Skip to content

Generative Engine Optimization

Day 171: Cancelling a Booking Does Not Rewind the Business

The appointment is cancelled. The preparation has already been done. The confirmation has already been read.

An Undo button cannot make all three disappear.

For founders and marketing leaders commissioning agent-assisted booking, recovery belongs in the commercial budget. A successful booking creates commitments; reversing it can require new work, leave costs behind and finish somewhere other than the original starting point.

This is not the retry problem. Retrying asks the service to complete the same intended action without duplicating it. Cancellation asks for a different outcome after that action succeeded.

Day 170: Compare the Item, Not the Product Family

The cheapest price and the largest capacity can belong to the same product family without belonging to the same product.

Put both in one comparison row and you have created something the buyer cannot order.

For a marketing director approving a competitor comparison, the first question is therefore not “Who wins?” It is “What exactly occupies each row?” A brand, a family and an individual variant answer different questions. Treating them as interchangeable can make the price judgement meaningless before anyone starts debating features.

Here is a procurement redline: an explicitly fictional buying comparison, not a captured AI answer, live catalogue or ZSA client result.

Day 169: More Prompts Do Not Create More Treated Cases

A team rewrites one product guide, then asks an answer engine about it repeatedly. The report grows with every capture. The number of separately treated guides stays at one.

That distinction matters when a CMO asks whether to fund the same rewrite across a catalogue. More observations of one intervention are not more independently assigned cases of that intervention.

For Day 169, the design choice is where to spend an evaluation budget: deeper observation of one changed unit, or a comparison across separately assigned units. The two sketches below are fictional methodological planning—not an observed test, client case or validated ZSA method.

Day 168: Let Readers Use the Chart Without Seeing It

The chart has a headline. The bars have labels. A reader using a screen reader encounters “Research findings”.

They have been told that research exists, but given none of it.

For a marketing director commissioning a benchmark report, the missing deliverable is substantial: a way to inspect the information without seeing the graphic. That belongs in the original research budget, alongside analysis, writing and visual design.

My proposed commissioning choice is to buy the chart and its complete text equivalent together. Below is a brief for one hypothetical graphic, not a delivered ZSA accessibility project or a tested publication.

Day 167: Commission the Free Sample of Your Paid Publication

A specialist publication has to give a prospective subscriber something worth reading before asking them to pay for more.

How much should that first reading experience deliver?

Treat the answer as an editorial commission. Someone must choose the useful material a non-subscriber receives, the promise made at the paywall and the expense the publisher can sustain. Leaving those choices to a default subscription setting leaves part of the product undesigned.

For Day 167, here is a proposed experiment memo for a fictional industrial-energy publication. No publisher, subscriber behaviour or commercial result was observed. The proposal is to test a lead-in: a useful opening portion of each selected article, followed by paid access to the remainder.

Day 166: Keep Product Research Open When Stock Runs Out

Keep the specifications available when the shelf is empty.

A buyer researching a product may still need its dimensions, compatibility or care instructions before deciding whether to wait. Removing that information because checkout is unavailable makes a stock decision do the work of a publishing decision.

For a merchandising or marketing lead, the useful distinction is between the product someone can investigate and the transaction the shop can currently offer. A temporary stockout can close purchasing while leaving research open.

The practical choice is what to retain, what to stop selling and when to accept a future order.

Day 165: Give Customer Service Its Share of AI Referral Value

Someone arriving from an AI assistant may already pay you.

They may need to change an account setting, understand an invoice or get a product working again. A useful visit could end with the problem resolved and no sales conversation at all.

For a marketing leader, that creates two ways to misread the same channel: claim customer-service demand as new pipeline, or dismiss useful self-service because it produced no lead.

The commercial question is broader: what work did the visit help someone complete? AI-originated traffic can carry prospect research and existing-customer needs through the same entry channel. Those workloads deserve different accounts of success.

Day 164: Faster Intake Is Not Faster Service

Imagine a company that replies to every enquiry immediately. Its agent gathers requirements, checks fit and offers the next step without waiting for a person.

The buyer still has to wait for a specialist to become available before the work can begin.

The agent may classify every request correctly. The problem is more basic: speeding up one stage does not increase the capacity of the stage that must deliver the service.

For founders and marketing leaders, this is the difference between buying a faster intake funnel and improving the customer outcome. The first can relocate waiting.

Day 163: Keep One Booking Intent Through Every Retry

The buyer asked for one booking. The agent sent the request, but the acknowledgement never arrived.

Should it try again?

A timeout cannot answer whether the first request failed. Stripe’s current error-handling documentation says that, after a network error, a client may not know whether the server received the request.[1] AWS describes the same distributed-systems dilemma: simply repeating a resource-creation call can create a second resource if the first call succeeded but its response was lost.[2]

That ambiguity turns one commercial intention into a transaction-design problem. If an authorised agent retries, the service must be able to recognise “this is the same booking attempt” without confusing it with “the buyer wants another booking”.

Day 162: Your Buyer May Need Two Products, Not One Winner

A shortlist assumes the buyer must choose one winner.

But some buying problems are not solved by selecting the best product inside one category. The buyer needs several products to work alongside one another. Treating every adjacent brand as a competitor can hide the more useful commercial question:

What else must the buyer acquire or organise before our product can complete the job?

Economics gives us a simple distinction. A substitute can replace another product to some extent. Complements are often used together, so using one tends to enhance the use of the other.[1][2] That concept does not prove demand for any particular B2B combination. It does give marketing teams a better way to inspect the market around the buyer's whole job.