Thank you for sharing your experience and expertise. What makes the difference with standard ETLs? Around minute 8, you talk about function and the need to transform data, so Segment understands it. When talking about identity stitching, it's only possible when you can identify, like in your example, subscribe with an email. What about a simple web session? Is the tracking plan through protocol the way to retrieve the interaction aka user interactions like a purchase? Last but not least, and I work for marketing automation software ^^ What about personalization because the audience is cool, but how do you personalize the message channel, time, and content? Sometimes I have the feeling that CDP is what we did 15 years ago with a couple of developers and ETLs. But I might be totally wrong.
Hi Jonathan, great questions. Let me unpack them. "CDP is what we did 15 years ago with a couple of developers and ETLs" To some degree, this is 100% true. A lot of products have been done for years as custom applications within companies. The difference in Segment's case is the real-time aspect. They also offer batch ETL but this is not really part of their CDP approach. "What makes the difference with standard ETLs" The two major differences is the routing to different destinations like CRMs, analytics platforms, or ad platforms in real-time. And also the data collection which is basically a stream of incoming data. The storage in a database is similar to an ETL. What about a simple web session? In a web session, people get identified first via a cookie (when existent), or you identify them after they signed in or provided an email. "the way to retrieve the interaction aka user interactions like a purchase" Not really. You achieve that by sending events from the frontend or backend. The tracking plan makes sure that you send it in the right format. "but how do you personalize the message channel, time, and content?" The major difference between Segment to other tools is that Segment only passes on the information. So you can do the personalization in a different tool with Segment data. Let me know if that helps.
Is it more cost effective (not counting the dev time to implement) to use snowplow to and census to handle collecting the data, and then sending the data onward to marketing platforms or back into the crm? Or is another tool necessary like Holistic or looker to create the personas/segments? Or does the cost balance out? And now factor in the dev time… I saw your other video interviewing snowplow, where 2/3 way through it surfaces that hiring snowplow to do your implementation makes things work/it’s not really advised for devs who are not already familiar with snowplow to attempt to read the docs and then diy. Or does snowplow + census not handle the user stitching without some other tool? User stitching and it’s making roi hard to properly report is the key driver of why we started researching CDP.
Thanks for the great question - this would already deserve a different video. Cost effective is difficult as sole decision parameter here. Some things to take into account for making a decision: - do you need marketing data passed on immediately? It is potentially doable with Snowplow but harder - is personas important for you? If yes, it becomes hard to do this with Snowplow - Holistics can't really help you there, unless you develop a way to take a Holistics filtered dataset, pass it down and then sync it with an ad platform Snowplow and any Reverse ETL way (tool or own scripts) work really well if batch is ok and segmentation happens in the marketing tools (eg. customer.io has a pretty powerful segmentation).
I would not say this is ALL about Segments. Ima developer here to learn how to make this work, and there is none of that. So maybe the title should be...This is a general cursory, non technical overview of Segment. 🙏
Thank you for sharing your experience and expertise.
What makes the difference with standard ETLs? Around minute 8, you talk about function and the need to transform data, so Segment understands it.
When talking about identity stitching, it's only possible when you can identify, like in your example, subscribe with an email. What about a simple web session?
Is the tracking plan through protocol the way to retrieve the interaction aka user interactions like a purchase?
Last but not least, and I work for marketing automation software ^^ What about personalization because the audience is cool, but how do you personalize the message channel, time, and content?
Sometimes I have the feeling that CDP is what we did 15 years ago with a couple of developers and ETLs. But I might be totally wrong.
Hi Jonathan, great questions. Let me unpack them.
"CDP is what we did 15 years ago with a couple of developers and ETLs"
To some degree, this is 100% true. A lot of products have been done for years as custom applications within companies. The difference in Segment's case is the real-time aspect. They also offer batch ETL but this is not really part of their CDP approach.
"What makes the difference with standard ETLs"
The two major differences is the routing to different destinations like CRMs, analytics platforms, or ad platforms in real-time. And also the data collection which is basically a stream of incoming data. The storage in a database is similar to an ETL.
What about a simple web session?
In a web session, people get identified first via a cookie (when existent), or you identify them after they signed in or provided an email.
"the way to retrieve the interaction aka user interactions like a purchase"
Not really. You achieve that by sending events from the frontend or backend. The tracking plan makes sure that you send it in the right format.
"but how do you personalize the message channel, time, and content?"
The major difference between Segment to other tools is that Segment only passes on the information. So you can do the personalization in a different tool with Segment data.
Let me know if that helps.
@@timodechau family time but thank you 🙏 I definitely will learn a lot and keep you posted
Is it more cost effective (not counting the dev time to implement) to use snowplow to and census to handle collecting the data, and then sending the data onward to marketing platforms or back into the crm?
Or is another tool necessary like Holistic or looker to create the personas/segments?
Or does the cost balance out?
And now factor in the dev time… I saw your other video interviewing snowplow, where 2/3 way through it surfaces that hiring snowplow to do your implementation makes things work/it’s not really advised for devs who are not already familiar with snowplow to attempt to read the docs and then diy.
Or does snowplow + census not handle the user stitching without some other tool?
User stitching and it’s making roi hard to properly report is the key driver of why we started researching CDP.
Thanks for the great question - this would already deserve a different video.
Cost effective is difficult as sole decision parameter here.
Some things to take into account for making a decision:
- do you need marketing data passed on immediately? It is potentially doable with Snowplow but harder
- is personas important for you? If yes, it becomes hard to do this with Snowplow - Holistics can't really help you there, unless you develop a way to take a Holistics filtered dataset, pass it down and then sync it with an ad platform
Snowplow and any Reverse ETL way (tool or own scripts) work really well if batch is ok and segmentation happens in the marketing tools (eg. customer.io has a pretty powerful segmentation).
What is the best analytics tool? I hate GA4
There is not really the best, unfortunately. It always comes down what you need. Even GA4 can be a solution. So, why do hate it?
I would not say this is ALL about Segments. Ima developer here to learn how to make this work, and there is none of that. So maybe the title should be...This is a general cursory, non technical overview of Segment. 🙏
Fair point. I will put it on my list to create a video like "All the things you need to know about Segment as a developer".