Freebee Ads docs

Product docs

What we changed: the November rethink

Last updated Nov 16, 2026 · @praveen · 9 sections

In four days we turned the plan in What we sell, from first principles into working pages. A buyer now starts with what they want, sees plans with an expected result, and sees each screen's numbers before paying. The mall, ops, Sales, the shopper and the shop counter each got the pieces that plan asked for. Every signed-in page also got a new look. This page says what was built, for whom, where we moved away from the plan, and what is still open.

In short

  • Before: a buyer picked screens, slots and dates, paid, and only then saw any number about people.
  • Now: a buyer picks one of four reasons, a budget and dates. We show two or three plans, each with an expected range and a cost per result. Every screen has a scorecard anyone can read before paying.
  • After the ad runs: the results page shows one big number, the one the buyer paid for, and then the chain that led to it.
  • New desk: a Sales person has their own pages to bring a shop from "hello" to a second booking.
  • One rule got stricter: a shopper's phone number never goes to an advertiser. Not with a yes, not for a price.

The eight directions, and where each one stands

These are the eight directions from the first-principles doc, in the same order.

#DirectionStatusWhat exists now
1Sell results, keep slots underneathBuiltBooking starts with the reason, the budget and the dates. Two or three plans, each with a range, a cost per result and one line that says why. "Choose screens myself" is still there as a side door
2A scorecard for every screenBuilt28 days per screen: people by hour, how many looked, seconds of looking, games, scans, coupons. Open to buyers before they pay. The mall sees it for its own screens. A screen with no camera says "not measured"
3Start with shops and restaurants in the mallBuiltThe mall can invite its shops. Sales has a desk built around a shop owner. A shop with no video can make an ad in two minutes
4Call meHalf doneThe shopper says yes to one brand, then sees the shop's own number and a "Call this shop now" button. We count the yeses and the taps. "We connect the call" waits for a lawyer
5Sell "Try me" by hand to big brandsHalf doneSales can make a Try me plan and send it. The sampling report for brands is not built
6Make the ad for themBuiltA library of ready-made ads, sorted by kind of business and by reason. The shop adds its name and offer. Ops still approve it like any other ad
7The top-up promiseHalf doneWe work out when looks fell below the range, and ops record "owed" and "given". The free plays are still added by hand
8Prove it by switching offTo doNot started. It needs a shop willing to share its sales

What changed, person by person

The advertiser

  • Booking by reason. The new start asks one question first: what do you want? Remember me, Try me, Come now, or Call me. Each reason also counts the ones before it, so a Come now campaign still reports looks and games.
  • Plans. From the reason, the budget and the dates we build two or three plans. Each one names its screens, its price, the range we expect ("4,000 to 6,000 looks") and the cost per result. The range comes from the real day-to-day spread of the last 28 days. A screen we cannot measure gets no range.
  • Scorecards before paying. A list of screens and one page per screen, with the people numbers that used to be hidden until after payment. Each screen is marked walk-past or sit-down.
  • Ready-made ads. A shop with no video picks a ready ad for its kind of business, types its name and offer, and sends it for approval.
  • Results by reason. The campaign report now opens with one headline number against the range we promised, the cost per result so far, and the full chain: plays, people near, looked, played, coupon, counter, rupees, came back. Where we have no measure, it says so.
  • Kind of business. Asked once at sign-up and kept in settings. It picks the right ready-made ads and plans.
  • Home page. Rebuilt in the new look. It keeps everything the old page showed, and adds a "Needs you" list.

The mall operator

  • Its own promotions. The mall uploads and schedules its own content in its share of the loop. Ops approve it before it plays.
  • Its own yes or no on ads. The mall answers each ad itself, with a reason for a no. A no on a playing ad takes it off that mall's screens.
  • Scorecard for each of its screens. The same numbers a buyer sees.
  • Booked beside paid. The overview and the statement show both, with one line that explains the gap.
  • Shops. A list of the mall's shops, and an "Offer screen time" invite for each. The mall can see who was invited, who signed up and who booked.

Freebee ops

  • Ready-made ads library. Add, tag, switch on and off, preview, and see how often each one is used. The library starts empty on the live site, so ops add the first ads there.
  • Plans sold. Every plan, its promise, and where it stands. This is also where "owed" and "given" top-ups are recorded.
  • Screen kind. Each screen is set to walk-past or sit-down on its own page.
  • Counting on unsold screens. Players can count people all day, not only while a booked ad plays. This is built and switched off on the live site until we decide to turn it on.
  • New numbers. Prize cost per visit, counts of yes and no for consent, and an alarm when a camera stops counting.
  • Review board. Ads that need our decision now come first.

The sales person

A new desk, described in full in The Sales desk.

