Product docs
Freebee Ads: what we sell, from first principles
Last updated Nov 13, 2026 · @praveen · 12 sections
Note, 16 November. This was written before the build. What we changed says what is built now and where the plan moved. One thing moved a lot: we never give a shopper's phone number to an advertiser. Where this page says "phone numbers" under Call me, read "the shopper calls the shop, and we count the calls".
Freebee should stop selling slots and start selling four results: being remembered, being tried, a visit, and a phone call. Our screens already record most of the proof for each one. Today the booking page shows an advertiser only slots and a price, and everything we measure is shown after the money is spent.
The business in six plain sentences
- People walk around a mall.
- We put screens where they walk and sit.
- A business pays us to show its ad on those screens.
- The business pays because it hopes some of those people will later buy from it.
- The business cannot see that hope come true, so it needs proof.
- We are the only ones standing at the screen who can count what happened.
So the thing we sell is not screen time. Screen time is the raw material. The thing we sell is a change in what people do, with proof.
An ordinary mall screen company sells sentence 3 and stops. It quotes one footfall number for the whole mall and gives a log that says the ad played. Our whole difference is sentences 5 and 6. Today we use that difference as a report at the end. It should be the product at the start.
The chain from money to a sale
Every rupee an advertiser pays has to travel through eight steps before it becomes a sale. People drop out at every step. We can prove the first six. Nobody else in a mall can prove more than the first.
| Step | In plain words | Can we prove it? | How exact |
|---|---|---|---|
| 1. Shown | The screen was on and my ad played | Yes | Exact. Every play is written down with its start time and length |
| 2. Near | People were close to the screen while it played | Yes, on screens with a camera | An estimate. Within 10 to 20% of a hand count |
| 3. Looked | Some of them turned their face to the screen | Yes, on screens with a camera | A floor. The real number is higher |
| 4. Did something | They played the game, tapped, or scanned the code | Yes | Exact |
| 5. Took it home | They got a coupon on their phone, or a product from the machine | Yes | Exact |
| 6. Came to the shop | They showed the coupon at the counter | Yes, when the shop enters the code | Exact, but it depends on the shop's staff |
| 7. Bought | They paid, and how much | No | We do not record the bill amount |
| 8. Came back | They became a regular, or remembered the brand | No | Not measured at all |
Three things follow from this table.
- Each advertiser cares about a different step. A drinks brand cares about step 3. A restaurant cares about step 6. A builder cares about a phone number, which sits between steps 5 and 6. Selling them all the same "slot" ignores this.
- The price of a result grows as you go down. A play costs a few rupees. A look costs more. A visit costs much more. A buyer for a flat can be worth lakhs. The deeper the step we can prove, the more we can charge.
- We should only charge per result where we do the counting. We count plays, looks, games, scans, phone numbers and machine drops. The shop counts step 6. If we billed per coupon used, a shop could simply not enter the codes.
Everything a screen records
This is the full list, read from the code, in ten groups. The last column says when an advertiser gets to see each thing: before paying, after paying, or never.
A. The screen itself
Ops type these in once. They do not change day to day.
| What | In plain words | Advertiser sees it |
|---|---|---|
| Mall, city, zone | Where the screen stands, such as Food Court | Before paying |
| Screen name | Its label | Before paying |
| Shape | Tall or wide | Before paying |
| Type | Normal loop screen, touch totem, half mirror, or holographic | Before paying |
| Size | Width and height in pixels | Before paying |
| Touch | Can people play on it with their hands | Before paying |
| Camera | Does it have a camera. Without one, groups D and part of E are empty | Not told clearly |
| Sound | On or off | Not told clearly |
| Opening hours | Daily hours, plus special days and late nights | Before paying |
| Loop length | How long before the whole set of ads repeats | Never |
| Slot length | How long one ad plays. 10 seconds on the demo screens, 15 on new ones | Before paying |
| Split of the loop | 40% paid ads, 30% Freebee content, 30% mall content | Never |
| Ad slots per loop | How many slots are for sale | Before paying |
| Price | Rupees per slot per day | Before paying |
| Auction | Whether slots on this screen can be bid for | Before paying |
| Counting switch | Whether people-counting is on for this screen | Never |
| Repair, off sale | Whether it is down or being removed | Before paying |
Four numbers come straight out of these facts. We never show them, and they are the first thing a buyer needs: how many times a day one slot plays, how many minutes a day that adds up to, how long the wait is between two plays of the same ad, and how many slots are already sold.
B. Ad plays
The player writes one line for every single play. Lines are never edited or deleted. This is exact.
| What | In plain words | Advertiser sees it |
|---|---|---|
| Screen and device | Which screen played it | After paying |
| Ad | Which picture, video or game | After paying |
| Booking | Which paid order it belongs to | After paying |
| Position | Which place in the loop | Never |
| Start time | To the second | After paying |
| Length | How long it really stayed on screen, in thousandths of a second | After paying |
| Pushed out | Whether an emergency message replaced it | After paying |
| Play number | A unique id, so no play is counted twice | Never |
| Late plays | Plays reported late because the screen lost its connection | After paying |
| Proof picture | A screenshot taken from the screen | After paying |
From these we work out plays delivered against plays promised, split by day, hour, screen and ad.
