Anomaly Detector

Anomaly Detector

Understanding the positive and negative trends in business
Anomaly detection is a technique used to identify unusual or out-of-the-ordinary behavior that does not conform with the normal trend of a business. For example, a sudden dip in purchases, lead conversions, or an increase in students dropping out of courses can indicate a negative deviation from the regular pattern. 

However, an anomaly need not always indicate a negative trend. If there is a surge in the purchase frequency or website visits, it indicates a positive trend. You can observe these anomalies, draw a conclusion about what they mean, and take the necessary follow-up actions.
On the whole, an anomaly detection offers meaningful insights about the business's trends which can play a critical role in influencing key decisions. Decision makers can change existing processes or assimilate better ideas and plans by observing what brings positive changes to their overall business health. 

Availability
Permission Required
Users with administrative profile permission can access this feature.
Enterprise edition - 10 anomaly detectors
Ultimate edition - 25 anomaly detectors.

Anomaly detection in Zoho CRM 
Real business scenarios are often complex and full of uncertainties, therefore, chances of deviations from an expected pattern may be common. It can be difficult to pinpoint the deviations in huge volumes of data manually. Zoho CRM's anomaly detector can help you identify deviations by populating the real-time data sets into an expected pattern and flagging data points that lie outside this pattern. 

Zia uses contextual or conditional anomaly detection to detect irregular behavior in the usual trend, wherein a data instance will be marked as anomalous only in a specific context (but not otherwise). That is, she will compute the expected trend by monitoring the current and past sales trends and whenever any data unit shows deviation from this expected trend, it will be captured as an anomaly.  

Here are some anomalies that you can detect in Zoho CRM: 

Make changes to customer engagement activities based on frequency of deal closure:  The sales team can detect a dip in the deal closure rate, by analyzing the irregularities in the number of deals closed during a specific period. They can compare this with the lead conversion pattern, by observing the number of contacts that were created during the period. A decrease in the number of contacts can be addressed by changing the lead generation processes and enhancing customer engagement activities. 
In the screens below, we have calculated the rate of deal closure as a single metric that shows 3 anomalous data points. 

With the related metric, we get the ratio between number of deals closed and number of contacts created. The ratio shows only two data points as anomalous, the other point is not considered an anomaly because the number of deals closed is proportional to number of contacts created.  

Restock inventory as the product demand increases: E-commerce companies can proactively restock their inventory by monitoring purchase frequency patterns. A spike in the number of products that are being sold compared to the number of products in inventory can help indicate when the inventory is running low on the product and prompt those responsible for stock levels to reorder accordingly. 

Quickly address customer complaints:  Real-time anomaly detection can help companies resolve customer complaints quicker. An unusually high number of support cases will be picked up by the anomaly detector. The support teams can quickly investigate what is causing the issues and avoid escalations due to delays and related impacts on customer experience. 

Increase traction to the website:  The anomaly detector can indicate if your web page is receiving fewer visits. You can take measures to increase the traffic to your page by promoting it on social media, advertising it on other pages, guest blogging, or using internal cross linking. 

Setting up the anomaly detector
In Zoho CRM, you can create an anomaly detector for both standard and custom modules. You may include the following additional parameters while setting up the anomaly detector:
  1. Related Metric: You can compare the anomalous metric with a related metric to get a ratio of the two. Drawing comparisons can be helpful if the behavior of one metric can be correlated with the other, such as deal closure vs. number of contacts created, enrolment for a course vs. number of opt-outs, number of cases created vs. number of solutions provided. Remember that, you cannot select average of any metric.   
  2. Objective: You can set whether an increase in value is considered positive or negative. For example, an increase in the number of deals that were closed in a positive anomaly, but an increase in the number of support cases is a negative anomaly. 
  3. Grouping: If you prefer the results to be grouped under a particular category, you can choose from the available dimensions: time, currency, date, or any pick-list field. For example, you can group leads by their source or deals by their stage. If you want to compare two metrics you will not be able to group the results under a single category.

Calculation of anomalous behavior
  1. The anomaly will be calculated based on the organization's timezone that is set in Zoho CRM. For example, if the business hour is set to the US timezone, then a user at India will view the anomaly at the org's time zone only. This is because Zia will make the calculations based on the organizations time zone and not individual user's working hours.

  2. The total number of records on the day of anomaly calculation will remain intact even if a change is made later on. For example, if the deal closure pattern shows an anomaly on 21st December with total number of deals as 12; then the record count will display 12 even if one of records is moved to a different stage for any reason. This happens because Zia computes the anomaly based on the observation at that point of time and doesn't replace it with a real-time change. 
To add an anomaly detector
  1. Go to the  Dashboards tab and click  Add Component .
  2. Click  Anomaly Detector from the list of components.

  3. In the  Add Anomaly Detector page, enter the  Component Name.
  4. In  Anomaly For , do the following:
    1. Select a module for which you want to determine the anomaly.
    2. Select the measure , which must be a number field (sum, count, average, maximum etc.)
    3. Choose a date field (last activity time, created time, modified time etc.) based on which the records will be considered.
    4. Click  Criteria Filter to specify whether selected records should be considered for detection. 
  5. Check  Compare with another metric and select the module measure , and  date field as in the previous step.
  6. In  Anomaly Duration , select the  Duration (this week, today, last 360 days etc.) and the  Frequency (Daily, Weekly, or Monthly) from the drop-down lists.
  7. In  Objective , select one of the options:
    1. Consider increase in value as positive
    2. Consider increase in value as negative
       

  8. Check Group By and select a grouping from the drop-down list.
  9. Click  Save. 
 
    • Related Articles

    • Analytical Components

      Every business needs to process and organize data on a regular basis. Also, your daily activities can become tedious when you have to get regular updates on trends, targets, performance indicators, pipeline stages, meetings, follow-up calls and ...
    • Zoho CRM Plus - Getting started guide for Sales teams

      Every business goes through a sales cycle, from acquiring, nurturing, and following up on prospects to closing business deals with them. There may be many activities involved in each phase, such as scheduling appointments with your customers, ...
    • Zia - An Overview

      Zia is an AI-driven support assistant who makes your customer support team more effective through data mining and machine learning. She provides support agents with analyses, customer forecasting, alerts and suggestions through a pre-configured ...
    • Trend Analysis

      Let us imagine that you are away from office on a business trip for a week and there's been a drop in the number of deals closed and revenue. Unless you check for the status manually by yourself, chances are you miss to take action at the right ...
    • Zia - An Overview

      In today's world, businesses deal with mammoth sizes of data that come from multiple sources. It could be a simple phone enquiry that you just had, a new follower on your Facebook page or a bounced email in your marketing campaign. Making sense of ...