Here’s what I learned this week

If all these companies with pixels and scripts were tracking the right data and able to translate it into action then every ecommerce company would be wildly successful.

What we see though is a lot of companies that are trying to use existing data to measure that existing data in a different way to produce a different result.

Reread that last bit, the whole premise of what these companies are trying to do is to read data better than the software that makes more in a day than they’ll make in multiple years and all of a sudden they have the magic ability to do a better job with less data points than organizations which arguably know us better than ourselves.

Yeah, that’s not going to happen.

So they have been clutching to signal loss and the ability to increase the amount of I guess “abandon cart emails” that can be sent to people.

Let me help you out with that, just send a newsletter that people actually look forward to reading every week, no need for abandon browse emails or abandon cart emails, just remind people that your brand exists consistently every week.

I feel like we’re trying to control people’s actions instead of creating experiences that would make people want to interact with us.

Which brings me to what most people still don’t understand about zero party data and what we’ve been building or why…

How Algorithms Work

First off what is an algorithm?

You feed a computer rules to that it can calculate data combinations that deliver a set result.

Here’s a great example of how we use an algorithm that’s dependent on multiple variables to produce marketing advice:

So in the above examples we have inputs with corresponding math that goes on behind the scenes to populate a spread of the variables in relation to key markers and suggestions we have for potential actions.

These are dynamic though, if you segment all your answers by a combination of variables the numbers change because they are linked by the rules and the calculations that are tied to corresponding actions recommended.

In fact the whole model is multi-variate meaning that we can create an algo that looks for a certain corresponding pattern with multiple combinations to produce suggestions that are more likely to produce a result.

Have I lost you yet? I hope not, but if I have, just know that it’s data inputs that have been told what to do within a set of rules.

Here’s why this stuff is important and unique and in my humble opinion the only way forward in the world of advertising.

You see, there’s no way for us to feed our algorithm and corresponding data back into any of the advertising platforms, which means they can’t train on it.

The best we can do is look to leverage it internally to impact all the customer journey touchpoints that we have and potentially feed it lookalike audiences.

So these algorithms now provide a greater level of insight than any other pixel, script, or other method of evaluating data.

To gain an advantage against these large advertising spy machines, you need to collect alternate forms of data directly from people relevant to their customer journey.

This is the only way to IMPROVE your understanding of your customers on a whole rather than “try” to build a better algorithm based on the space junk of actions that real customers as well as bot traffic takes on your website.

This is my biggest problem with everyone chasing the idea of attribution in the space.

1. It’s never going to be perfect

2. It’s all using existing “signals” trying to make more out of them to provide value

The truth is most of those algorithms that people are using are all behind a black box where they don’t actually break down how they work.

I actually don’t think that Facebook’s reporting is all that off, is it all first touch, naw, but seeing an ad a few times can prompt people to type it into a search bar for sure.

One of the stats we started pulling around this is the percentage of first time purchases that are coming in from people that subscribed.

Then varying the level of our popup aggressiveness to force this number to be over 50%.

If I have prepurchase intent data from 50% or more of all first time purchasers I can do a few really cool things.

1. I know the breakdown of who’s most likely to convert so I can change my entire advertising approach to just target people like them

2. I know what products they purchased regardless of the answers provided so I can advertise a combination of the top products with the highest revenue

3. I now take the black box of 50% of sales and having zero data about who’s purchasing and turn it into an engine to look at behaviors of these specific people (Shopify’s server side pixel is in beta) pay attention to this as it’s not available in the API for app configurations

Wrapping up here on this topic, the success of AI will be a combination of creating and leveraging data sets in new ways.

My personal view on this is that based on a few algorithms that we currently are running, adopting and AI layer with better prompting should produce the results we’re looking for to provide any ecommerce store with a step by step checklist of what actions to take on a weekly basis.

It’s never been about combining data, it’s been about building additional data layers necessary from a large portion of first time customers (given that 70% never come back) and modeling their intent points to build reliable systems and guidance.

My prediction

Currently, we’re seeing people using “AI” which is really just chat based interfaces to do work that can be done manually through User Interfaces.

That’s not AI, that’s a trained Large Language Model on internal docs to help a poor UI problem that previously would have been solved with help articles and search.

This is the same we saw with quizzes, which are usually used to “collect data” but really cover up for poor navigation.

If you forced people to click on navigation on your homepage rather than long form, you’d get better data than through a quiz.

The future for a lot of websites will just be better tagging and better layouts that combine data collection with customer journey - this is what I see as being the future.

A few thought nuggets from last week for you to ponder

Treat customer support as part of your marketing budget, they already purchased your product, just treat any good customer service as a marketing expense complete with their own percentage of the daily ad spend to fix problems.

We all have 4k video cameras in our pockets yet we rarely see a ton of video on product pages, it’s an easy fix, it doesn’t even have to be perfect.

Be wary of any companies telling you they built an algorithm to better read your first party data on your website and apply that in any meaningful way, it’s like trying to spell words in alphabet soup, you see what you want to see even if it wasn’t intentional.

The data arms race has begun, it takes 90-180 days of data to start to build reliable models, if you haven’t started collecting pre-purchase intent data yet, you’re already behind.

Update on Jon as a Service (JaaS)

I retired Jon as a Service to rethink what I wanted to do with it, but we’ve been onboarding a lot of brands on the smaller end that turns out would be a great fit for it, so I’m probably going to relaunch it with a small cohort of brands.

Also Shien is going to market it’s infrastructure network and technology for all brands to leverage. Read about it here

The Takeaway

Data layers, pay attention to them. You need new ones to create real business insights and most of them aren’t connected to anything in a meaningful way.

Join the slack

Have a great week!

-Jon

Catch up on past posts: https://ecomwithjon.beehiiv.com/

You can learn from me: jonivanco.com