Filter and preprocess telemetry in the Application Insights SDK

You can write and configure plug-ins for the Application Insights SDK to customize how telemetry can be enriched and processed before it's sent to the Application Insights service.

  • Sampling reduces the volume of telemetry without affecting your statistics. It keeps together related data points so that you can navigate between them when you diagnose a problem. In the portal, the total counts are multiplied to compensate for the sampling.
  • Filtering with telemetry processors lets you filter out telemetry in the SDK before it's sent to the server. For example, you could reduce the volume of telemetry by excluding requests from robots. Filtering is a more basic approach to reducing traffic than sampling. It allows you more control over what's transmitted, but it affects your statistics. For example, you might filter out all successful requests.
  • Telemetry initializers add or modify properties to any telemetry sent from your app, which includes telemetry from the standard modules. For example, you could add calculated values or version numbers by which to filter the data in the portal.
  • The SDK API is used to send custom events and metrics.

Before you start:


This technique gives you direct control over what's included or excluded from the telemetry stream. Filtering can be used to drop telemetry items from being sent to Application Insights. You can use filtering in conjunction with sampling, or separately.

To filter telemetry, you write a telemetry processor and register it with TelemetryConfiguration. All telemetry goes through your processor. You can choose to drop it from the stream or give it to the next processor in the chain. Telemetry from the standard modules, such as the HTTP request collector and the dependency collector, and telemetry you tracked yourself is included. For example, you can filter out telemetry about requests from robots or successful dependency calls.


Filtering the telemetry sent from the SDK by using processors can skew the statistics that you see in the portal and make it difficult to follow related items.

Instead, consider using sampling.

Create a telemetry processor (C#)

  1. To create a filter, implement ITelemetryProcessor.

    Telemetry processors construct a chain of processing. When you instantiate a telemetry processor, you're given a reference to the next processor in the chain. When a telemetry data point is passed to the process method, it does its work and then calls (or doesn't call) the next telemetry processor in the chain.

    using Microsoft.ApplicationInsights.Channel;
    using Microsoft.ApplicationInsights.Extensibility;
    public class SuccessfulDependencyFilter : ITelemetryProcessor
        private ITelemetryProcessor Next { get; set; }
        // next will point to the next TelemetryProcessor in the chain.
        public SuccessfulDependencyFilter(ITelemetryProcessor next)
            this.Next = next;
        public void Process(ITelemetry item)
            // To filter out an item, return without calling the next processor.
            if (!OKtoSend(item)) { return; }
        // Example: replace with your own criteria.
        private bool OKtoSend (ITelemetry item)
            var dependency = item as DependencyTelemetry;
            if (dependency == null) return true;
            return dependency.Success != true;
  2. Add your processor.

ASP.NET apps

Insert this snippet in ApplicationInsights.config:

  <Add Type="WebApplication9.SuccessfulDependencyFilter, WebApplication9">
     <!-- Set public property -->

You can pass string values from the .config file by providing public named properties in your class.


Take care to match the type name and any property names in the .config file to the class and property names in the code. If the .config file references a nonexistent type or property, the SDK may silently fail to send any telemetry.

Alternatively, you can initialize the filter in code. In a suitable initialization class, for example, AppStart in Global.asax.cs, insert your processor into the chain:

var builder = TelemetryConfiguration.Active.DefaultTelemetrySink.TelemetryProcessorChainBuilder;
builder.Use((next) => new SuccessfulDependencyFilter(next));

// If you have more processors:
builder.Use((next) => new AnotherProcessor(next));


Telemetry clients created after this point will use your processors.

ASP.NET Core/Worker service apps


Adding a processor by using ApplicationInsights.config or TelemetryConfiguration.Active isn't valid for ASP.NET Core applications or if you're using the Microsoft.ApplicationInsights.WorkerService SDK.

For apps written by using ASP.NET Core or WorkerService, adding a new telemetry processor is done by using the AddApplicationInsightsTelemetryProcessor extension method on IServiceCollection, as shown. This method is called in the ConfigureServices method of your Startup.cs class.

    public void ConfigureServices(IServiceCollection services)
        // ...

        // If you have more processors:

Example filters

Synthetic requests

Filter out bots and web tests. Although Metrics Explorer gives you the option to filter out synthetic sources, this option reduces traffic and ingestion size by filtering them at the SDK itself.

public void Process(ITelemetry item)
  if (!string.IsNullOrEmpty(item.Context.Operation.SyntheticSource)) {return;}

  // Send everything else:

Failed authentication

Filter out requests with a "401" response.

public void Process(ITelemetry item)
    var request = item as RequestTelemetry;

    if (request != null &&
    request.ResponseCode.Equals("401", StringComparison.OrdinalIgnoreCase))
        // To filter out an item, return without calling the next processor.