C. Screen health
The screen checks in every minute. This is exact.
| What | In plain words | Advertiser sees it |
|---|---|---|
| Minute state | Up, down, or switched off, for each minute of the day | After paying, as an uptime % |
| Plays in the minute | How many ads played that minute | Never |
| Outages | When a screen went down, when it came back | After paying |
| Data age | How fresh the numbers on the page are | After paying |
| Device health | Processor load, memory, disk, heat, battery, network | Never. Ops only |
| Camera state | Working, blocked, or missing | Never. Ops only |
| Counting mode | Counting bodies, counting faces only, or off, and how fast | Never. Ops only |
D. People near the screen
How it works, simply. The screen's own camera takes a small picture about two and a half times a second. A program inside the screen finds bodies and faces in that picture. It follows each body for as long as it stays in view. When a body has been gone for two seconds, it stops following. Only the counts leave the screen. No picture and no face is ever saved or sent.
| What | In plain words | How exact | Advertiser sees it |
|---|---|---|---|
| Walked past | Different people seen while this ad was on screen. Each counted once per play | Within 10 to 20% of a hand count, at 2 to 8 metres. Tight groups are undercounted | After paying |
| Looked | Faces turned to the screen for one second or more | A floor. A face must be within about 3 to 4 metres to be found | After paying |
| Looking time | All the seconds those faces stayed on the screen, added up | A floor, for the same reason | After paying, as an average |
| Age groups | How many lookers fell in each age band | A guess from the face. Only for people close to the screen | After paying. Hidden when under 10 people |
| Men and women | How many lookers of each | A guess from the face | After paying. Hidden when under 10 people |
| Looking right now | Live count of faces on the screen | Live | After paying |
From these we work out the attention rate (looked divided by walked past), the average looking time per play, and lookers by hour, day, screen and ad.
The counting runs in one of three modes. Bodies and faces is the full mode. Faces only is the fallback, and there "walked past" equals "looked". Off gives zeros, and the pages then show sample numbers marked as samples.
E. Games
This is exact. The screen counts its own game.
| What | In plain words | Advertiser sees it |
|---|---|---|
| Games started | How many times someone began playing | After paying |
| Games finished | How many played to the end | After paying |
| Scores | Total and best | After paying |
| How it started | By stepping in front of the camera, or by a tap | After paying |
| Code shown | How many times the phone code came up | After paying |
| Cricket game | Nickname and picture chosen, runs, balls faced, out or not out, milestones, prize step reached, start and finish time | After paying, as totals |
| Slot game | Spins used, score, best prize level, and for each spin its prize level and symbols | After paying, as totals |
From these we can work out four strong numbers: games per 100 people who walked past, how many finish, how long an average game lasts, and total minutes people spent playing with the brand. The last one is not shown today.
F. Phone scans
| What | In plain words | Advertiser sees it |
|---|---|---|
| Scan | One line each time someone scans the code: which ad, which screen, when, and their score | After paying |
A scan carries no name and no phone number. From it we work out scans per 100 times the code was shown.
G. Coupons
This is exact. Every coupon has its own line and its own history.
| What | In plain words | Advertiser sees it |
|---|---|---|
| The offer | Prize text, terms, expiry, which outlets accept it | Sets it up themselves |
| Prize levels | Up to three levels, each with its odds, a daily limit and a stock count | Sets it up themselves |
| Coupon given | Level, title, code, time given, expiry | After paying |
| How it travelled | Which route the person used to collect it | After paying |
| Delivered and opened | When it reached the phone, when it was opened | After paying |
| Used | When, and at which outlet | After paying |
| Code checks | Every time a counter looked up a code, even if it was not used | Never |
| Brand-reported uses | When a brand with its own codes tells us a code was used | After paying |
| Cost of each coupon | What one coupon costs the advertiser | After paying |
| Cancelled or reversed | Coupons taken back, and why | After paying |
| Stock left | Prizes remaining at each level | After paying |
From these we can build the full chain for one campaign: won, collected, opened, used. We can also show which outlet and which hour, and how long people wait before using a coupon. The page today shows totals, not this chain.
H. Shoppers
Only kept when the shopper says yes.
| What | In plain words | Advertiser sees it |
|---|---|---|
| Phone number | Stored scrambled and locked | Never |
| Yes or no answers | Four separate questions: camera for the game, being counted in ad numbers, giving a number for the coupon, and being 18 or over. We keep both yes and no, with the language and the screen | Never |
| Last seen | When this phone last came back | Never |
| Points | Balance and full history | Never |
We do not yet work out the things this group could tell us: new players against returning ones, how often one phone comes back in a month, and how many brands one shopper has played with.
