Does the data change?
Yes! At Headset, we use cutting edge real-time technology and point-of-sale integrations to allow us to access sales data as the sales are made. We combine these real-time integrations with equally real-time machine learning algorithms which allows us to forecast data every day in real-time. Unlike other providers of data, we do not make our customers wait weeks or months for updates. This allows our customers to get the most up-to-date information so they can be confident they are making decisions with the latest information. At Headset, we forecast SKU level market share daily based on a stratified, bootstrapped sample of retailer partners' data. These market share forecasts are then grossed up to state total sales based on sales data received from each state or provincial government.
What data changes?
All of it! Every night we re-forecast the data with the new information we have gathered. This means changes can happen to historical data daily. Adjustments of data are generally small (within a few percentage points at the brand category level). Larger adjustments can happen when a state or provincial government restates its sales totals. This happens monthly in most states and quarterly in California. At Headset we anticipate these adjustments and forecast accordingly; however, occasionally these adjustments can be outside of what we anticipated. Historically, these adjustments have been +/-4% of total sales. These adjustments are "peanut butter spread" across all products within the month.
Why does the data change?
The numbers can change for a lot of reasons. Knowing a bit more about our process helps to understand why. At a high level, our process uses a sample of store data to project market share at the product/day level. So we get a market share forecast for every product on every day. From there we gross up those market shares to state totals. The first reason the forecasts can change is the receipt of new, or adjusted, state sales totals. Until we receive state data we forecast state totals (this lag ranges from 2 weeks to 180 days depending on the state/province). When the state publishes totals we adjusted our forecasts. The second reason our forecasts change is because we adjust our market share forecast. This happens most commonly when combine brands or products as well as adjust forecasts for excluding things like sample products. Finally, our market share forecasts can change when we add new products to the data. Our overall forecasts can fluctuate, though usually not by large amounts unless we have an atypical event, for the most recent 6 months. Generally, after 6 months have passed we will see relatively little changes in the data.
What are the best practices for working with real-time forecasts?
It is best to access the Insights tool each time for your analysis. If you choose to download the exportable data and do your own analysis we advise pulling a full-refresh of all historical data each day. Avoid using incremental files.
Why doesn't the downloaded data match Insights?
The most common reason is that you are accessing a partial file. Some of our data exports (ex. Package Size Download) contain only a portion of sales. Due to the extensive volume of products available in the market, our more granular data downloads limit the data that can be downloaded so that the files are still easy to use. Please read the details at the top of each data download section to understand if you are working with a full or partial file. For example, in the Brand download file, we expose the historic sales for the Top 50 brands in each category in each state (removing brands with less than 500 units sold in the last month). All other brands are then labeled as "Other Brand". This file will change each month; for example, if a brand grows from a rank of 51 to 49 they will move out of "Other Brands" and will be shown under their brand name. The reverse will happen if a brand's rank declines below 50, they will move from showing under their own brand name to being in the "Other Brands" line item. The designation of a "Top 50 Brand" occurs at the end of each calendar month and as such the Brands included in the historical file change each month.