    // Send everything else

Filter out fast remote dependency calls

If you want to diagnose only calls that are slow, filter out the fast ones.


This filtering will skew the statistics you see on the portal.

public void Process(ITelemetry item)
    var request = item as DependencyTelemetry;

    if (request != null && request.Duration.TotalMilliseconds < 100)

Diagnose dependency issues

This blog describes a project to diagnose dependency issues by automatically sending regular pings to dependencies.

JavaScript web applications

Filter by using ITelemetryInitializer

  1. Create a telemetry initializer callback function. The callback function takes ITelemetryItem as a parameter, which is the event that's being processed. Returning false from this callback results in the telemetry item to be filtered out.

    var filteringFunction = (envelope) => {
      if ( === 'tobefilteredout') {
          return false;
      return true;
  2. Add your telemetry initializer callback:


Add/modify properties: ITelemetryInitializer

Use telemetry initializers to enrich telemetry with additional information or to override telemetry properties set by the standard telemetry modules.

For example, Application Insights for a web package collects telemetry about HTTP requests. By default, it flags as failed any request with a response code >=400. But if you want to treat 400 as a success, you can provide a telemetry initializer that sets the success property.

If you provide a telemetry initializer, it's called whenever any of the Track*() methods are called. This includes Track() methods called by the standard telemetry modules. By convention, these modules don't set any property that was already set by an initializer. Telemetry initializers are called before calling telemetry processors. So any enrichments done by initializers are visible to processors.

Define your initializer


using System;
using Microsoft.ApplicationInsights.Channel;
using Microsoft.ApplicationInsights.DataContracts;
using Microsoft.ApplicationInsights.Extensibility;

namespace MvcWebRole.Telemetry
   * Custom TelemetryInitializer that overrides the default SDK
   * behavior of treating response codes >= 400 as failed requests
  public class MyTelemetryInitializer : ITelemetryInitializer
    public void Initialize(ITelemetry telemetry)
        var requestTelemetry = telemetry as RequestTelemetry;
        // Is this a TrackRequest() ?
        if (requestTelemetry == null) return;
        int code;
        bool parsed = Int32.TryParse(requestTelemetry.ResponseCode, out code);
        if (!parsed) return;
        if (code >= 400 && code < 500)
            // If we set the Success property, the SDK won't change it:
            requestTelemetry.Success = true;

            // Allow us to filter these requests in the portal:
            requestTelemetry.Properties["Overridden400s"] = "true";
        // else leave the SDK to set the Success property

ASP.NET apps: Load your initializer

In ApplicationInsights.config:

    <!-- Fully qualified type name, assembly name: -->
    <Add Type="MvcWebRole.Telemetry.MyTelemetryInitializer, MvcWebRole"/>

Alternatively, you can instantiate the initializer in code, for example, in Global.aspx.cs:

protected void Application_Start()
    // ...
    TelemetryConfiguration.Active.TelemetryInitializers.Add(new MyTelemetryInitializer());

See more of this sample.

ASP.NET Core/Worker service apps: Load your initializer


Adding an initializer by using ApplicationInsights.config or TelemetryConfiguration.Active isn't valid for ASP.NET Core applications or if you're using the Microsoft.ApplicationInsights.WorkerService SDK.

For apps written by using ASP.NET Core or WorkerService, adding a new telemetry initializer is done by adding it to the Dependency Injection container, as shown. This is done in the Startup.ConfigureServices method.

 using Microsoft.ApplicationInsights.Extensibility;
 using CustomInitializer.Telemetry;
 public void ConfigureServices(IServiceCollection services)
    services.AddSingleton<ITelemetryInitializer, MyTelemetryInitializer>();

JavaScript telemetry initializers


Insert a telemetry initializer using the snippet onInit callback:

<script type="text/javascript">
!function(T,l,y){<!-- Removed the Snippet code for brevity -->}(window,document,{
src: "",
crossOrigin: "anonymous",
onInit: function (sdk) {
  sdk.addTelemetryInitializer(function (envelope) { = 'This item passed through my telemetry initializer';
}, // Once the application insights instance has loaded and initialized this method will be called
cfg: { // Application Insights Configuration
    instrumentationKey: "YOUR_INSTRUMENTATION_KEY"

For a summary of the noncustom properties available on the telemetry item, see Application Insights Export Data Model.

You can add as many initializers as you like. They're called in the order that they're added.

OpenCensus Python telemetry processors

Telemetry processors in OpenCensus Python are simply callback functions called to process telemetry before they're exported. The callback function must accept an envelope data type as its parameter. To filter out telemetry from being exported, make sure the callback function returns False. You can see the schema for Azure Monitor data types in the envelopes on GitHub.