I. Vending machine
| What | In plain words | Advertiser sees it |
|---|---|---|
| Drop request | Asked, confirmed, what the machine answered, how any problem was settled | After paying, as a count |
J. Selling
| What | In plain words | Advertiser sees it |
|---|---|---|
| Bookings | Screens, slots, dates, hours of the day, price | Their own |
| How full a screen is | Slots sold against slots for sale | Before paying, as free slots |
| Money booked | By screen and by advertiser | Never. Ops and mall only |
| Holds, price requests, bids | Who asked for what price | Their own |
| The advertiser | Business name, phone, GST number, address, city, and one of 14 ad categories | Their own |
One gap in this group matters a lot. We record an advertiser's ad category, but not what they want the ad to do. We never ask.
We also record seven anonymous events about how advertisers use the website, from sign-up to first approved booking. That is about us, not about screens.
What we cannot see
Thirteen things are missing. Five of them are cheap to fix, and those five matter most for selling.
| What we cannot see | Why | Can we fix it |
|---|---|---|
| How many different people | Someone in view across two plays is counted in both. Someone who steps away for two seconds and returns is counted again. We count "times seen", not "people" | Partly, and cheaply. Count each followed body once per hour per screen |
| Footfall at a screen, apart from any ad | People are only counted against the ad that was playing | Yes, cheaply. The same new count gives this |
| How long a walker stays in view | We keep looking time for faces, not time in view for bodies | Yes, cheaply. Send the average time each body was followed |
| Anything on a screen with no working camera | No camera, no people numbers | Add cameras, or say plainly "not measured" |
| People looking from far away | A face must be within about 3 to 4 metres | No. We say "at least" |
| Whether they understood or remember the ad | A camera cannot see a thought | Later. Ask one question on the phone after a game |
| Whether they heard the sound | Not measured | No |
| Who the person is | We chose not to know. This stays | No, by design |
| Where they went next | We lose them when they leave the camera | Yes, cheaply. A code at the shop door that they scan for points proves a visit |
| What they bought and the bill amount | The counter enters a code, not a bill | Yes, cheaply. Let the counter type the bill amount when it uses the code |
| Whether they would have come anyway | We have nothing to compare against | Later. Switch a campaign off on some days and compare |
| The mall's own door count | Not connected to us | Ask the mall for its daily number |
| Children, tight groups, shop dummies | Small or overlapping bodies are missed. Dummies and posters can be counted | Partly, by aiming the camera well |
One more unknown is about us, not the screens. I cannot log in to the live site, so I do not know how many live screens have a working camera or how much real data they have collected. Everything in the directions below depends on that.
The maths of one slot
One slot on a demo screen means 216 plays a day, 36 minutes of screen time, and a wait of 3 minutes 20 seconds between plays. Here is how that comes out.
- The mall is open 10:00 to 22:00. That is 12 hours, or 43,200 seconds.
- The loop is 200 seconds long, so it runs 216 times a day.
- A slot is 10 seconds, so the loop holds 20 slots. Eight of them are for paid ads.
- One slot plays once per loop: 216 plays, 2,160 seconds, 36 minutes. That is 5% of the screen's day.
The four demo screens are priced like this. These are the prices in our test data, not a live rate card.
| Demo screen | Price per slot per day | Cost of one play |
|---|---|---|
| Escalator Court, North | ₹1,500 | ₹6.94 |
| Escalator Court, South | ₹1,200 | ₹5.56 |
| Food Court, Entry | ₹1,000 | ₹4.63 |
| Food Court, Seating | ₹800 | ₹3.70 |
The same word "slot" means different things on different screens. A screen made with the system's standard settings has a 600 second loop and 15 second slots. One slot there is 72 plays a day, 18 minutes, with a 10 minute wait between plays. A buyer comparing two screens by slot price is comparing nothing.
Every cost an advertiser cares about is the same sum, one step further down the chain:
- Cost per play = price ÷ plays
- Cost per look = price ÷ looks
- Cost per game = price ÷ games started
- Cost per visit = (price + cost of the coupons) ÷ coupons used at the counter
- Cost per phone number = price ÷ numbers handed over
We hold every number on the right-hand side except phone numbers. We show none of these sums.
Now turn it round and ask what a buyer needs back. Take a ₹1,000 slot for one day. The earnings per result below are my guesses and need checking with real owners.
| Buyer | Roughly earns from one result | Results needed per day to cover ₹1,000 |
|---|---|---|
| Restaurant | ₹200 on one extra order | 5 extra orders |
| Shop | ₹500 on one extra sale | 2 extra sales |
| Coaching centre or builder | ₹500 or more for one good phone number | 2 numbers |
This is the question each buyer is silently asking: can this screen bring me five orders a day? We are the only ones who can answer it. If the honest answer for a screen is no, then the price or the offer is wrong for that buyer, and it is better we find out before they pay than after.
Two kinds of screen
A person walking past a screen has about a 1 in 11 chance of seeing any one ad. A person sitting near a screen sees every ad, several times. These are two different products, and today we sell them as one.
Here is the reasoning. Say a walker is in view for 8 seconds. That figure is my assumption. My ad is on for 10 seconds out of every 200. My ad and the walker overlap only if the walker arrives during my 10 seconds or in the 8 seconds before. That is 18 seconds out of 200, or 9%.
