If you're looking for a tutorial article on how to navigate the event mapping page for a specific integration, head over here instead!
If you are auditing, refreshing, or just maintaining your MadKudu instance, the event mapping is an important piece. Here is how to edit it.
Pre-requisites
- You know what an Event mapping is
This article covers
- Your data pipeline
- Prepare your data
- How to add new events to the event mapping?
- How to edit events already mapped in the event mapping?
Your data pipeline: One key thing to keep in mind
The event mapping is the backbone of your behavioral model. Therefore, if adding events does not impact your model, editing existing events (in your integration or in the event mapping) can have an impact on your scores depending on what you are editing.
- If you change events at the entrance (in your integrations) or in the middle (in the event mapping) of this data pipeline there would be implications downstream of this pipeline, which means your scores can be impacted.
- If you add events in your integration, without including them in the event mapping, the model will either not include them in scoring or bucket them into "Other activity"
After you make any modifications, we recommend you to:
- Go into the Studio
- Re-build your training dataset to include the new events of your updated event mapping
- Make sure the weights of each event make sense
- Make sure the thresholds make sense
Prepare your data
If you connected several integrations containing behavioral data, you will have several event mappings to configure: 1 per data source. So before diving into MadKudu and configuring your event mappings, you will need to plan things ahead.
What events to include?
Some activities performed by your users are key to include in your Likelihood to Buy model. You need to make sure these important events are mapped to be taken into account by the model.
Events showing strong engagement are:
- Account creation
- Account activation
- Free trial or Demo request
- Contact Us or Booked a meeting
- Signed into App
- Attended a webinar
- Downloaded content from website
- Opened and/or clicked an email
Other classic events to make sure are mapped are:
- Chat activities
- Filled out forms
- Viewed certain website pages (pricing, blog, case studies, partners)
- Registered for a webinar
- Registered for and/or Attended an event
- Subscribed to a newsletter
- Unsubscribed from a newsletter
What events to exclude?
Not every single data point stored in your integrations will be valuable for your Likelihood to Buy model. You need to make sure passive activities not performed by your leads are excluded from your event mapping because they don't show any real intent from your leads to buy your product. For example, a lead was sent an invite to an event.
We advise you to specifically map these events, and select the "Non User Activity" Activity type for them in the event mapping dropdown. This will ensure these events are never considered in the Models, in the Signals or in the MSI activities.
Why can't you just 'not map' those events? Because Madkudu has built-in logic to make sure we don't miss out on anything interesting from your integration. Unmapped events will fall in a bucket called 'Other [integration] activity' (referred to as a 'catch-all' since it catches all events not explicitly mapped) that is automatically created by MadKudu. This catch-all event is not visible in the event mapping but you will see this event in the Data Studio. From there, you can review its importance and decide whether or not you want to allocate weight to it, or map some of the events inside it to get more granularity into your model.
What is the source of truth for each event?
If you track the same event in different integrations connected to MadKudu, you need to choose only one source for this event. You'll map the event from the source integration's event mapping and you'll exclude it from the other integrations' mappings (i.e. you'll let it fall into the catch-all event 'Other [integration] activity').
For example, if you have an event fill_out_form for your 'Contact Us' form in Marketo, but also have a Salesforce Campaign of type Contact Us whose members' statuses change to Responded when they filled out your 'Contact Us' form, we advise you to:
- In your SFDC event mapping, map your Campaign type Contact Us with Campaign member status Responded as the MadKudu event 'Responded to Contact Us' and select the activity type 'Marketing'
- In your Marketo event mapping, do not map your event fill_out_form with the correct attributes to identify your 'Contact Us' form, it will fall into the catch-all event 'Other [integration] activity' that you will later exclude from the model and the signals.
Another example: if you track your web visits both in Marketo and Segment, don't send the web visits data from Segment to MadKudu; only map your important web pages from your Marketo event mapping.
When to refresh your event mapping?
You and your team probably create new events (a new Salesforce campaign, a new webinar, etc.) on a regular basis.
These new events won’t be automatically taken into account by your Likelihood to Buy model.
Before you go to your event mapping and explicitly map them as a standalone event, they will fall in a catch-all named 'Other [integration] activity'.
We advise you to set up reports within your integrations to be aware of the creation of new events so you can be reminded to come to the MadKudu App and add them to your event mappings.
You can also see the list of events falling into the catch-all for each integration in-app, within the Insights tab.
During your weekly check-in on your models, please have a look at this Insight report to identify events not mapped.
How to add new events to the event mapping?
