Our work
Work we have built
These automations are built and running today at a family owned local business with a retail store, a service crew and an online shop that we know well. We built them before Open Afternoon had a name. Each card shows the problem, what we built and what the project did.
- 9paid orders nobody had seen, found and resent
- 14,000+past customers matched by one follow up rule
- 32 of 32days of sales a report matched to the cent
We leave out names and private figures. Every number comes from a count or a check on the project itself. We have not measured sales effects for most of these projects, and some are too new to count. A person stays in charge of anything that needs judgment.
Reports
Know where the business stands
Numbers that used to take four logins and a spreadsheet now arrive on their own.
-
A morning summary
The problemSales, online orders, web traffic and ad spend lived in four tools. The owner added them up by hand and compared them to last year.
What we builtOne private page, rebuilt every morning, with each number next to the same period last year. It only reads. It changes nothing in those tools.
What it didThe first full check compared 165 rows against their sources and found no mismatch. Each run repeats that check.
-
Store sales that match the register
The problemStaff printed four register reports a day and typed the net sales into a sheet.
What we builtA nightly job reads the day's sales from the point of sale system and writes one row per day. It then refreshes the forecast and item tabs.
What it didThe first check matched 32 of 32 days to the cent. Item sales now go back more than five years, and every month with a register figure ties to the cent. It also showed that the old sheet's yearly totals were off.
-
A price check between shop and website
The problemThe price on the website and the price at the register could drift apart, and nobody would see it.
What we builtA regular check compares every web price to the register price and lists the differences. It never changes a price. A person decides.
What it didThe first run scanned 2,697 product rows. 365 were priced in both places, and 35 of those differed. The owner reviews the 35.
-
Ad numbers you can trust
The problemThe ad dashboard reported a large conversion value. Almost all of it came from default values set years ago for page views and store visits, not from sales.
What we builtA scoreboard that counts only tracked sales and leads: calls, form sends and clicks to call. Modeled numbers show apart and never count as money.
What it didOver five months, tracked revenue was about one part in twenty of the value the dashboard reported. The scoreboard shows the tracked figure.
-
Analytics that count purchases
The problemOver 90 days, analytics recorded zero purchases across more than 100,000 visits while the store was taking orders.
What we builtWe found the cause, a duplicate purchase event that fired before the site's own, and shipped a fix.
What it didThe fix is live. In the first week after it, analytics recorded 25 purchases. Matching them to store orders is still open.
-
Real visitors, not junk traffic
The problemVisits climbed by more than 22,000 sessions in a year. The gain looked too good.
What we builtAn analysis that separated bots, other sites firing the analytics tag, and our own test visits from real people.
What it didMore than 18,000 of the added sessions were junk. Real visits rose from more than 82,000 to more than 86,000.
For a shop like yours. A plumber could get yesterday's jobs, open quotes and unpaid invoices in one morning message. An HVAC company could see which tune up customers are overdue. We would build the version that fits your tools.
Customers
One customer list, and follow ups that wait for you
The routine part runs by itself. The owner reads anything that goes to a customer before it goes.
-
One customer list
The problemCustomers lived in three places: the register, the email platform and the website. Nobody could say who was who.
What we builtA nightly job merges all three into one sheet. It keeps one row per person, combines duplicates and keeps the newest contact details. It also marks who can be texted, from a read only copy of the text inbox.
What it didMore than 78,000 records became more than 60,000 customers. More than 3,700 of them appeared in all three systems.
-
Email drafts to the right buyers
The problemEvery campaign email started with building an audience by hand.
What we builtA segment rule runs over purchase history, for example customers whose first purchase was over a year ago and who bought nothing from a matching group of products in the last 180 days. The tool keeps the ones on the email list and writes a draft campaign to them. It never sends. The owner reads the draft and presses send.
What it didOne rule across many products matched more than 14,000 customers, and more than 1,700 of them were on the email list. Only sales tied to a named customer count.
-
Survey answers become website reviews
The problemSomeone had to read every survey answer and copy the good ones onto the site.
What we builtEvery night the tool reads new answers to the survey sent after a finished job. Top scores with no written comment go to the reviews section on their own. Any comment waits for the owner's yes.
What it didThe reviews section updates itself each night, and a person still sees every comment first.
-
Forms that land in the right place
The problemLeads came in through a separate form tool, and analytics could not see them.
What we builtWe moved the forms onto the business's own site. Each counts as a tracked lead. The design request form writes all 17 of its answers into the service team's sheet.
What it didNothing is left on the old tool, and analytics now sees each form send.
-
Which inquiries become jobs
The problemThe owner could not tell why one service line trailed the rest of the business.
What we builtAn analysis that joins every inquiry to the customer list and the invoices. Tabs show the funnel, the trend and what jobs are worth.
What it didIt covers more than 2,000 inquiries over four years, so the path from inquiry to paid job is visible.
-
A customer lookup for staff
The problemFinding one customer meant opening several sheets.
What we builtA staff only page. Signed in staff search one merged record that updates every night.
What it didIt is live today.
For a shop like yours. A pest control company could draft a reminder to every customer due for a seasonal visit. A contractor could nudge every open quote. You read the draft first, every time.