Now say someone sits in the food court for 20 minutes. That is 1,200 seconds, so the loop goes round six times in front of them. They see my ad six times.
| Walk-past screen | Sit-down screen | |
|---|---|---|
| Where | Escalators, entries, corridors | Food court seats, queues, lift lobbies, outside the cinema |
| Time in view | A few seconds | Minutes |
| Chance one person sees my ad, with one slot | About 1 in 11 | Certain, and several times |
| What I get | Many people, once, for a moment | Fewer people, many times, with time to act |
| What the ad must do | Work in two seconds. Big name, five words. A phone code is wasted here | Ask for something: play, scan, say yes to a call |
| What buying more slots does | Raises the chance. Two slots is about 1 in 6 | Mostly repeats the ad to the same people |
| Best for | Remember me | Try me, Come now, Call me |
This is the kind of reason you asked for. "Choose this bundle because people here sit for 20 minutes and see your ad six times" is a reason a shop owner understands.
We cannot yet sort screens into these two kinds from data, because we do not record time in view. Until we do, two numbers we already have point the same way: looking seconds per looker, and games started per 100 people who walked past. Someone can also just stand at each screen for ten minutes and write down which kind it is.
Why advertisers pay: four reasons
Every advertiser, whatever they sell, pays for one of four reasons. The reason decides which screens, which kind of ad, which number proves it worked, and what we can charge for.
| Reason | In the advertiser's words | Step in the chain | The one number that proves it | We count it ourselves |
|---|---|---|---|---|
| Remember me | "Make people know my name" | 3. Looked | Looks, and seconds of looking | Yes |
| Try me | "Let people play with it, taste it, install it" | 4 and 5. Did something, took it home | Games played, minutes played, samples dropped, scans | Yes |
| Come now | "Bring people to my counter today" | 6. Came to the shop | Coupons used at my counter | No. The shop's staff count it |
| Call me | "Give me people I can phone" | Between 5 and 6 | Phone numbers, given with a clear yes | Yes, once we build it |
Two corrections to what I told you last time.
- "Get people playing" is not a reason. Nobody wants a game for its own sake. A game is a tool. It earns looking time for Remember me, it hands over the sample for Try me, it gives out the coupon for Come now, and it is the moment we ask for a number for Call me. So the game belongs inside all four, not beside them.
- Asking "what kind of business are you" is not enough. A restaurant opening next month wants Remember me. The same restaurant on a slow Tuesday wants Come now. The reason can change from one booking to the next, so we ask it every time.
Eight kinds of advertiser
The rupee figures in this section are my guesses. They are here to show the shape of the thinking. Talking to real buyers will replace them.
| Kind | Main reason | How fast they decide | Fit with what we have today |
|---|---|---|---|
| 1. Restaurant, café or food stall in the mall | Come now | A day. The owner decides | Strongest |
| 2. Shop in the mall | Come now, in sale weeks | A week or more | Strong |
| 3. Cinema in the mall | Come now | Days | Good |
| 4. Film makers and distributors | Remember me | Weeks, through agencies | Weak |
| 5. Big national brands | Remember me, Try me | Months, through agencies | Weak for reach, strong for sampling |
| 6. High-price sellers: flats, cars, loans, insurance, coaching, clinics | Call me | Weeks | None yet. Biggest prize |
| 7. Local business outside the mall | Come now or Call me | Days | Fair |
| 8. Apps and online brands | Try me | Weeks | Fair |
1. Restaurant, café or food stall in the mall
- What they want: more orders in the quiet hours, today.
- What one result is worth: an order of about ₹200 to ₹800.
- What they fear: paying and seeing nobody come. People using the same coupon again and again. Extra work for staff at the counter.
- Why we fit: the buyer is already in the building, hungry, a short walk away. Every step of the chain up to the counter works today.
- The offer: "Fill my quiet hours." Food court screens, 3 pm to 6 pm, with a coupon that is good only today.
- The line that sells it: "People sitting at these tables see your ad about six times while they wait."
- What is missing: an easy way to buy by time of day. A ready-made ad, because most have none. A check that one phone cannot take the coupon many times. Cost per used coupon on the results page.
2. Shop in the mall
- What they want: walk-ins during a sale or a new arrival.
- What one result is worth: a bill of ₹1,000 to ₹20,000 or more.
- What they fear: looking cheap. Not knowing which walk-ins came from us.
- The offer: "Sale week." Entry and escalator screens so many people see it, plus one sit-down screen that gives a small gift for visiting the shop.
- The line that sells it: "Everyone who enters passes this screen, and your shop is forty steps from it."
- What is missing: a code at the shop door, so a visit counts even when nothing is bought. The floor and direction shown on the ad.
3. Cinema in the mall
- What they want: to fill seats for tonight. An empty seat is worth nothing after the show starts.
- What they fear: most tickets are bought on phone apps, so the counter coupon misses them.
- The offer: "Tonight's shows." From 4 pm, show times with free popcorn on a ticket bought today.
- What is missing: a way to count tickets bought on an app after seeing our screen.