Important: if an event mapping is configured in madML (displays a SQL code), events should only be added in this code and not in the other connector sections. If the latter is done, the madML configuration will be overwritten and therefore your scores will be impacted.
Now that you know what activities you are going to map and from which integration they'll come from, you can get started and work in the App to map your events that will be ingested by your Likelihood to Buy model.
Since each integration has its own data architecture, the interface of the platform will slightly differ from one to the other. We made an article for each integration so you can have a detailed guide by your side when you’re working on a given event mapping, but here are indications that apply across all integrations:
- Connect to your account on app.madkudu.com
- Navigate to Mapping > Event mapping
- Select the integration you would like to add an event from
- Click on Start draft
- Click on Add new events. A new row will appear at the top.
- Configure the event you would like to map.
- Event: select or input the condition which defines the user activity. This column should be read as "IF event = ' .....' [AND sub condition] THEN the event will be mapped into:
- Negative user Activity: check the box if the event represents a negative behavior (unsubscribe, delete account, etc.)
- MK Event name (Signals): choose a user-friendly name that describes the visitor/user action. This is how this event will appear in the Likelihood to Buy Signals in your CRM. Think of this as the label name of the event that you can change at your own will.
- MK Event name: choose a user-friendly name that describes the visitor/user action. This one will be used to configure the weight of each event in the Likelihood to Buy model in the Data Studio. Think of this as the technical or API name that you won't want to change over time. If you think some events are too granular, we recommend grouping them together by giving them the same MK Event name & MK Event name (Signals). It helps to have more volume for the same event to have statistical significance but also to make sure it's clear for the signals shown to Sales.
-
Activity type: choose what best defines the category of the event:
- Web Activity: interactions with your website
-
App Usage: interactions with your product
-
Marketing Activity: webinars, events, content downloading
-
Email Activity: email interactions, (un)subscriptions
- Sales Activity: sales calls
-
Non User Activity: passive activities, email sent, email soft bounced
Tip: In the signals, you may want to display more granular events while grouping some events together under the same name (in MK event name) to use in the model.
For example: in the signals in your CRM, you would like to display the fact that a visitor downloaded your most recent eBook called "Let's get this party started" different from your previous eBook "How to send your first invites". Nevertheless, at the end of the day from an activity standpoint, you still want to group them both under "Downloaded eBook" and configure it this way in your model.
Thus, you would map 2 events as shown in the screenshot below.
Important: the order of how the events appear in this interface is important. If you map the following events:
- "Event = form submission and title contains ABC, THEN MK event name = Downloaded ABC"
- "Event = form submission and title contains ABCDEF, THEN MK event name = Downloaded ABCDEF"
You will never see the event "Downloaded ABCDEF" because only the first condition met maps the event. All form submissions of ABCDEF would instead be mapped "Downloaded ABC".
7. Click on Save draft when you are done, or frequently if you add more events, to make sure your changes aren't lost if you get disconnected. This will save your edits in the draft version.
8. Any event that MadKudu pulls but is not caught by any condition in your integration's event mapping will appear in the model and the signals under a bucket called 'Other [integration] activity'. Since new events created in your integrations are not automatically mapped, they will end up in this 'catch-all' bucket before you map them.
9. To deploy your changes to production, click on Publish changes.
10. Once the event mapping changes have been published, you will need to configure the importance of these events in a draft model. To do so, follow those steps >> How to make newly mapped events appear in the Data Studio Event Weights tab?
Please note that your newly published events won't be available in the Studio before the process map has run at least once after publishing. You can monitor this in the App > Settings > Processes. Learn more
How to edit events already mapped in the event mapping?
We mentioned that editing events in the event mapping can impact the scores. Here are more details about why and how.
- Per the instructions above, the column "MK Event name" essentially stores and controls the names of the events used in the behavioral model. Therefore, if you change the name in this column without reloading the dataset of the model, the people who have performed this event won't get points from this event anymore because the model will not recognize the event. Therefore, their score will decrease automatically in your CRM.
For example: If you change an event called "Requested a demo" to "Submitted a demo request", the model will not recognize that "Requested a demo" anymore.
For example: if you change from [Event = 'form_submission" AND title contains contact] into [Event = 'form_submission" AND title contains demo], then all the people who submitted a form "contact-us" will not be scored with this event anymore but the people who submitted a form "demo request" will now be scored with this event.
Therefore, make sure to keep both conditions instead of replacing one with the other ones if both of these forms still exist and mean the same for users.
Otherwise, create a separate event and make sure to configure its importance in the Data Studio.