Case Study: Cutting Two Dishes and Rewriting Four Descriptions

Background: Meera Iyer owns Nara House. Like most independent owners, she had years of experience running her floor and a strong sense of her menu built on instinct, guest comments, and watching what came back to the kitchen untouched. What she didn't have, until she started using real dish-level analytics, was a clear, evidence-based picture of which dishes were actually pulling their weight.

The situation going in

Nara House's menu had grown over time the way most restaurant menus do, additions here and there, nothing formally reviewed against performance. Some dishes were popular guesses that turned out right. A few were sitting on the menu mostly out of habit, with no one quite sure whether they were worth keeping or if they'd just always been there.

What the data surfaced

With dish-level analytics available for the first time, two clear patterns emerged. Two specific dishes showed consistently low order and reorder rates relative to the rest of the menu, evidence rather than a hunch that they weren't earning their spot. Separately, four dishes with decent order volume showed a gap between how often they were ordered and how positively guests responded once they had, suggesting the written descriptions weren't representing them well.

• Two underperforming dishes identified through consistent low reorder data
• Four dishes flagged for a mismatch between description and actual guest response
• Decisions based on a full month of real order and feedback data
• Changes implemented within that same first month of using the dashboard

The action taken

Meera cut the two underperforming dishes from the menu, freeing up prep time and menu space for things actually working. She rewrote the descriptions for the four flagged dishes, aligning the written pitch with what guests who'd already tried them clearly responded to. Both moves came directly from the data in front of her, not a guess about what might be wrong.

Why this matters beyond the specific numbers

The value here isn't really about the exact count, two dishes, four descriptions, it's about the shift from decisions made on instinct to decisions made on evidence, in a fraction of the time it would have taken to reach the same conclusions by observation alone.

What happened after the changes went live

Cutting a dish always carries some risk that a handful of loyal fans will notice and ask about it, so the two removals at Nara House were watched closely in the weeks after. The dashboard made that easy to track directly, no complaint volume tied to either removed dish, and no dip in overall table traffic that could be attributed to the change. The four rewritten descriptions showed the more interesting result: order rates for those specific dishes moved up noticeably against their own prior baseline, suggesting the original descriptions really had been undervaluing what guests were experiencing at the table.

The part that's easy to overlook: staff reaction

A change like this doesn't just affect the menu, it affects the people preparing and serving it. Removing two dishes freed up real kitchen time that had been going toward low-volume prep, time that got redirected toward the dishes actually carrying the menu. Servers reported an easier time describing the four rewritten dishes to guests, since the new descriptions matched more closely what guests actually said when they liked something, rather than a generic pitch written before anyone had real feedback to work from.

What this case study doesn't claim

It's worth being precise about what this case actually shows. It isn't a claim that every restaurant should cut exactly two dishes or rewrite exactly four descriptions, those numbers are specific to Nara House's own menu and its own data. What generalizes is the process: get real dish-level evidence, act on the clearest patterns first, and check the results afterward rather than assuming the change worked just because it made sense on paper.

This is the kind of outcome TasteGraph's dashboard is built to produce: real dish-level analytics that turn a vague sense that something's off into a specific, actionable list, the same way it did for Meera Iyer at Nara House, start to finish within a single month of real use.