4. Film makers and distributors
- What they want: everyone to know the film in release week.
- What they fear: nothing much. They buy on habit, through agencies, and want sound and long video.
- Why we fit badly today: our slot is 10 to 15 seconds. A trailer is 20 to 30 seconds or more, with sound.
- What is missing: two or three slots back to back. A list of screens with sound. A release-week pack across many screens.
5. Big national brands
- What they want: reach across many malls, or people trying the product.
- What they fear: numbers nobody outside has checked. They also pay late.
- Why reach fits badly: one mall is too small for them. This changes only with many malls.
- Why sampling fits well: nobody else can hand over a sample and prove who took it.
- The offer: "Sampling with proof." A game with the brand in it, a sample from the vending machine, and a report: games played, minutes played, samples given, age groups.
- How to sell it: a person sells it, not the website.
6. High-price sellers
Flats, cars, loans, insurance, coaching centres, colleges, clinics.
- What they want: names and phone numbers of interested people.
- What one result is worth: they already pay for every phone number, from online ads and from staffed stalls in malls. One good number can be worth ₹500 or far more.
- What they fear: junk numbers. Breaking the law on personal data.
- Why we do not fit today: we never give a phone number to an advertiser. That rule is right as it stands, because nobody has said yes to it.
- The offer: "Interested people." A short quiz or game. Then a clear question: "May this brand call you?" The box starts empty. The person ticks it. We check the phone with a one-time code. We ask one or two questions, such as budget or timing. Then we hand the lead to the brand.
- What is missing: the yes-per-brand step. The hand-over. A price per lead. A legal check before any of it.
7. Local business outside the mall
Gyms, salons, clinics, tuition centres, nearby restaurants, events.
- What they want: first-time visitors or enquiries.
- What they fear: wasting a small budget. They have no ad and no time to make one.
- The offer: a small fixed-price pack with a ready-made ad and a coupon or a call-back.
- What is missing: the ready-made ad. The door code, because their counter is not in the mall.
8. Apps and online brands
- What they want: installs and sign-ups.
- The number: scans, and cost per scan.
- What is missing: a link that tells them which screen each install came from.
The mall itself
The mall does not pay us, but it matters in three ways. It owns 30% of every loop. Its tenants are our first buyers. Its leasing team can add screen time when it signs a new shop.
What is wrong with the product today
Seven things, each one following from the sections above.
- We ask the buyer to do our job. They must pick screens, slots and dates. They cannot know which screen is good. We can.
- We hide our best thing until after they pay. I checked the booking page in the code. It shows slots and a price, and no people numbers at all. Every number about people appears only on the results page, after payment.
- One unit means different things. "One slot" is 216 plays a day on one screen and 72 on another. A walk-past screen and a sit-down screen are sold the same way, though they do different jobs.
- We never ask what the ad is for. So the results page shows every number we have and cannot say "it worked" or "it did not".
- The coupon chain is shown only as totals, and the phone-number product does not exist. These are the two results buyers value most.
- Most small buyers have no ad. "Upload a video" is a wall for them. We already have a tool that makes a game ad. We should lean on it.
- Sample numbers can mislead. On a screen with no camera we must say "not measured", not show numbers that look real.
Directions to take, in order
- Sell results. Keep slots as the engine underneath. The buyer tells us the reason, the budget and the dates. We show two or three plans. Each plan has an expected range, such as "4,000 to 6,000 looks", and one line that says why. "Choose screens myself" stays as a side door for buyers who want it.
- Build a scorecard for every screen first. Nothing above can be honest without it. We do not start from zero: our ops team already has most of these numbers on its own page for each screen. The work is to show them to a buyer before they pay, and to fill two holes. We have no count of people by hour. And a screen with nothing sold has no people numbers at all, because we only count while a booked ad plays. For the last 28 days, per screen: people per hour, how many in 100 looked, seconds of looking, games per 100 people, coupons used, best hours, age mix, walk-past or sit-down, camera or no camera. Add the two small counts we lack: people passing per hour, and how long a person stays in view.
- Start with shops and restaurants inside the mall. The whole chain works for them today. They decide in a day. They are inside the building. Their results become the stories we show everyone else. It also fits "one mall first".
- Build "Call me" second. It is likely worth the most per screen, because these buyers already pay per phone number. It needs the yes-per-brand design and a legal check before any code.
- For big brands, sell "Try me" by hand. Do not chase them for reach until there are many malls.
- Make the ad for them. A ready ad for each kind of business, done in two minutes.
- Make a top-up promise. If the looks we deliver fall below the low end of the range, we add free plays. Only on screens with a working camera.
- Later: prove it by switching off. Run the ad on some days and not on others, and compare the shop's sales. This is the only way to reach steps 7 and 8 of the chain.
Build order, when changes start
| Step | What gets built | Why it comes here |
|---|---|---|
| 1 | Scorecard per screen | Every later step reads from it |
| 2 | Booking that starts with the reason and shows plans | The buyer sees why before paying |
| 3 | A results page per reason: one headline number, then the chain | The buyer can tell if it worked |
| 4 | Phone numbers with a yes per brand | Opens the highest-value buyers |
| 5 | Sampling report | For brands, sold by a person |
The same plan, person by person
Six kinds of people use Freebee. Each one opens it with one question in mind. If a page does not answer that question, the page is not doing its job.
