~/codigoagente
AI agents

Pricing and decisions: designing the page where the reader finally pays, with marketing-psychology

In part 1 you installed the skill. In part 2 you learned to persuade without tricking. This part closes the guide where the money is actually made: the pricing page, the closing block of your articles, the screen where the reader decides to pay or close the tab. We will apply the five skill models that move the most conversion on a choice page (anchoring, charm pricing, decoy, Hick's Law and activation energy) with examples from this blog and the companion Hotmart store.

A reminder of what powers all of this: the marketing-psychology skill is MIT (coreyhaines31/marketingskills) and we combine it here with cro from the same repo to decide where each element goes.

A well-made pricing page is not longer: it is three numbers ordered with care · codigoagente

Anchoring: the first number rules the rest

When the reader sees price A before price B, their brain judges B as relative to A. This is not a trick: it is mental arithmetic. We use it in two concrete spots in the blog.

Technical anchor: the context paragraph in the slopus post reads "training an LLM from scratch costs 500k to 10M USD". After that, the Claude Max plan (200 USD/month) reads as a steal. Real datum, not invented, and it respects the ethical rule from part 2.

Commercial anchor: on the Hotmart kit page the "Claude Code de cero a producción" ebook (19.99 USD) appears before the Kit (4.99 USD) in the secondary list. The reader reads the kit as "the cheap starter", not as "the main product". If the kit were first, 4.99 USD would act as the anchor and the ebook would feel "four times more expensive".

Rule for your agent:

snippettext
In a pricing page or product list, order prices from high to low when you want
the small one to feel accessible, and from low to high when you want to raise
the average ticket. Never mix without a plan: order is not decoration.

Charm pricing and rounded pricing: pick a voice

The skill distinguishes two opposite pricings:

  • Charm pricing (0.99 or 0.95): a "value" signal. The brain processes the leftmost digit. 19.99 "starts with 19".
  • Rounded pricing (0.00): a "premium" signal. 20 USD sounds like a professional product; 19.99 sounds like a sale.

In codigoagente the Hotmart products use charm pricing (19.99, 14.99, 9.99, 4.99) because the audience is price-sensitive. The Safe Build Labs plan when we sell custom development uses rounded pricing: "starting at 2,000 USD" sounds like a company, not a midnight deal.

Tell your agent to decide before writing the price:

snippettext
Before setting a price on a page, tell me if the voice is "value" or "premium".
If value: 19.99, 14.99, 9.99. If premium: 20, 50, 200.
Never mix: a premium plan with charm pricing reads as clearance.

Decoy: the sacrificial plan that pushes toward the middle

The decoy effect, proved by Dan Ariely in *Predictably Irrational*, says that a clearly worse option makes the one you care about look obvious. Classic example from his book: *The Economist* sold digital at 59 USD, print at 125 USD and both at 125 USD. 84 % chose the bundle because print-only exists. Without that ugly option, 68 % chose digital only.

On the kit page we were tempted to add a decoy plan. We decided not to for an ethical reason: if you create an option you do not expect anyone to buy, you are manipulating the decision, not informing it. Instead we applied paradox of choice (one recommendation) and framing (reciprocity in the supporting text).

Rule for your agent:

snippettext
Do not propose a decoy plan you do not expect anyone to buy.
If the "sacrificial" plan makes sense for someone (e.g. a real starter for a small segment),
document that segment. If it fits nobody, delete it.

Paradox of choice: fewer options, more decisions

Three plans convert better than seven. The SKILL.md leans on Iyengar's study (24 jams at 3 %, 6 jams at 30 %). The same logic shows up when a page moves from "here are four equal products" to "this is the recommended one and the other three sit below if you prefer": the decision turns into a simple comparison rather than a table.

The concrete pattern that works in codigoagente:

  • A big card with the main product, "Buy the kit" button and secondary "See what's inside".
  • Below it, a small list with the other three products and a link to each.
  • Nothing at the top of the page that duplicates the decision.

If your agent proposes four plans in parallel, ask it to justify with data why they are not three. Most of the time the fourth plan is inertia, not strategy.

Hick's Law: every extra button is a pause

Hick's Law formalizes the obvious: the more options there are, the longer the decision takes. On a payment page that translates into abandonment.

