Yogi Berra is often quoted as saying, “The future ain’t what it used to be.” Steve Jobs echoed the same sentiment as a preface to his optimistic extrapolations about the future of technology. In today’s world, this couldn’t be closer to the truth. Last Thursday, The Parker Avery Group hosted our 4th and final session in our Reconstructing Retail webinar series, with a focus on Predicting the Future and how very different it will be for retailers and consumer brands—not only the outcome, but the elements involved in creating these critical predictions. This blog is a recap of key messages, as well as questions raised from the webinar attendees.  If you weren’t able to join the live sessions over the past few weeks, we invite you to visit our website to review the replays, download the presentations, and read the blog recaps.

Review Reconstructing Retail Webinar Recaps & Replays

Most of us have adapted to major changes by learning from the past, but the last few months have put us in an interesting position. Retail as we know it has never seen anything like this, so learning from the past is not likely an option. The future is not certain, and many retailers will be learning as they go. In our Predicting the Future webinar, we discussed three phases with tools and actions that retailers can take as they reopen in order to prepare and forecast for the future.

  • Learn and Assess
  • Recover and React
  • Evolve and Adapt


Learn and Assess
Let’s talk sales forecasts. One of the first things you are going to notice coming out of the COVID-19 crisis is that your sales patterns look significantly different than they have in previous years. This is important because your forecasting systems rely heavily on historical sales. Especially true with traditional time-series based models, many of the forecasting systems that are built into the execution tools that retailers and consumer goods companies rely on are going to need varying levels of intervention.

Allocation and replenishment solutions and methods will likely be hit hard, and you will need to be as efficient as possible. Talk with your teams. Make sure they know when to switch methods and what other levers they need to adjust. Ensure they know that it’s OK to make changes and understand who to partner with if they have questions. This is not a time to do things because “That’s how we did it before COVID.” Explore your inventory flow options like pack and hold and do what’s right for the current situation, but with an eye on future business decisions and strategies.

Also, keep an eye on your teams and be careful they do not get into analysis paralysis. Your teams will want to be as accurate as possible knowing inventory investments are high with new shipments coming with no place to go. There will be stress. It will be easy to get stuck in multiple forecast scenarios trying to answer questions and be as precise as possible.

Recover and React
As we move out of the initial shock of reopening, we will move into the recovery phase—granted this will vary across brands. Retailers and consumer brands rely on forecasting systems for support in many functional areas across the enterprise—from planning and pricing to fulfillment and operations. We recommend starting with the easiest and most intuitive mitigation strategies first. Impute demand lost during the COVID-19 crisis using simple proxies such as LY/LLY and look at the reasonableness of the resulting forecast. Then, move on to more complex solutions as time and resources permit.

Prolonged store closures could have a significant effect on sales mix as well. Many retailers will move to more to allocation and rely less on replenishment as they will likely have a higher mix of fashion product in their inventory. Online sales could be a good indicator of this change especially with the seasonality by region in a great deal of flux. Who every thought it would snow in May? If you are using more advanced machine learning approaches that learn from different channels, then you should be able to recover faster.

Also, test where you can. Be flexible and allow for the option to roll out goods slowly by region where possible. Some areas will likely recover more quickly than others, and you will need to maximize on those. Online will give you a good view, but it will not show you what customers are willing to set foot in brick-and-mortar.

In any case, you should have a set of easy metrics that you follow as your stores open. A simple rolling sales mix will give you an easy way to validate your analytical output. Again, easy and intuitive is where you start and then prove to yourself that more complexity is required.

Advanced analytical systems can bring significant value to help you through this crisis, but they can also provide you with very spurious results. Remember, the analytical systems have not seen anything like this before either. The best systems will use a plethora of input and will adjust quickly with minimal hiccups, but if that’s not an option, have a mitigation plan for manual forecasting and communicate it to all stakeholders. This is not a time to be siloed.

Evolve and Adapt
Lastly, we don’t believe many retailers can, should, or will go back to pre-pandemic behaviors. As we emphasized in last week’s blog, “Fortune Favors the Bold,” those that leverage learnings and newly tapped skills from COVID-19 and continue to evolve and adapt will come out stronger. From a forecasting and analytic perspective, disruptions in the business make it more important to utilize additional streams of input. Decision tree based and deep learning approaches can take in a variety of inputs and do not simply rely on continuous time series data. These algorithms make it easier to utilize inventory, plan, hierarchy, product, and location attributes such as COVID-19 case by trade area.

