What we learned building Talk2UsNow

Talk2UsNow helps local businesses turn happy customers into Google reviews. A customer scans a QR code at the counter and rates their visit. Then every customer, whatever their rating, gets the same choice: post a Google review, or send the owner a private note.

The idea is simple. Building it well took a series of decisions that weren’t obvious at the start. These are the ones we’d share with anyone building a first product.

Design for the moment, not the feature list

The whole product depends on one moment: a customer at the till with their phone out, feeling good about their visit. That moment lasts seconds. If the page is slow, asks for an account or has too many steps, the phone goes back in the pocket.

Holding onto that moment made many decisions easy. No app to download. No sign-up. One question and five stars. Each star gets a short label as you choose it, from “Poor” to “Loved it”, so the choice feels quick and clear.

Cutting was the most valuable work

At different points the plan included SMS alerts, text-message verification codes, an AI assistant to write replies, several pricing tiers and a large admin console. Almost all of it was cut or postponed.

Verification codes are a good example. They would have reduced spam. But asking a customer at the counter to wait for a code and type it in would have lost most responses. Spam is handled in the background instead, and the customer never sees it.

Each cut made the product simpler, quicker to build and cheaper to run. None of them removed anything a customer or owner actually needed.

Rules and policy are part of the design

Sending only happy customers to Google while quietly keeping unhappy ones away is known as review gating. Google’s policies don’t allow it, and it isn’t fair to the people leaving feedback.

So the design had to give owners a private chance to fix problems without steering anyone. Our first version sent low ratings to a private form first, always with a visible link to review publicly. Reading the policy more strictly, we went further: now every customer sees the same invitation to post on Google, and the private note is simply a second option. We treated this as a product requirement rather than a legal detail, and it shaped the whole flow.

Not everything needs AI

Owners see each piece of private feedback tagged by topic, with a suggested next step. It would have been easy to use a language model for this. Plain keyword rules were enough. They’re predictable, easy to check and free to run, which matters for a product meant for small businesses. We wrote more about this in why adding AI doesn’t automatically make a product better.

Make the numbers honest

Owners judge the product by the numbers on their dashboard, so those numbers need to be right. If someone reloads the page or shares the link, that shouldn’t count as a new customer. Scans and redirects are counted once per visit. It’s a small detail that protects trust in everything else.

Write for the owner, not the developer

Business owners shouldn’t need technical words to use a product. Internally, each location has a “slug”. Owners never see that word. They see “Your link”, with a preview of the address. Finding a Google review link confuses many people, so the dashboard shows step-by-step help right next to the field.

Tests show the rules work, not that it scales

Talk2UsNow has more than 130 automated tests covering the rating flow, permissions, billing and security rules. They catch a lot. But they say nothing about how the system behaves under heavy traffic. That needs separate load testing, and it’s easy to forget when every test is passing.

What we’d tell someone starting out

• Find the one moment your product depends on and protect it.
• Keep a written list of what you decided not to build, and why.
• Treat rules, policies and privacy as design inputs, not paperwork.
• Use AI where it clearly helps, not because it’s expected.
• Don’t build detailed analytics before you have data to analyse.

Talk2UsNow is live now. We’ll share real results once there’s enough usage for them to mean something.