The cost of attribution data an agent cannot read
Unstructured measurement is a tax paid twice: in assembly time now, and again when the record cannot be re-read, checked or moved later.
Three platforms, three answers, one budget to defend. Your GA4 export already holds the answer, and the first finding is free and never leaves your browser. Get your first finding free
By Joris van Huët, Founder & CEOUpdated 2 min read
Measurement locked in an interface is a tax paid twice: once in assembly now, and again when the record cannot be re-read later.
The two payments
| Payment | When | Size |
|---|---|---|
| Assembly | Every cycle | Hours, and it is the visible one |
| Irrecoverable history | When you switch tools or check your own past | Larger, and invisible until it lands |
The assembly tax
If the numbers live only in a rendered view, every report starts with somebody transcribing. That is hours per cycle, and it is the reason reports get skipped in busy months, which is when the spend was highest. The cost of that pattern is in the cost of measurement that runs when remembered.
The irrecoverable history tax
This is the expensive one. A year of screenshots preserves impressions and loses the fields.
So when you want to check whether last year's decisions were right, or run a second method on last year's window, you cannot. The comparison that would have validated your whole approach is unavailable, and it was unavailable from the moment the data was rendered rather than recorded.
The agent connection
An agent is just the most demanding consumer of the same property. If a model can reason over your measurement correctly, you can re-read it in three years. If it cannot, neither can you. Designing for one gets the other free.
What it costs to have this
The source stays in your own analytics account, so any window you can export is a window any method can read later.
The question to answer this week
Can you produce, as a file, the per-channel estimates and intervals you used for a budget decision six months ago. If not, that decision is unauditable, and so will this month's be.
Related answers
Frequently asked questions
What does unstructured measurement data cost?
Assembly time every cycle, which is visible, and irrecoverable history, which is not.Why does losing the fields matter later?
Because it makes an annual review impossible. You cannot check whether last year's decisions were right, or run a second method on last year's window, if the record kept only the pictures rather than the numbers behind them.How do agents relate to data portability?
An agent is the most demanding consumer of the same property. If a model can reason over your measurement correctly, you can re-read it yourself in three years. Designing for one gives you the other at no extra cost.
Go deeper: Causal attribution, explained.
Sixty-second versions of these ideas: Causality Engine on YouTube Shorts.
Have an audience of ecommerce operators? Earn 20% recurring commission by recommending the €99 read.
Keep reading
Terms in this article
- AnalyticsAnalytics is the systematic computational analysis of data. It reveals customer behavior and measures campaign performance.
- AttributionAttribution identifies user actions that contribute to a desired outcome and assigns value to each. It reveals which marketing touchpoints drive conversions.
- ImpressionAn Impression counts each time an ad or content displays on a user's screen. It measures exposure, not engagement.
- ShopifyShopify is an ecommerce platform for creating online stores and selling products. Attribution modeling shows which marketing channels drive traffic and conversions within Shopify.