The parts before this one are written from the advertiser's side, because the advertiser pays. This part goes through all six people. Each person gets the same four things: what they want, what they see today, what is missing, and what to add. Each one also gets a picture.
| Person | Their one question | What they see today | The biggest gap |
|---|---|---|---|
| Advertiser | Did my ad work? | Screen names, slots and prices before paying. People numbers only after paying | The proof comes after the money |
| Mall operator | What did my screens earn, and which spots are good? | Totals for the whole mall, live status of each screen, a monthly statement. Read only | No numbers for each screen over time. Cannot put up its own content |
| Freebee ops | Are the screens up, sold and paid for? | Everything, including people numbers for each screen across all advertisers | The numbers exist, but nobody turns them into advice for a buyer |
| Shop counter staff | Is this coupon real? | Valid or not, the offer, today's count, the last 50 coupons | No bill amount. No weekly total |
| Shopper | What do I get, and is my number safe? | Their coupons, their points, and a promise that the number is never shared | Nothing asks whether a brand may call them |
| Sales person | "Who is stuck, and what do I do next?" | Separate pages for the waiting list, holds, prices, bookings and unpaid bills | No one page follows a customer from the first hello to a second booking |
Who sees which numbers today
The ten groups below are the same ten groups from "Everything a screen records". This table shows who can see each group today.
| Group of numbers | Advertiser, before paying | Advertiser, after paying | Mall operator | Freebee ops | Shop counter | Shopper |
|---|---|---|---|---|---|---|
| A. The screen itself | Name, place, price | Same | Its own screens | All | No | No |
| B. Ad plays | No | Its own | Total for the mall | All | No | No |
| C. Screen health | No | Uptime % and outages | On or off now, last check-in | All, with device and camera health | No | No |
| D. People near the screen | No | For its own ad, with age and gender estimates | Totals for the mall. For one screen, only people looking this minute | For each screen, across all advertisers. No age or gender | No | No |
| E. Games | No | Its own | Total for the mall | For each screen | No | Their own score |
| F. Phone scans | No | Its own | Total for the mall | All | No | No |
| G. Coupons | No | Its own, given and used | No | Full report | One coupon at a time, today's count, last 50 | Their own |
| H. Shoppers | No | No | No | Counts only. Never a phone number | No | Their own number and points. Can delete |
| I. Vending machine | No | A count | No | Stock and every drop | No | Their own PIN |
| J. Selling | Prices and free slots | Its own bookings and bills | Its share of the money | How full, who paid, who is owed | No | No |
Three things stand out.
- The column "before paying" is almost empty. That is the main fault in the product, and it is why Direction 2 comes so early.
- Ops already hold most of what a buyer needs. The people numbers for each screen are already on the ops page for that screen. The work is to show them to the buyer, not to invent them.
- Nobody outside Freebee sees what happens at the counter in rupees. We do not record the bill, so even ops cannot say what a coupon was worth.
The advertiser
What they want. One answer: did my ad work? "Work" means a different thing for each advertiser. A film wants to be remembered. A new drink wants to be tried. A restaurant wants people at the counter today. A builder wants people to phone.
What they see today. Before paying: a list of screens with names and prices, and which slots are free. After paying: plays, people who walked past, people who looked, games, scans and coupons, for their own ad.
What is missing.
- Nothing asks what they want the ad to do.
- Nothing before paying tells them how many people a screen reaches.
- They must bring a finished ad. Most small shops have none.
- After paying, every advertiser gets the same page of numbers. Nothing says which number is theirs.
What to add. Turn the path around. Ask the reason first. Show plans with expected results. Put the one number for that reason at the top of the results page.
The four pictures below show the pages that change the most. They are layouts only. Every number in them is made up.
Page 1. The start of a booking. Today it opens on a list of screens. It should open on one question.
Page 2. The plans. After the advertiser gives a reason, a budget and dates, we build two or three plans. Each plan says what to expect and gives one line that says why.
Page 3. The scorecard for one screen. This page is open to anyone, before paying. It is the proof behind every plan. Ops already have most of these numbers on their own page for the screen.
Page 4. The results. The top of the page is one number, in the advertiser's own words. The chain sits under it. The page ends with what to do next.
The mall operator
What they want. The mall gives us wall space and footfall. In return it wants three answers. How much did my screens earn? Which spots in my mall are good? Are the ads fit for my mall?
What they see today.
- Totals for the whole mall: plays, people who looked, and money.
- Each screen, live: on or off, which ad is playing now, and how many people are looking this minute.
- The bookings that are coming up.
- A monthly statement. It shows the mall's share of what advertisers paid. The share is 30 in 100 unless we agreed something else, and it never drops below the agreed minimum.
The mall can only look. It cannot change anything.
What is missing.
