What? Cocktail Pokédex is evolving!

Welcome to Product Cocktail, where the takes are as polarizing as a shot of Fernet—but the insights come together like a perfectly crafted daiquiri.

Cocktail Pokédex was a joke.

At least, the name was.

I called it Cocktail Pokédex because it was funny, and maybe, in the back of my mind, a way to beat the LinkedIn algorithm.

Perhaps a small part of me remembered my days in the trenches with Pokémon Red on my Gameboy Color, grinding out tedious battles with errant Zubats while trying to catch 'em all to fill my Pokédex.

It stopped being a joke when Lovable produced a counter reading "14 of 151 specimens" with empty slots for unseen entries. 151 is the number of original Pokémon (real ones already clocked that).

I hadn't asked for a denominator. I'd asked for the opposite. The brief I handed all four tools back in June has a whole section devoted to keeping this thing from becoming a Pokémon parody. Every tool obeyed it — there isn't a single XP bar in any of the four.

Collectible was all it took.

When I originally saw this, I chuckled and moved on. Then I saw it reappear in the plan I co-wrote with Claude Code when I started to build this for real:

By including "collectible," I'd policed the paint and green-lit the foundation.

I started to wonder how much of the spec came from the product, and how much came from the name.

Every one of these is false for cocktails. I already knew this from the JTBD I wrote:

Every piece of this is about a specific encounter. None is about cataloging cocktail "species". None is about completion.

I wrote that sentence in the same document that named the app. Two lines apart — the name first, the job second. It disagreed with the metaphor from the first draft, and I read past it for two months.

When I built my test data set for the spike, I ended up with 12 of 15 pairs that were house originals. Two of the remaining three "classics" were house riffs: The O.F. and Privateer Select Mai Tai.

This is about cataloging uniqueness, not checking 151 boxes just to say you've had every canonical cocktail.

Last issue ended on the spike rewriting the UI backlog and data model. This is the same spike and same answer key, doing it a second time, but this time it's rewriting the product concept rather than the backlog.

I'd parked the catalog question — curated list, or emergent from what you capture — while I collected test data and ran the spike. On day three I started the agent back up and found a batch of open questions already closed — on data rather than argument. This one read:

Two counters: one for progress against a curated list, one for unbounded discovery count inclusive of house originals.

It's a reasonable idea. It preserves the metaphor, acknowledges the data, and it fits on a screen.

It fails the Poké-test. It's metaphor-preservation dressed up as product design.

With two counters, 80% of your activity (at least, according to my ordering habits) lands in the bucket that doesn't count toward completion. I'd be building a progress bar that barely moves while the user does the thing that the app is for, and then adding a second number next to it as an apology.

Read that quoted line again: "or the core metaphor breaks." That's not a product constraint. It came from the name, and by then it was central enough to shape a recommendation made entirely out of correct data.

I pushed back without much confidence: "I'm wondering how much the 'pokedex' metaphor is polluting this. That was a shortcut." Directional taste, not a conclusion. It was the right thread to pull.

I'm not trying to influence ordering habits with this app, I'm trying to replace the Notes app, or forgotten captions on geotagged images of cocktails.

A bottle of wine is a SKU with a barcode. Vintages vary, but a given bottle is the same liquid anywhere on earth. Identify it and you're done.

The catalog is the product. Vivino's design follows from that. So does Untappd's, for beer.

A cocktail is a performance. Sometimes an art form, sometimes a bastardization (try ordering a margarita at TGI Fridays). It breaks the catalog model twice.

First: 80% of what I captured has no catalog entry at all. Nobody has published a spec for a drink one bar invented last spring. My camera roll skews toward unusual drinks, so 12 of 15 is directional — but it isn't close.

Second, and this one doesn't depend on my sample at all: for the 20% that does have an entry, the entry doesn't tell you what you drank. The IBA will confirm a Negroni is equal parts gin, Campari, and sweet vermouth. It won't tell you which gin, which vermouth, or how long it was stirred — and on a Negroni, that's most of the drink.

I had been unwittingly building an identification product for a category where identification is the least interesting thing you can do.

The evidence is the artifact. Under a collection framing, you only need to identify the species. The characteristics live in a database. Charmander is a 2' tall lizard. Under a preservation framing, the observed tier (glassware, ice, garnish, color) is the artifact. Menu-only API calls scored 0-7% on these four fields. Last issue argued that the drink photo was worth keeping in the API call for the marginal performance improvement. It was never the tiebreaker, it's the entire point of the capture.

The schema adds a field it never had. The JTBD's final clause is "whether I'd order or make it again." Further down the same document, I listed the thirteen fields a specimen entry must carry: name, bar, date, photo, reconstruction,
ingredients, method, glassware, garnish, flavor profile, spirit base, notes, specimen number. The final clause is not among them. The metaphor was already doing the selecting.

I brought it back as two fields, because they're answered at different moments: would_order_again at the bar, want_to_make at home once you've got a reconstructed spec in front of you.

The curated list survives, as a feature. It returns as a Cocktail 250 — an IMDb Top 250 for cocktails, an unranked checklist ticked off as encounters happen to match. It's a view over encounters, not a schema constraint. There's no second counter. This metaphor wasn't worthless, it was mis-scoped.

Calm down, nerds. It would probably be closer to 102 based on official International Bartenders Association (IBA) standards.

The product was never a Pokédex. I renamed it days after I reframed the product vision.

I asked four AI prototyping tools for an app in the style of a "vintage naturalist field guide / specimen catalogue" whose entries would be observed, classified, illustrated, and preserved. That's a preservation product in the same document that called it a Cocktail Pokédex. The art direction and the title disagreed from day one, and the title is what drove the product direction.

Lovable took collectible and built a bounded universe with blank slots for drinks nobody has invented. Meanwhile, the art direction was already describing a field guide. The brief contained the wrong metaphor and the right description of the product. Building the schema told me which one to keep.

The subtitle survived. The name didn't. It's Proof — a field guide to cocktails.

Pokédex is Nintendo's IP and this ships to two app stores, so it was going to change whether or not I'd come to this insight. (I'm sure the cease and desist letter is already in the mail after the couple dozen times I've used the P-word in my newsletter.)

Your metaphor is influencing your spec, whether you authorized it or not. The check is cheap, but few actually run it: state what your metaphor asserts as a list of claims. Mark each one true or false for your actual product.

I didn't miss the catalog decision, I parked it. The damage happened while the decision sat on a list marked "decide later."

Next week, I'll cover the measurement issues during the spike that almost pulled the cocktail amounts feature.

I've wrapped up some critical schema definition work and initial testing on the App/Play Store SDKs. I'm moving onto porting the design from v0 and getting an initial version into the app stores. I'll keep you posted on anything insightful that comes out of these steps.

Send me questions, feedback, and cocktail recipes:
[email protected].

_{Icons made by Icongeek26 from www.flaticon.com.}