Are all real time tools created equal? Not quite.
In our previous two posts in this series, we discussed the problem of overlooked metrics in both ecommerce and fintech. With so many layers of business of technology, traditional BI analytics tools often encounter unpleasant surprises, discovering issues and opportunities long after the financial or reputational damage is already done. We focused on the multitude of specific metrics which arise when a generic metric like daily revenue is broken down by product, geo region, browser, etc. – the combinatorial explosion in the number of these very granular metrics highlights how dimensionality undermines those BI tools effectiveness.
While monitoring highly dimensional data in real time is a challenge in the adtech vertical, adtech companies frequently come up against another challenge: detecting variability due to seasonality in their metrics and having to account for this with their analytics tools, which are a far cry from the real time solutions their vendors claim they are.
Seasonality meets real time bidding: beyond the reach of traditional BI tools
Some, but not all, of a company’s metrics will exhibit some aspects of seasonality. Seasonality refers to the presence of cyclical patterns in the time series data. The period of the cycle can span from hours to a full year or more. The main sources of seasonality could be climate, institutions, social habits and practices, or calendar. Seasonal patterns are changes we expect; they are part of the normal behavior of a given metric and thus must be included in the model of that metric. For some metrics, however, there are no seasonal patterns. And sometimes, multiple seasonal patterns are present in a time series. Since this can be mis-identified as outliers that might deserve attention, seasonal variability must be identified, filtered out and ignored.
In a complex and extremely time sensitive vertical like Adtech – where the bidding, buying and selling of online advertising occurs in millisecond timescales, they also need to consider how seasonality plays in each metric. The problem is that traditional real time BI tools can’t account for seasonal trends.
The immense volume of data generated from transactions over online ad exchanges, together with the speed at which those transactions take place, are exacerbated by the challenges of data dimensionality and seasonality. If something slips through the cracks while trying to monitor all this, it could undermine the reputation of the ad exchange, wreaking havoc for the ad buyers and sellers who use the exchange. That’s why it’s so important to be able to quickly spot anomalies in KPIs for metrics like click rates, bid durations, cost per click, and page views served and correlating across all relevant metrics, providing visibility to all other related anomalies.
Taking a closer look at the numbers
Intelligently monitoring huge amounts of data could benefit from real-time anomaly detection. Eyeview, a video advertising company monitoring for approximately 200,000 metrics, had to constantly update its static thresholds as traffic increased. Seasonality complicated their reporting, for instance, online ad views would start climbing in the morning and peak around lunchtime, but drop during rush hour (since many of those eyes are focused on the road, as they should be) and peak again during prime time (the late evening) before dropping and staying low through the night.
Everything from fraud to ad blocking software to server problems can disrupt the reliability of the whole ad tech environment. Adtech companies that can take charge of their data true real time analytics can empower the business to react at speeds that the market moves.
Delivering “real” real time analytics for adtech
Adtech faces challenges around large volumes of high speed data, the need for real time processing of that data, and having to return the relevant results before the user leaves the page. Traditional BI tools cannot process this in a time effective manner to address these challenges, especially when considering the impact of seasonality.
Unlike other BI solutions which cannot account for seasonal trends, by learning the normal behavior of an adtech company’s data and determining seasonality for each metric, Anodot highlights issues that analysts would not have known about otherwise. With real-time analysis, companies can know if a specific DSP stops responding to bids, getting alerts and correlations to determine if an issue needs to be fixed.
In the image below for a campaign, focusing on data for number of bid requests, Anodot highlights an anomaly (in orange) accounting for seasonality:
The ease of use of Anodot’s business incident detection capability, together with its automatic correlations makes it straightforward for account managers to quickly see the problems that need to be addressed with their clients immediately. When Anodot’s AI-powered analytics detects a fall in revenue, it correlates with other relevant metrics; for example, blocks could be up, timeouts could be up, bid responses and pricing could be down. With these correlated metrics, adtech companies can quickly pinpoint where to investigate.
Only AI-powered analytics is up to the task of detecting business incidents real time, from server hiccups to ad affiliate changes, important signals in often overlooked metrics that can save revenue by the millisecond.