  • My day. Who is stuck today, and on what.
  • Customers. One list, and one page per customer that shows their whole path from first contact to a second booking.
  • Where a customer gets stuck. No login, an email never opened, no ad, an ad sent back, a price, an unpaid bill, a campaign that ended. Each step of the path says what the customer is stuck on and gives Sales the action that clears it.
  • Plans for a customer. Sales builds a plan with the customer, sends it as a link, and may book for them only after the customer has agreed.
  • Team view. For the sales manager.

The shopper and the shop counter

  • Bill amount. Counter staff may type the bill when they redeem a coupon. It is optional and never blocks the sale.
  • Shop door code. A code at the shop door. A scan counts as a visit, once per shop per day for each shopper.
  • Come-back coupon. Given in the wallet when a coupon is used at the counter. It works from the next day and counts as "came back".
  • "May this brand call you?" A separate yes for one brand at a time, with the exact words kept. The shopper can take it back. The advertiser gets a count and nothing else.
  • Call this shop now. After a yes, the shopper sees the shop's public number and a button to call. We count taps per campaign per day. We do not know who tapped.
  • Plain-scan coupons. A daily limit for each campaign. Only coupons with a checked phone are billed.

The new look

Every signed-in page moved to one shared look, drawn from three reference dashboards the founder picked.

  • A compact sidebar in groups, with counts beside the items that need work.
  • A top bar that stays in place, with the path to the page and a quick jump.
  • Cards for the main numbers, each with a change chip and a small chart.
  • Soft pill badges for status, and one shared tab row.
  • A "Needs you" list at the top of each home page.

The rule we kept: a new look must not hide anything the old page showed.

Where we moved away from the plan

The plan saidWhat we didWhy
Call me gives the brand phone numbers, each with a clear yesNo number ever leaves us. The shopper calls the shopA number handed to an advertiser can be passed on and used for scam calls. We cannot take it back
Count people on unsold screensBuilt, but off on the live siteIt is a decision for you, not for us
Top-up adds free playsWe work out what is owed and keep a ledger. Ops add the playsAdding plays by itself touches the loop. We wanted the arithmetic right first
A sampling report for brandsNot builtNo brand is waiting for it yet

The decisions the plan asked for

The first-principles doc ended with twelve decisions. The build had to assume an answer for each. These are the answers it assumed. Say so if any should change.

DecisionWhat the build assumes
The four reasons come before "pick a screen"Yes. The old way stays as a side door
The first buyers are shops and restaurants in the mallYes
A phone number goes to a brand when the person says yesNo. Stricter than the plan: a number never goes to a brand
Price stays per slot per day underneathYes. The buyer sees a plan price and a cost per result
The top-up promiseYes, on measured screens only, as a ledger
Any buyer may see a scorecard before payingYes
Count people on unsold screensBuilt, and off on the live site
The mall uploads its own promotionsYes, and ops approve each one
The mall sees the scorecard for its screensYes
Counter staff type the bill amountYes, as an optional step
A separate yes for one brand at a timeYes, with new words. The lawyer check is still open
A limit on plain-scan couponsYes, a daily limit per campaign. Only phone-checked coupons are billed

Open and waiting

For a lawyer

  • The words on "May this brand call you?" (version 2).
  • "We connect the call": a call through our number, so neither side sees the other's. Not built. It waits for this check.

For you to decide

  • Turn on counting for unsold screens on the live site, or leave it off.
  • Sales may confirm a customer's email by hand and book on their behalf after an "Agreed" note. Keep both, or narrow them.
  • Sales keeps a business owner's name and phone as plain contact details. They never go to logs or the audit trail. Say if you want them locked the way shopper phones are.
  • Top-ups stay a ledger for now. Say when free plays should be added without a person.

Still to do

  • Add the first ready-made ads on the live site. The library starts empty there.
  • Mark each live screen as walk-past or sit-down.
  • A page to change the shop's phone number after booking. Today it is set at booking.
  • A check of the Telugu and Hindi words by a native speaker.
  • The sampling report for brands.
  • The switch-off test.

How it was tested

  • 2,371 backend tests pass.
  • 249 browser tests, run after the last merge with main: 247 passed, 1 passed on its second try (a payment test that is sometimes slow), 1 is switched off on purpose.
  • Every page was opened and photographed at desktop and phone width.
  • The database changes were run many times on the local test database. They were not run on a copy of the live database before going live. They only add new tables and optional columns, and change nothing that exists.

For engineers

  • Nine database changes, in this order: rt_scorecard, rt_sales, rt_ops, rt_shopper, rt_mall, rt_plans, rt_results, rt_plans_flow, ui_call_shop. They run by themselves on deploy.
  • One new optional setting, AMBIENT_COUNTING. Unset means on everywhere except production.
  • New pages: /book/new, /screens, /screens/[id], /creatives/ready, /settings/business, /mall/promotions, /mall/approvals, /mall/shops, /mall/screens/[id], /join/shop/[token], /ops/sales and its sub-pages, /ops/plans, /ops/ready-ads, /ops/coupons/door, /outlet/door, /w/door/[token].
  • The API for all of it is in API contract.
What we changed: the November rethink · Docs