- Numbers for one screen over time. The mall sees a screen right now, but not last week, and not hour by hour. So it cannot tell a good spot from a bad one.
- Its own content. The mall owns 30 in 100 of every loop. But it cannot put anything there by itself. It must ask Freebee staff.
- Its own yes or no. Our Sales team records the mall's approval of an ad. The mall does not press the button.
- One money figure. The overview counts money that is booked. The statement counts money that is paid. The two never match, and nothing on the page says why.
- Its own shops. The mall cannot see how many coupons were used at the shops inside it.
- Sign-up. A mall cannot join by itself. We make every login by hand.
What to add.
- Give the mall the same scorecard for each screen that the advertiser gets.
- Let the mall upload and schedule its own promotions in its own share of the loop.
- Let the mall say yes or no to an ad inside its portal.
- Show both money figures side by side, each with a clear name: booked and paid.
- Give the mall's leasing team a simple way to offer screen time to its own shops. The mall already talks to every shop. That makes it our best sales partner.
Freebee ops
What they want. Our own team keeps the machine running. They want every screen on, every slot sold, every ad checked, and every rupee collected.
What they see today. Ops see more than anyone else. They already have a page for each screen with the people numbers on it: how many walked past, how many looked, and for how long. It covers all advertisers together. This is most of the scorecard a buyer needs. It is just not shown to buyers.
What is missing.
- People by hour. Ops see "looked" by hour. They do not see "walked past" by hour, so they cannot say what share of people look at 11 am against 7 pm.
- Unsold screens. We only count people while a booked ad is playing. A screen with nothing sold shows no people numbers at all. These are the screens we most need to sell, and we have no proof for them.
- Who the people are. The age and gender estimates are not on the ops pages.
- Yes and no counts. Ops cannot see how many shoppers said yes to giving a phone number and how many said no.
What the new model adds to each job.
- Set up a screen. Also mark it as walk-past or sit-down. The plans depend on this.
- Approve ads. Also keep a small set of ready-made ads for each kind of business, so a shop with no video can start the same day.
- Keep screens playing. A broken camera now matters as much as a dark screen. With no camera there is no scorecard, and with no scorecard there is nothing to sell.
- Fill the slots. Stop selling slots. Build two or three plans for each buyer from the scorecards.
- Collect and pay out. Where we count the result, bill by the result. Give free plays when the looks fall short of what we promised.
- Stock the prizes. Show what the prizes cost for each person who came to the counter.
The shopper and the shop counter
These two people never buy from us. But every result we sell passes through their hands. If the shopper does not play, or the counter does not redeem, there is nothing to report.
What the shopper wants. A bit of fun and a real prize, with no trouble. They do not want to be chased on the phone.
What happens today. The shopper sees a game or an offer. They play by typing a name only. To claim a prize they give a phone number, type a one-time code, and tick a box: 18 or older, save my coupons. They get a coupon or points. They show the coupon at the shop. Staff type the code, press Check, then press Redeem. The coupon works once.
The promise we made. When the shopper gives a number, our words say it is never shared with advertisers. This is a real limit. No number we hold today can ever be handed to a brand. So "Call me" needs new words and a separate yes, asked each time, for one brand only.
The limits that keep it fair. One coupon for each phone for each offer. Five prizes for each phone in a day. Five innings in a day. These limits work for coupons saved to a phone number. A coupon taken by a plain scan asks for no phone, so it has no limit for each person. One person can scan again and again. We should close this before we charge for coupons used.
What the counter sees today. One coupon at a time. Today's count. The last 50 coupons. After 10 wrong codes in 15 minutes the page locks.
What is missing.
- The bill. Staff do not type what the shopper spent. So we know a visit happened, but not what it was worth.
- Visits with no coupon. A shopper who walks in because of the ad, but has no coupon, is never counted.
- Coming back. We do not know if the shopper returned a second time.
- A weekly total for the shop. The counter sees today only.
What to add.
- A second question, for some ads. "May this brand call you?" The box starts empty. A yes goes to that brand only.
- A code at the shop door. The shopper scans it for points. It counts a visit even when nothing is bought.
- The bill amount. Staff may type it when they redeem. It is optional. It turns visits into rupees.
- A reason to come back. A second offer after the first visit shows who returns.
The sales person
Freebee has a sixth kind of user: its own sales person. The first customers are shops and restaurants inside the mall. The owner decides, often in a day, and most of them have no ad. A sales person is not a clerk who types bookings. Their job is to remove whatever the customer is stuck on.
What they want. To take a business from the first hello to a running ad, and then to a second booking. Each morning they have one question: who is stuck, and what do I do next?
What they see today. Freebee already has two sales roles, Sales and Sales Manager. They can approve ads, agree prices, manage bookings and holds, follow up unpaid bills and read the waiting list. Each of these is a separate page. In the table "Who sees which numbers today" they sit under Freebee ops, with fewer rights.
What is missing.
- A business does not exist in Freebee until it makes an account. A shop that Sales visited last week is nowhere.
- No page shows where each customer stands. Sales must open five pages to find out.