Real example: the "It helps us grow" block at the end of posts lists WhatsApp, LinkedIn, X and Facebook. We chose not to add Telegram or Threads even though they are real channels: a fourth button is already the comfortable limit before the reader starts reading and hesitating. A primary CTA should be unique; share buttons work as a secondary cluster, not scattered.

Rule for your agent:

snippettext
On any page with a decision to make, list every clickable element.
If there are more than 4 CTAs above the fold, remove the least used ones
by your analytics. On a pay page, the goal is one highlighted CTA and at
most 1 secondary link (FAQ or questions).

Activation energy: lower the cost of starting

The SKILL.md calls this activation energy: the minimum energy to take the first step. If you ask for too much before the reader pays, they close the tab even if they are interested.

Two blog examples:

  • The newsletter form only asks for email. No name, last name or company: every extra field you will not use is friction the reader pays for and you do not.
  • The "Use it now" block lets you copy the prompt without downloading anything. If the reader wants, they download the .md with a button. Friction sits at both levels so they can choose.

Rule for your agent:

snippettext
Before asking for data, ask yourself the absolute minimum needed.
If you can start without an account, do it. If you can start without a card, do it.
Barriers go in once the reader has momentum, never upfront.

Checklist you can hand to your agent

The end of the marketing-psychology skill has a "Quick Reference" table. Here is the version we use to audit a pricing page. Tell your agent:

snippettext
Review this pricing page with marketing-psychology. Check one by one:

1. Anchoring: is the first number the reader sees the highest or lowest? Is it coherent with the voice (value or premium)?
2. Charm vs rounded: are the prices coherent with the product voice?
3. Decoy: is there a plan nobody is expected to buy? If yes, justify it or remove it.
4. Paradox of choice: more than 3 plans? If yes, justify or reduce.
5. Hick's Law: how many CTAs above the fold? If more than 1, which to drop.
6. Activation energy: what is the first thing asked of the reader? Can they start with less?
7. Peak-end: what is the last sentence the reader sees after picking a plan?
8. Loss aversion: is the refund or guarantee visible? Does it state the exact condition?

For each item answer with the exact line of code or text and the proposed change.

Feed it your page HTML and you get a concrete audit, with a diff ready to commit. If anything in the diff crosses the ethical line from part 2, do not accept it.

FAQ

Do I need to be a psychology expert? No. The skill is exactly so you do not have to be: it loads the models, you give product context, your agent applies the right part. You decide whether to publish.

What about A/B tests? The skill does not replace them. A mental model is a hypothesis; an A/B verifies it. For the test mechanics, the open-seo post covers the design small teams usually run before they have volume.

Can all of this apply to a README? Yes. A README of an open source project is a pricing page where the "price" is your time. Anchoring (install badge), decoy avoided (do not list inferior alternatives), activation energy (npm i in the first code block).

What is next

If you read all three parts, you already have everything the skill offers without reading the 12 KB SKILL.md. The real next step is the diff: copy the prompt from part 1 and run the audit on your own project. If you want to keep exploring this skill family, check the reverse-engineer-marketing post and the four skills to build a landing page guide.

// source code

Repository in this article
The code this article covers is public: browse it, star it or clone it from here.

Help us grow

codigoagente is free, with no paywalls. If this article helped you, any of these keeps the blog going.

// newsletter

One email a week with the repos worth your time

The week's new articles, tested against real code. No spam; unsubscribe in one click.

AI agents

guide · The complete ECC guide for Claude Code · #8

AgentShield, ECC's scanner, checks CLAUDE.md, settings.json, MCP, hooks and agents. We ran it on a deliberately vulnerable setup and on ECC's own install.
  • Security and code review
  • Claude Code
read article
AI agents

guide · The complete ECC guide for Claude Code · #7

We measured ECC's context footprint with /ecc:context-budget --audit (about 12,500 tokens on the minimal profile) and how to trim it: model, compaction and MCP.
  • Claude Code
  • Skills for your agents
read article
AI agents

guide · The complete ECC guide for Claude Code · #6

What ECC's 28 hooks do, how GateGuard stops the first edit, how to save and resume sessions and how continuous learning instincts work, with a real test.
  • Claude Code
  • Skills for your agents
read article