When everyone zigs, we zag: five things ZAG does the other way
Sell a data platform, chase search rankings, chase prompts. Much of the industry moves in the same direction. These are five places where ZAG deliberately goes the other way, and why.
"When the world zigs, zag." The line was written in 1982 for a launch of black jeans, at a time when every pair of jeans was blue1. When everyone moves one way, go the other. It's the idea behind our name.
Going the other way isn't the point in itself. In product data and AI, we keep running into places where most of the industry does one thing and it doesn't work, and in those places we've chosen another path. Here are five of them.
1. Not another data platform: we work inside your workflow
Most vendors sell a new data platform. Adopting it means rebuilding years of proven workflow around the platform.
ZAG works inside the workflow you already have. The PDF catalogs, the internal spreadsheets, the PIM or ERP you run today all stay. We bring our engineers and our modules for intake, normalisation and output into that workflow and move it forward as it is.
We don't hand over software and leave you to master it; we take on the work itself. Some call this Service-as-a-Software, or forward-deployed AI engineering. Nothing has to be stopped or rebuilt to adopt it, so getting started is light.
2. Not chasing rankings: making product data readable by AI
Most companies still chase search rankings.
ZAG puts product data an AI can read first. In our study of 110 websites, mostly manufacturers, specs were written into the HTML on most sites, but model numbers were recognised as SKUs or GTINs on 1%2. A higher ranking doesn't help if an AI can't read your specs well enough to compare you.
3. Not chasing prompts: chasing personas
Most companies starting on AI search track a fixed list of prompts.
ZAG decides who is asking before writing any questions. A persona is not a job title or a demographic: it is a kind of buyer who asks an AI a question before buying, proven by real companies making a real, dated move right now. Most buyers never type your name, so the work is being the answer to the questions where nobody asked for you. More in our post on personas.
4. Not asking once: asking again where answers waver
Most measurement asks each question once and reports the share of answers that mention you.
ZAG asks the same question many times. AI answers change from one ask to the next, and whether they waver depends largely on the question. We spend repeats on the questions that waver, and every judgement carries the AI's own sentence and its sources. More in our post on measurement.
5. Not chasing fill rate: leaving a blank when unsure
Most data clean-up reports how many fields got filled.
ZAG leaves a value blank and flags it for review when the reading isn't certain. Everyone can see that a blank is still open. A plausible wrong value travels quietly into quotes and comparison sheets. More in our post on PDF catalogs.
Why we zag
All five come down to the same thing: what actually happens in the customer's work matters more than numbers that look good. Don't make people rebuild their workflow. Don't chase rankings an AI can't read. Don't count before deciding who is asking. Don't judge on a single answer. Prefer a blank to a wrong value.
This blog is where we'll write about those choices and the measurements behind them.
Sources
-
Written by Barbara Noakes at the agency BBH for Levi's black jeans, launched in the UK in 1982. The Marketing Society, "When the world zigs, zag". https://www.marketingsociety.com/the-library/when-world-zigs-zag ↩
-
ZAG research. Product detail pages on 110 websites, mostly manufacturers (n=110). Details in our product-data AI readability report. ↩