- No notes and no next-step date. Follow-ups depend on memory.
- No owner. Two people can chase the same shop. When a sales person leaves, their customers are lost.
- No safe way to help a customer who cannot use a computer.
What to add. A sales desk with four pages.
- My day. What needs me today: follow-ups that are due, customers stuck at a step, holds about to end, unpaid bills, ads that were sent back, and campaigns about to end.
- Customers. Prospects and signed-up customers in one list. Add a business, bring in the waiting list, filter by step and by owner.
- One customer. The path as a checklist, with one help action at each step. The steps are: prospect, account made, email confirmed, kind of business set, reason chosen, ad ready, ad approved, booked, paid, playing, first results shared, renewed. Under the checklist sit the contact log, the owner, and the campaigns with their numbers.
- Team. For the Sales Manager: each person's customers by step, the stuck ones, and a way to move customers from someone who has left.
The rule: do it with them, not as them. Sales prepares. The customer agrees and pays.
- Sales never logs in as the customer. There is no such button.
- The customer sets their own password from an invite. Sales never sees it.
- When a customer cannot use a computer, Sales does the typing. Before that, Sales records the customer's yes: what they agreed to, and how. The customer still pays.
- Every sales action is written to the audit trail.
The hard cases. Each one has a fixed answer, so two sales people handle it the same way.
| Case | What the desk does |
|---|---|
| The same business is added twice | The desk checks before saving. Same email or phone as another record: refused, with a link to the first one. Same email as an account: it offers to link the account. A similar name only: a warning. |
| The customer never confirms their email | Shown as stuck, with days waiting. Sales can send the link again, or confirm by hand after speaking to them. Both are recorded. |
| The customer has no ad | The desk suggests a ready-made ad for their kind of business. Nothing is uploaded in their name without them. |
| The customer cannot use a computer | Sales types for them after recording their yes. The customer pays. |
| Two sales people want the same shop | Everyone sees the whole list, so nobody chases a shop twice. One owner for each customer. Only the first claim works; the second person is told who has it. |
| A sales person leaves | Their customers show "owner has left". The Sales Manager moves them all in one step. Notes and next steps move with them. |
| A plan was sent, then the slots were sold | A hold keeps slots until its end date. A price without a hold does not. The desk lists "slots not kept" and holds ending within a day. |
| A price cut beyond the limit | The desk does not set prices. It links to the price pages, where the existing limits and approvals apply. A note is never a price promise. |
| The ad was sent back | The step shows the reviewer's reason. It appears on My day. |
| Payment is late or failed | Shown on My day after two days. Sales follows up. Nothing about money is changed on the desk. |
| Results fall below the promise | Sales sees the numbers first, and "not measured" where no camera counted. Free plays go through the top-up rule, not through a favour. |
| The customer wants to stop | Closed as "stopped", with a reason. Cancelling and refunds stay with the bookings page and its rules. |
| The customer goes quiet | No contact for 14 days and nothing playing: listed as "gone quiet". |
| Private details in notes | The note box says "business facts only". Phone numbers and emails typed into a note are removed before saving. |
| What Sales must never see | No shopper phone numbers and no wallet data. No change to money records or system settings. |
Decisions needed and things to find out
Decisions for you
- The four reasons replace "pick a screen" as the first question. I recommend yes.
- The first buyers we build for are shops and restaurants inside the mall. I recommend yes.
- A phone number goes to a brand only when the person says yes to that brand. I recommend yes.
- The price stays per slot per day underneath. The buyer sees a plan price and an expected cost per result. I recommend yes.
- The top-up promise: yes or no. I recommend yes, on measured screens only.
More decisions, from the person-by-person view
- Scorecard before paying. May any buyer see a screen's people numbers before they pay? The whole plan rests on a yes.
- Counting on unsold screens. May we count people all day, even when no booked ad is playing? Without this, an unsold screen has no proof.
- Mall content. May the mall upload its own promotions, or do we keep doing it for them?
- Mall numbers. May the mall see the scorecard for each of its screens?
- Bill amount. Do we ask counter staff to type the bill amount when they redeem a coupon? It is one more step for them.
- Call me wording. Do we add a separate yes for one brand at a time? This needs new words on the shopper's page, and a check by a lawyer.
- Plain-scan coupons. Do we put a limit on coupons taken by a plain scan, before we charge for coupons used?
Things I could not check
I cannot log in to the live site. Someone with an ops login has to read these.
- How many live screens have a working camera, and in which mode.
- The real loop length, slot length, opening hours and price of each live screen.
- Four weeks of real numbers per screen.
- Whether each live screen is walk-past or sit-down. Stand there for ten minutes and watch.
People to talk to
My rupee figures are guesses until these talks happen.
- Five shop or restaurant owners in the mall. What do they spend on ads now? What proof would make them pay again?
- Three buyers of phone numbers: a builder, a coaching centre, a car showroom. What do they pay per number today?
- One media agency. What numbers do they need, and who must check them?
Legal
- Check that we may share a phone number with a brand when the person has said yes. Add the answer to
docs/LEGAL_REVIEW.md.