As a result, excess stocks are trimmed in markets that will not match strong comp sales levels, reducing eventual markdown costs. Moreover, retailers can increase inventory levels in markets that are likely to rebound strongly from weak, weather-dampened prior year sales. This results in fewer lost sales where consumer demand is elevated.

• In-season markdown decisions: The next opportunity to proactively manage markdowns presents itself in-season, after sales have peaked. By considering when, where, and how much the upcoming weather will impact consumer demand, a retailer can take advantage of favorable conditions to delay markdowns for a period of time or reduce the depth of markdowns (e.g. 30% off instead of 50% off). There is no reason to throw away margin if the weather will be boosting demand naturally. For example, a December or January snowstorm may produce strongly positive weather-driven demand projections for hats and gloves and snow blowers and more in affected markets. The retailer that adjusts the timing or degree of markdowns can capture more higher-margin sales on their snow categories while still drawing down inventories as the season winds down.

• Digital marketing to drive more full-priced sales: Targeting audiences that will be experiencing a positive weather backdrop increases message relevancy and conversions and is yet another way to leverage weather-driven demand metrics. Take a situation where a particular region has had soft sales, and stores are currently carrying excessive inventory that will eventually need to be marked down. If a favorable weather-driven demand environment is on the way, ramping up digital advertising in these markets would help drive sales that reduce stock levels at full price or before deeper markdowns become necessary. Businesses consistently capture larger than expected sales boosts when they coordinate marketing activities to capitalize on favorable, weather-assisted demand environments. More effective, targeted deployment of advertising spend is a proven way to increase sales overall, and in many cases, reduce the very inventories that could most use a helping hand before turning to the blunt and profit-eroding instrument of markdowns.

Retailers that use proven predictive demand analytics stand to reap a variety of financial benefits that come from aligning inventories with consumer purchasing trends. With visibility into how the key external factor of the weather will influence demand, retailers can smartly plan and manage inventories across stores and regions in a way that minimizes markdowns and preserves margin without sacrificing availability and sales. Learn more by visiting

Click here to read the full original article.

view details »

The Planning & Forecasting Fix that Fashion Retailers Need to Increase Profit Margin

Nothing repeatedly challenges retailers and erodes margins as the misalignment of inventories with consumer demand.  For clothing, footwear, and accessories retailers this misalignment often leads to lost sales due to out-of-stocks or higher markdowns when they have overestimated demand.

Obviously, a retailer’s revenue and profit take a hit when an item is not available when a customer wants it.  And today, with shoppers quickly pulling out their cell phones to seek alternatives, retailers are at greater risk than ever of losing the current sale – and potentially all future ones –while negatively impacting the customer experience and their brand.

On the other hand, heading off lost sales with heavier inventories comes with huge costs as well. A survey conducted by Coresight Research and Celect estimated that non-grocery retailers in the U.S. absorb markdown costs of about $300 billion annually, or about 12% of overall sales.

This research report found that “misjudged inventory decisions—including overbuying, buying the wrong type of products and misallocating inventory—account for an estimated  53% of unplanned markdown costs for retailers”. Diving deeper, the research identified that the largest factor leading to unplanned markdowns was “reduced demand due to external factors such as unseasonable weather, sudden changes in consumer behavior and competitors’ unplanned promotional activities”

Factor in External Demand Drivers to Optimize Inventories

Traffic levels (store and online) and the sales of particular products in specific locations vary significantly from day-to-day and week-to-week throughout the season and no other external variable influences these demand shifts as consistently, meaningfully, and directly as the weather.

Too many retailers ignore the impact of weather and this adds error to plans and demand forecasts. Planalytics predictive demand analytics give companies the visibility they need to proactively adjust allocation and replenishment decisions based on when, where, and how much changes in the weather will influence purchasing.

Businesses use Planalytics’ product-specific, localized demand adjustments within their SAP environments to improve accuracy for core retail activities.  As Planalytics’ preferred SAP systems integrator, Groupsoft provides seamless integration into SAP applications. This includes SAP’s Customer Activity Repository (CAR), where retailers can get a clear read on their performance by having a true understanding of the weather’s impact on sales. In addition, these weather-driven demand analytics can feed into SAP’s forecasting system (F&R / UDF) as well as Merchandise Planning, Allocation, Replenishment, Pricing, Promotions, the SAP Analytics Cloud, and more. Once integrated into SAP, these scalable and sustainable analytics drive measurable benefits across the business.

What We Deliver

As Planalytics’ exclusive SAP systems integrator, Groupsoft offers implementation and consulting services for operationalizing Weather-Driven Demand (WDD) analytics

Groupsoft specialises in Industry Solutions for Fashion, Retail and Wholesale Verticals.





view details »

Retail Sales Impacts & Outlook for Spring 2022

Join UBS and Planalytics for a look at key factors expected to influence retail sales performance during the spring season.

Topics include:

  • A discussion of the current business environment and the macroeconomic factors (inflation, interest rates, etc.) on consumer demand
  • High-level review of year-to-date opportunities and challenges confronting specific retail segments
  • An outlook on how weather-driven demand is expected to affect comp sales performance across regions/markets and key seasonal product

view details »

Weather Permitting: How to Use Weather Data in Retail Forecasting

“Retail is detail. And there is a lot of ‘detail’ for retailers to manage and factor into the business every day. When it comes to external variables, nothing is more consistently and directly impactful on demand than the weather.

This is because the weather influences consumer buying behavior everyday – and it never stops changing; no other external variable shifts demand trends as immediately, frequently, and meaningfully. Despite knowing all this, too many retailers ignore the impact of weather and this adds error to plans and demand forecasts.”

Read ToolsGroup’s entire post.

view details »