You can modify cloud_RoleName by changing the attribute in the tags field.

def callback_function(envelope):
    envelope.tags[''] = 'new_role_name'
# Example for log exporter
import logging

from import AzureLogHandler

logger = logging.getLogger(__name__)

# Callback function to append '_hello' to each log message telemetry
def callback_function(envelope): += '_hello'
    return True

handler = AzureLogHandler(connection_string='InstrumentationKey=<your-instrumentation_key-here>')
logger.warning('Hello, World!')
# Example for trace exporter
import requests

from import AzureExporter
from opencensus.trace import config_integration
from opencensus.trace.samplers import ProbabilitySampler
from opencensus.trace.tracer import Tracer


# Callback function to add os_type: linux to span properties
def callback_function(envelope):['os_type'] = 'linux'
    return True

exporter = AzureExporter(
tracer = Tracer(exporter=exporter, sampler=ProbabilitySampler(1.0))
with tracer.span(name='parent'):
response = requests.get(url='')
# Example for metrics exporter
import time

from import metrics_exporter
from opencensus.stats import aggregation as aggregation_module
from opencensus.stats import measure as measure_module
from opencensus.stats import stats as stats_module
from opencensus.stats import view as view_module
from opencensus.tags import tag_map as tag_map_module

stats = stats_module.stats
view_manager = stats.view_manager
stats_recorder = stats.stats_recorder

CARROTS_MEASURE = measure_module.MeasureInt("carrots",
                                            "number of carrots",
CARROTS_VIEW = view_module.View("carrots_view",
                                "number of carrots",

# Callback function to only export the metric if value is greater than 0
def callback_function(envelope):
    return[0].value > 0

def main():
    # Enable metrics
    # Set the interval in seconds in which you want to send metrics
    exporter = metrics_exporter.new_metrics_exporter(connection_string='InstrumentationKey=<your-instrumentation-key-here>')

    mmap = stats_recorder.new_measurement_map()
    tmap = tag_map_module.TagMap()

    mmap.measure_int_put(CARROTS_MEASURE, 1000)
    # Default export interval is every 15.0s
    # Your application should run for at least this amount
    # of time so the exporter will meet this interval
    # Sleep can fulfill this

    print("Done recording metrics")

if __name__ == "__main__":

You can add as many processors as you like. They're called in the order that they're added. If one processor throws an exception, it doesn't impact the following processors.

Example TelemetryInitializers

Add a custom property

The following sample initializer adds a custom property to every tracked telemetry.

public void Initialize(ITelemetry item)
  var itemProperties = item as ISupportProperties;
  if(itemProperties != null && !itemProperties.Properties.ContainsKey("customProp"))
        itemProperties.Properties["customProp"] = "customValue";

Add a cloud role name

The following sample initializer sets the cloud role name to every tracked telemetry.

public void Initialize(ITelemetry telemetry)
    if (string.IsNullOrEmpty(telemetry.Context.Cloud.RoleName))
        telemetry.Context.Cloud.RoleName = "MyCloudRoleName";

Add information from HttpContext

The following sample initializer reads data from HttpContext and appends it to a RequestTelemetry instance. The IHttpContextAccessor is automatically provided through constructor dependency injection.

public class HttpContextRequestTelemetryInitializer : ITelemetryInitializer
    private readonly IHttpContextAccessor httpContextAccessor;

    public HttpContextRequestTelemetryInitializer(IHttpContextAccessor httpContextAccessor)
        this.httpContextAccessor =
            httpContextAccessor ??
            throw new ArgumentNullException(nameof(httpContextAccessor));

    public void Initialize(ITelemetry telemetry)
        var requestTelemetry = telemetry as RequestTelemetry;
        if (requestTelemetry == null) return;

        var claims = this.httpContextAccessor.HttpContext.User.Claims;
        Claim oidClaim = claims.FirstOrDefault(claim => claim.Type == "oid");
        requestTelemetry.Properties.Add("UserOid", oidClaim?.Value);

ITelemetryProcessor and ITelemetryInitializer

What's the difference between telemetry processors and telemetry initializers?

  • There are some overlaps in what you can do with them. Both can be used to add or modify properties of telemetry, although we recommend that you use initializers for that purpose.
  • Telemetry initializers always run before telemetry processors.
  • Telemetry initializers may be called more than once. By convention, they don't set any property that was already set.
  • Telemetry processors allow you to completely replace or discard a telemetry item.
  • All registered telemetry initializers are guaranteed to be called for every telemetry item. For telemetry processors, SDK guarantees calling the first telemetry processor. Whether the rest of the processors are called or not is decided by the preceding telemetry processors.
  • Use telemetry initializers to enrich telemetry with additional properties or override an existing one. Use a telemetry processor to filter out telemetry.

Troubleshoot ApplicationInsights.config

  • Confirm that the fully qualified type name and assembly name are correct.
  • Confirm that the applicationinsights.config file is in your output directory and contains any recent changes.

Reference docs

SDK code

Next steps