Looking forward at supply chain visibility and data accessibility to make faster decisions will also play a critical role. Those companies that have to piece together information will have a tougher time moving forward and keeping up. The future will be about agility and being able to pivot as different scenarios arise.

When it comes to current tools and processes, come up with a plan to make what you have smarter. Then prioritize changes that will need to be made in the short term, mid-term and longer term. In some cases, you can adapt and change what you have now to limit spend. Also, remember it’s not all about technology. Just evaluating how you’ve always done it and making changes to your existing business processes can go a long way. This approach also sets you up for a successful technology change in the future if that is the route you choose. Ensure you are keeping up where you can and evolving your processes to stay engaged and relevant.

Wrapping up
Overall, keep in mind these five key takeaways and hopefully your future will look much brighter.

  • Understand your end game – easily consumable metrics for validation are key
  • Remember your benchmarks and don’t overcomplicate – LY and LLY compares can still help
  • Know your limits – many execution systems in place rely solely on traditional models for forecasting – have a mitigation plan
  • Don’t throw the baby out with the bath water – make your existing systems and processes smarter
  • Limit manual intervention – enable your teams to move quickly


Webinar Q&A
Q1.  Should we do an updated top down plan or aggregate the location plans to determine the company impact?
A.  Having a store group plan vs. a location plan in conjunction with a top down view is likely the best bet if you have this capability. The next few months will be fast moving with many changes. Doing detailed plans at store level will likely not be useful for a long period of time so it will be about getting directionally ‘OK’ and being able to have the flexibility to change.

Q2.  Video analytics / RFID – How do we leverage AI – video for contactless shopping / inventory / self-checkout?
A.  RFID is a great input for helping you understand your inventory position. Video analytics is another input that can utilized by the advanced models that we mentioned above. The availability of this type of data will continue to drive forecasting to more advanced methods that can best utilize the plethora of data that is becoming available in retail.

Q3.  I’ve read you can implement your demand planning system in as little as 6-weeks. Can you briefly discuss that if you have time.
A.  We have built an AI platform that produces an accurate and stable demand signal very quickly. We can integrate this signal into your execution systems to minimize the effect on your existing business processes.  We are happy to set up a time to discuss this in your own situation if you would like to contact Parker Avery at contact@parkeravery.com.

Q4.  This callout that backlogged inventory needs quick attention even before replenishment seems critical. How do you believe companies should be looking at SP21 Fashion? Is there a chance that many assortments may look ‘too safe’ a year from now?
A.  At this point for spring fashion, likely there are two choices. Retailers who can pack and hold will need to evaluate the inventory and the investment to hang on to it for a year. Others will need to markdown and move it while fully understanding the profitability model for the remainder of the year. Many wholesalers have already decided to pack and hold where they can so one way or another this year’s spring will be next year’s spring (whether we like it or not). So, we’re not sure if there is a model that would be termed ‘too safe,’ but it will come down to balance and ease of execution in the long run.

Q5.  Unemployment continues to increase – how can I get a good estimate of future demands?
A. Unemployment numbers can be tracked as location attributes and utilized for forecasting by machine learning algorithms such as decision tree-based models.

Q6.  In the short term – how many months should I forecast reduced sales for? E.g. the next 6 months?
A. This will depend on multiple factors such as when your stores are going to reopen and what portion of your sales come from the brick-and-mortar channel. There are of course the macro economic impacts which are even harder to predict. Likely thru the end of 2020 there will continue to be significant impact of some kind and sales reduction. You will need to continue to monitor relative time periods and be in a position to react as the customers decide. Just be ready for things to also ‘right size.’ There will continue to be strong online sales, but they may subside slightly as brick-and-mortar locations open and footsteps return. Placement of inventory into the correct channels will need to be carefully monitored.

If you have any questions about your own analytics systems and how to predict tomorrow and the months ahead for your company, please don’t hesitate to contact us.

Amanda & Sam

Cover image by Gerd Altmann from Pixabay