Case study · Barbarians Barbershop · Barbershop
From 18% to 42%
of customers recovered
And average ticket from $35 to $50, on the same WhatsApp number with the same team.
No new ad budget, nobody hired, and no change to the volume of conversations the shop already received.
The definition first
What counts as a recovered customer
Out of every 100 people who messaged on WhatsApp and did not book right away, 18 came back and booked later.
The same 100 people, with an instant reply, the full menu in the agent's hands, and a follow-up whenever something was left open.
The difference did not come from new advertising or from more people answering. It is the same volume of conversations the shop already had, handled without gaps. Twenty-four percentage points.
Where the other 82% leaked
Four holes, all in the same place: WhatsApp
It arrived after hours
The 9pm message got answered at 10am the next day. The urge to book does not wait for the shop to open.
It waited through the rush
Busy shop, every chair full, nobody on the phone. Replying in minutes converts far better than replying in half an hour.
They asked for a service nobody could quote fast
With a full menu and the phone ringing mid-cut, quoting a service, a price or an open slot from memory is slow, and getting it wrong costs the booking.
Someone promised to check and never came back
A Saturday slot, a price for the combo, a product in stock. The promise disappeared and the customer disappeared with it.
None of these holes was the team not trying. They are the four you can close without hiring anyone.
Hole by hole
What actually changed
| The hole | How it was | With the Maze agent |
|---|---|---|
| Hours | Answered when the shop opened | Answers in seconds, 24 hours a day, Sundays included |
| Peak-time delay | Messages queued while the shop worked the chairs | Serves everyone at once, no queue |
| Menu & prices | Quoting a service or product from memory or the wall | Searches the whole menu and only quotes what actually exists |
| Something they don't offer | Risk of promising a service or product the shop doesn't do | Says they don't offer that and suggests the closest option |
| Half-finished bookings | Confirmed the service and the conversation died there | Asks for the day, time and any add-on, and closes with the total added up |
| Forgotten promises | Said they'd check and nobody remembered | Alerts the owner, with what was promised and to whom |
Every line in the right-hand column is live and was tested in real conversations.
The number almost nobody looks at
Average ticket went from $35 to $50
per visit
per visit
Why it went up
They used to book one cut and get one cut. Now they get the cut, the beard trim, and the product that pairs with it, with the booking closed and the total added up.
Shows what's on the menu, not what it remembers
The search sweeps the entire menu, so the add-on or the product that usually gets forgotten makes it into the conversation.
Doesn't leave the booking half done
Confirming the service doesn't end the chat. It asks for the day, the time and any add-on, and adds up the total.
Arrives while they're still deciding
A reply in seconds catches the customer mid-decision, which is when they'll add the beard trim or grab the pomade.
Where 42% sits
Above the market average, closing on tuned automation
Average WhatsApp conversion
General average for service businesses, human agents included.
Well-tuned automation
The range for teams running a tuned agent. That is where there is still room to climb.
Replying in under 5 minutes
Converts 21 times better than replying after 30 minutes. The agent replies in seconds.
42% clears the top of the market average and is closing in on the range of tuned automation. The result is strong, and it is not an outlier.
Market ranges: Kubalabs, WhatsApp conversion benchmark 2026; EchoLeads, automation benchmark 2026.
The math
Per 100 conversations
18 customers × $35
$63042 customers × $50
$2,100revenue that was not there
$1,4702.3 times more recovered customers, each one spending 1.4 times more. The two effects multiply, which is why the revenue jump is much larger than the jump in percentage points.
Same number of conversations, same team, same ad budget.
Recovered revenue per month
Find your shop's row
| Conversations a month | Before (18% × $35) | After (42% × $50) | Difference |
|---|---|---|---|
| 200 | $1,260 | $4,200 | $2,940 |
| 500 | $3,150 | $10,500 | $7,350 |
| 1,000 | $6,300 | $21,000 | $14,700 |
The break-even point
At $39 a month, a single recovered customer at $50 more than covers the whole month. Everything past that is new profit.
One shop or the whole network
The table is per shop. In a chain, multiply by the number of locations answering on WhatsApp.
Formula: conversations × recovery rate × average ticket, at both points in time. All figures in US dollars ($).
Where the numbers come from
It all shows up in the dashboard, week by week
Conversations and replies
How many people messaged and how many were answered. The target here is 100%, and it is verifiable.
Time to first reply
The single factor that moves conversion the most. Logged conversation by conversation.
Closed bookings and value
Chats marked as won. This is where the recovery rate and the average ticket come from.
Owner alerts
A service the customer asked for that the shop does not offer, and a promise the agent made that somebody has to keep.
Both numbers come from the same place
The 18% and the 42% were read with the same criteria, and it is the same criteria that keeps running. Neither one is a presentation estimate.
How it was built, and how it repeats
Four weeks, start to first number
Live in an afternoon
Account created, menu imported from a spreadsheet, number connected. The agent replies the same day.
Baseline and tuning
Read the real conversations and fix anything the shop would have answered differently. This is where the agent learns the shop's style.
Closing the booking
The full booking over WhatsApp: day, time, add-ons and total added up. This is the step that moves the ticket.
The first number
Compared against the baseline, same criteria, and a decision on what to tune next.
What you need to start
A menu with services, prices and any products; the answers your shop repeats most on WhatsApp (hours, prices, which barber, walk-in versus booking, payment); and the number that will do the answering.
The part nobody puts on a slide
What Maze does not do
It does not track stock
The menu is a snapshot. If a product runs out or a price changes, the agent keeps quoting the old one until the spreadsheet is updated.
It does not know what is not written down
Hours, prices and policies only exist for it if they are in the setup. Whatever is missing, it says so instead of inventing.
It does not replace your barber
It answers, shows the menu, builds the booking and calls a human when the case needs one. The relationship still belongs to the shop.
See it answering your own customers first
The trial is free and never asks for a card. Upload the spreadsheet, connect the number, and watch the agent work your real conversations before you decide anything.