Your website
Customers find answers on their own
Common questions answered on the site, and a search that finds what visitors came for.
-
A site search that ranks by relevance
The problemThe old search returned results in the order the pages were created. Customers ran more than 61,000 searches in 13 months.
What we builtA search that ranks by relevance, forgives typos and shows a ready answer for common questions. It has 27 answer cards, and its ranking was tested against 335 real queries.
What it didIt is the live site search today.
-
A product finder that rebuilds itself
The problemCustomers browsing thousands of products needed a way to narrow down to what fits their needs.
What we builtA finder with 16 filters over 2,201 items. It rebuilds from the live catalog every night, so new products appear without anyone editing it.
What it didWe also cut the first 48 cards on a phone from 18.6 MB to 3.0 MB, so it loads faster.
-
Quantity calculators
The problemOlder pages worked out quantities using a different package size than the one for sale.
What we builtCalculators on product pages that state one package size and count from it.
What it didThe count a customer sees now matches the package they buy.
-
A weekly check on AI answers
The problemThe owner had no view of how AI answer tools describe the business to local people.
What we builtEach week the tool puts 50 fixed local questions to three AI search tools. It records whether the business is named and cited. It approximates the consumer apps and never writes to the site.
What it didIn a recent week, the business was named in 108 of 140 answers. The owner now knows which questions it misses.
For a shop like yours. An electrician could answer "do you serve this town" and "what does a panel upgrade involve" without a phone call. We write the answers with you and check them.
Back office
Back office work that runs while you work
The jobs nobody has time for, done on a schedule, with a person deciding anything that matters.
-
One catalog from many supplier lists
The problemStaff opened a dozen supplier lists, one at a time, to answer one customer question.
What we builtEach morning the tool collects price and stock lists from websites, PDFs and emailed files. It merges them into one searchable page for staff, with the shop's price or an estimate.
What it did22 supplier lists, more than 40,000 price and stock lines, one search.
-
An order email check
The problemA customer typed a wrong email address on a paid order. The confirmation bounced and nobody knew for two days.
What we builtA daily scan of recent orders for mistyped addresses and bounced order emails. It flags them for a person to fix.
What it didOn 649 real orders it flagged 3 bad domains and made no false flag among 42 domains.
-
Order emails that went missing, found and fixed
The problemThe web host's mail filter dropped new order emails to staff for a week. The store still said they were sent.
What we builtWe proved the cause with a controlled test, added a code fix and resent the lost messages.
What it didNine paid orders had gone unseen over one week. All nine were resent, and staff confirmed.
-
Dead links sent to working pages
The problemThousands of old links led visitors to an error page, and one old file alone drew over a thousand logged hits to nothing.
What we builtA one time cleanup, then a weekly watcher. It confirms each broken link is still dead and redirects the ones with one clear destination. It lists the rest for a person.
What it didThe cleanup sent more than 1,900 broken links to nearly 600 working pages, covering more than 8,000 recorded visits. The watcher's first run fixed more than 700 others, and it caught and fixed one wrong redirect.
-
Sales that end on their own
The problemSomeone had to remember to take each sale off the site.
What we builtWhen the owner approves a sale, it is set up with an end date. A scheduled job then removes it on that date, loads the public page logged out to check it and reports.
What it didSales are scheduled to end on their own when set up.
-
Social posts copied to the business listing
The problemThe same update went up in two places by hand.
What we builtA daily job copies new social posts to the business listing. We rebuilt it after one stuck post made it repeat itself for 27 days.
What it didIt runs daily and records every failure. It also alerts a person if it breaks.
-
A weekly video becomes short clips
The problemCutting, titling and scheduling clips from a weekly live video was manual work.
What we builtEach week the tool finds the new video, transcribes it and picks the best moments. It cuts four to six short clips per episode and schedules one a day.
What it didA finished episode turns into scheduled clips on its own.
-
Staff photo upload
The problemStaff photos had no route to the site.
What we builtSigned in staff upload photos. A daily job adds approved ones to the right product gallery, up to four photos on up to 30 products a run, and never touches the main image.
What it didStaff photos now have a route to the site.
-
Event tickets on the business's own site
The problemEvents sold through a third party that charged a fee.
What we builtAn events system on the site. Tickets are store products, stock is the room's capacity and registration takes one click.
What it didAll 19 events moved over, six sell tickets on the site, and the business stopped paying the ticket fee.
For a shop like yours. A garage door company could get a flag when a quote email bounces. A builder could have every supplier quote in one place. We start with the one that costs you the most time.
Still in testing
A product advisor chatbot
We are testing a chatbot that answers product questions from the business's own site. Code checks every claim and link before it prints. Only staff can use it today. It is not public, and we do not count it as finished work.
See how we would build one for you on website chatbots.
Questions about this work
Why does the business have no name here?
We leave out the name and the private figures so nothing here can be traced to the business or its customers. The projects are real. We show only counts and checks that come from the projects themselves.
Is this a list of clients?
No. Every project here is for one business we know well. This page says nothing about how many projects we have done, and it is not a promise of the same result for your business.
Can you build something like this for my business?
Often, yes. The free call is where we look at your own work and say what we would build first. Some projects here depend on tools your business may not use. We say so on the call.
Updated October 2026