Evaluating a new, unknown technology is a complicated task. Although you can articulate the goals you’re trying to achieve, you’re probably faced with multiple solutions that approach the problem in different ways and highlight varying features. To cut through the clutter, you need to figure out what questions to ask in order to evaluate which technology has the optimal capabilities to get the job done in your unique setting.
That’s where a Request for Proposal (RFP) document comes in.
What is an RFP Template?
An RFP template is essentially a framework of questions that covers all the key bases for the type of solution under assessment. A good RFP covers the technology from all angles, so that you know that you haven’t overlooked something majorly important in the procurement stage. In addition, RFPs enable you to compare solutions by evaluating them side by side, according to the same parameters.
If you don’t know what to include in your RFP template for an anomaly detection solution, no worries. You are more than welcome to download and use our free anomaly detection RFP template that will guide you through the process of solution evaluation.
But while the RFP dives into all the nuts and bolts of anomaly detection technologies, here’s a general overview of the nine key capabilities to evaluate:
1. Scope of Data Coverage
The robustness of an anomaly detection solution is closely tied to the scope of data it can monitor. The types of data the solution can ingest, it’s support of multiple data sources, and the data integration and consolidation processes highly determine the amount of coverage the solution can offer. For example, you’ll need to know whether the solution can process semi-structured data, and whether it can monitor infrastructure, application, revenue, customer experience and 3rd party data. Also relevant is the mode of collection and any resource limitations placed on it.
2. Level of Monitoring Automation
While most anomaly detection solutions apply some kind of ML-based automations, these widely differ in their scope and sophistication. Across the solution’s architecture AI can be used for auto-learning of optimal models, metric behavior, baselines, multiple seasonalities, and anomaly detection and scoring. The extent of automation will determine both the sophistication of the solution (it’s ability to detect anomalies), and the amount of manual work which will be required on your part to configure and vet the solution.
3. Scalable Metric Monitoring
Here, too, the efficacy of the technology is highly influenced by its ability to scale to hundreds of thousands of metrics for both time series and event streams. A metric cap, for example, should alert you to the monitoring scope offered by the tool. In this context you should also consider whether the tool offers real time detection and alerting, and whether it can cope with composite metrics and accommodate multiple dimensions — and under what limitation, if any.
4. Alerting & Investigation
In the final analysis, you require that the solution will detect any and all significant anomalies on the one hand, while reducing false positives and alert storms on the other. That sweet spot is achieved through the system’s alert reduction mechanisms, including the ability to score alerts according to severity and urgency, consolidate related alerts to reduce alert storms, tailor alerts by different alert reception groups, and offer multiple sources incident-based investigation capabilities.
5. Root Cause Analysis
Even if the anomaly detection achieves optimal time to anomaly detection, your end goal is anomaly resolution. That’s why root cause analysis should be top of mind. The solution’s ability not only to detect an incident but to also report about related anomalies and provide resolution insights is key. This is determined by the technology’s ability to correlate metrics and events to the reported incident be investigated locally within the product, and providing a granular report on the anomaly’s root cause.
6. Flexibility & Scalability
Can the solution be deployed according to your specific needs, i.e. both on-premise or SaaS/hybrid? Are off-premise/cloud connections required? The answers to these questions might determine if the solution is right for you currently, and also as your scope increases in terms of data volumes, data traffic, number of users and landscape. You should also consider here approximately how many weeks or months are required for implementation, and what is the solution’s average Time to Value rate, or how soon can you expect to see ROI.
Usability issues influence adoption and continuous use of the solution. According to your needs, you will need to understand whether the software is self-service, does it require a data science background, whether users can collaborate on the platform, can business users run ad-hoc queries or setup alerts without the help of IT professional, and to what extent are alerts and reports interactive? Also of interest are the solution’s visualization and reports capabilities, and the extent to which these can be customized.
8. Total Cost of Ownership
This is pretty self-explanatory. Questions regarding TCO will gather the information you need to figure out how much the solution will really cost you in long-term staffing costs, data storage, and other associated costs. They will also shed light on the general criteria for pricing, for example, whether it’s based on data sources/streams, number of metrics, number of instances, etc. This will also enable you to project future costs as your company and its monitoring needs scale.
9. Support & Service
Onboarding and ongoing support for a new technology can sometimes make or break its implementation, adoption and ultimate success. That’s why when you team up with a new vendor you need to understand how they handle new accounts, and to what extent and quality they offer post-sales support, in-depth training and resources. It’s also important to know if new customers are assigned a dedicated Customer Success Manager or equivalent, responsible for establishing their objectives and ensuring you’re consistently getting the most out of the solution..
While approaches to anomaly detection vary between vendors, what matters most is how each solution’s capabilities stack up to meet your objectives. While this is a lot to take in, the RFP breaks down the concepts, features and capabilities covered in this post into bite-size chunks that will enable you to evaluate competing technologies head to head. So download our RFP template, and don’t hesitate to reach out if you need further guidance. We’re here to help.