The traffic estimator looks at the average. 12-month search volume, while Spate can infer the exact month-by-month search volume by ingesting Google Trends data. This permits you to track in "real-time" the brand health and the impact of consumer demand, product launch, and marketing initiatives...
For each brand that we are tracking, we scrape the brand's website to generate a list of Keywords → then, an algorithm determines if the Keywords are branded Keywords (e.g. Glossier) or product Keywords (e.g. Boy Brow). Our team manually checks all classified keywords to exclude irrelevant ones (e.g., Matrix the Movie vs. Matrix Shampoo).
Every month, Spate scrapes the internet for new keywords relevant to the beauty category to ensure we capture the latest trends or product launches. A new product launch will be added to an existing Brand.
We aim to achieve 100% accuracy, but there can sometimes be little glitches. We are super transparent and share our keyword list with our users as we want to get feedback (e.g., in the dashboard, users can see the keyword lists and flag miss-categorization). We take feedback super seriously, and our team will process any input + re-pull all the data as quickly as we can (currently est. 1 week turnaround)
Spate uses state-of-the-art prediction algorithms to offer predictions in 12 months. We run the model based on the millions of historical data points we collect and have an accuracy of 72% for the next 12 months.
All predictions are not equal; for each prediction that we make, we will provide a confidence score; some projections will have a very high level of confidence (90% confidence), while certain predictions are labeled as uncertain (50% confidence → that’s a flip of a coin). We aim to be transparent and will tell you when we aren’t sure about the future, as our goal is to provide a quantitative tool so you can make better decisions.
What about brands?
We have applied the same model to brands - but I would take it with a pinch of salt. Brands are unpredictable, as future searches can easily vary due to a product launch, a marketing campaign, or another event. In general, I would take this as directional data.
The sharp decline at the end of the prediction is due to a statistical effect that tends to "exaggerate" the volatility with time. As we look more into the future, the model becomes more uncertain. I wouldn't focus on the month-to-month data point but rather the overall 12-month number.