Skip to content
Article 7 of 11 — part of the Interpretability Engine Optimization series.
Juan Carlos López

Free Delivery. But Will They Carry It Upstairs?

“Free delivery” answers one question. A shopper on a third-floor walk-up asked another. Here is how to state stairs and assembly as data.

delivery conditionsassemblyproduct data
A sofa at the foot of a staircase, with delivery conditions written as separate data pieces

In this article: the part of Interpretability Engine Optimization (IEO) that starts from the question, not from the data. Its job is to work out what a machine has to know to answer one specific question, and then check whether the store has given it exactly that, in that form.

Here is what happens without it.

A shopper finds the perfect sofa through an AI assistant. Right size, right fabric, right price. She has one question left: "I live on the third floor, no elevator. Will they carry it up and put it together?"

Your product page says Free delivery. True. And the answer to her question is spread over three other places. The delivery page says free delivery is to the building entrance, anywhere in the metro area. A line further down says carrying up stairs costs $25 per floor. The assembly service exists, but it's included only with "Room of Choice" delivery, an $89 option described on a different page.

The assistant has "free delivery" and nothing it can connect to stairs, floors or assembly. It can't say yes, and it shouldn't say no. So she buys from a store that answered her in one line: Delivery to your room, stairs included, assembly available.

You could have done all of that. You even said so, in four places. Nobody could put it together.

Why "free delivery" isn't an answer

"Free delivery" answers "what does it cost to ship?". She asked something else: "will it get into my apartment, assembled?". That's a different question, and it needs different facts: your service area, which floor delivery stops at, what stairs cost, whether assembly is offered and under which option.

The store knows all of this. It's in the delivery team's head and in a few pages written at different times. What it isn't is attached to the sofa in a form a machine can read for that question.

Google, for one, already asks for delivery facts as separate pieces rather than one phrase. Its Merchant Center takes handling time and transit time as separate values, in business days, and calculates an estimated delivery from the maximum of each, taking cut-off times and business calendars into account, so it can show a delivery estimate with the listing. That's a machine saying, in its own vocabulary, that one phrase about delivery isn't enough to answer a shopper's question about it.

The technical name: the decision context

In IEO the question a machine has to answer is called the decision context, and it's the unit of the whole practice. The same product page can be fully interpretable for "what's it made of?" and useless for "will they carry it upstairs?". A store's data isn't interpretable in general. It's interpretable for a task.

So the work starts at the task. For "third floor, no elevator, assembled", write down what the machine has to resolve, where the truth lives, and what it costs to get it wrong:

What the machine has to resolve Where the truth lives Cost of error
Does the store deliver to her area? Your service map A sale that can't happen
Where does free delivery stop? Your delivery terms A surprise charge at the door
Is stair carry offered, and at what price? Your delivery terms A sofa left in a lobby
Is assembly offered, and with which option? Your service options A box, not a sofa

All four are yours to state. Once they're stated as data, the machine can answer her and the next shopper on a fifth floor with an elevator. Until then, questions like hers go unanswered.

Where it lands: the pieces, not the phrase

Here is the sofa, before and after:

The shopper asks What the store says, and where What the machine had What it has now
Will it get to my apartment? Delivery page: "free delivery to the building entrance, anywhere in the metro area" "Free delivery" Standard delivery: to building entrance, free · Service area: metro area
Will they carry it up three floors? Delivery page, lower down: "stair carry available, $25 per floor" Nothing Stair carry: available · $25 per floor
Will they put it together? Services page: "assembly included with Room of Choice delivery, $89" Nothing Assembly: included with Room of Choice delivery · $89
So what does my delivery cost? Spread over two pages No answer Room of Choice ($89) plus $25 per floor of stairs, assembly included

Nothing in the last column is a service the store didn't already offer. It's the store's real delivery options, written as the pieces a machine needs for her question, and attached to the sofa so they travel with it.

What to do this week

  1. Collect the delivery questions your customers actually ask. Your support inbox has them: stairs, elevators, old-item removal, assembly, time windows, what counts as "the door".
  2. For each question, write the facts that answer it, as data: service area, where standard delivery stops, each add-on with its price and conditions.
  3. Attach them to the products they apply to. Big-and-heavy items carry different conditions than a cushion. Say which is which.
  4. Make the product page and the delivery page agree. "Free delivery" can stay, with its scope: free to the building entrance; room delivery and stairs available.
  5. Check it once a quarter, or whenever the delivery partner changes. Delivery conditions change more often than the sofa does.

Practical example

A bike shop's product page says: "Ships in 1–2 business days. Free shipping over $500." The bike is $800.

A shopper asks an assistant: "Will it arrive ready to ride?"

Before you read on, work out what the machine needs for that question and what the page gives it.

The page gives: processing time and shipping cost. Neither answers "ready to ride".

The question needs: how assembled the bike arrives, what the buyer has to do, which tools are included, and whether local assembly is offered.

What the owner confirms when asked: the bike ships 85% assembled; the buyer attaches the front wheel, pedals and handlebar; the needed tools are in the box; a partner shop offers full assembly for a stated fee in certain cities.

What the data should say: Assembly on arrival: 85%, front wheel, pedals and handlebar to attach · Tools: included · Professional assembly: available through partner shops in listed cities, fee stated.

Now the machine can answer: "Not fully. You'll attach the wheel, pedals and handlebar; tools are in the box. Full assembly is available through partner shops in listed cities, for the stated fee." Same bike, same page, one more question answered, because someone started from the question.

That's the decision context from section 7 of the IEO Methodology: start from what the shopper needs to decide, list what has to be resolved, and state each piece where the machine can find it.


Check what's explicit, what's missing and what contradicts itself on one of your products → Run the free audit

To go deeper → The IEO Methodology, section 7 (doi.org/10.5281/zenodo.22416784) · nsolvia.com/ieo

Sources: Google Merchant Center Help, "Set up estimated delivery time" (handling time, transit time, cut-off, calculation and display).

← "Organic", "Made in Italy", "Doctor-Founded". Which of Your Claims Can a Machine Trust? · ⌂ IEO · Next: Shoppers Search Your Store for Things You Sell. And Find Nothing. →

Series: Interpretability Engine Optimization · Knowledge Domain: Interpretability (IEO)

Interpretability Engine Optimization (IEO)
The UNDERSTAND Layer
nsolvia.com/ieo