Examples by industry
ZeroWidth adapts to how your business works and finds the decisions that moved your numbers the right way.
See where your team's effort goes, and which work moves your product numbers.
A software team decides in Claude Code sessions, pull requests, tickets and Slack threads. ZeroWidth reads those decisions, groups them by the part of the product each one touched, and checks each one against numbers like signups, activation and revenue. Before the next roadmap call, the team and the team's agents can see which kinds of work paid off.
- Decisions from
- Claude Code, ChatGPT and Claude, GitHub, Linear or Jira, and Slack.
- Checked against
- Signups, activation, retention and revenue, from Stripe or your own product data.
- What the team sees
- Where the work went this quarter, and which work moved a number.
Acme's decisions
“Let's move the weekly demo to Thursdays so sales can bring prospects.”Priya, in Claude
Demos booked
Worked. Demos booked rose 18%.
“Offer annual plans with two months free.”Marcus, in ChatGPT
The share of new customers paying yearly
Worked. Annual plans went from 12% to 31% of new customers.
“No feature request gets built without a call with the customer first.”Dana, in Claude
Customers using new features
Mixed. Use of new features rose 30%, but fewer features shipped each month.
“Build a self-serve plan for teams under ten people.”Leo, in Claude
New customers each month
No clear change. New customers rose 4%, within their usual ups and downs.
“Every new engineer ships one small fix in the first week.”Ava, in Claude
Weeks until a new engineer ships on their own
Worked. New engineers shipped on their own in four weeks instead of seven.
“Stop paid search ads on our own brand name.”Theo, in ChatGPT
Ad spend, without losing signups
Mixed. Ad spend fell $6,000 a month, but signups from search fell 8%.
“Move error alerts into one shared channel instead of email.”Kenji, in Slack
Time to fix errors customers see
Worked. Average time to fix fell from nine hours to three.
“Call every customer who cancels within the first month.”Sofia, in Claude
Cancellations in the first month
Didn't work. Cancellations rose from 9% to 12%. Most customers didn't pick up.
Connect marketing, pricing and store decisions to what customers bought.
A store's calls are spread across campaign plans written with AI tools, Slack threads with the warehouse and emails with suppliers. ZeroWidth brings those decisions into one record and checks each one against the numbers from your store, your CRM and your ad accounts. A promotion, a shipping change and a loyalty rule each show what they did to sales, side by side.
- Decisions from
- ChatGPT and Claude, Slack, email and meetings.
- Checked against
- Orders, average order, repeat customers and the cost of returns.
- What the team sees
- Which campaigns and policies raised sales, and which only raised costs.
Acme's decisions
“Send the restock email on Tuesday mornings instead of Fridays.”Sam, in ChatGPT
Sales from restock emails
Worked. Sales from restock emails rose 22%.
“Free shipping over $50 instead of $75.”Aisha, in Claude
Orders and average order size
Mixed. Orders rose 9%, but the average order fell 6%.
“Pack same-day orders on a separate line after 2pm.”Tom, in Slack
Orders shipped on time
Worked. On-time shipping went from 88% to 96%.
“Open a second warehouse on the East Coast.”Mei, in ChatGPT
Delivery time and shipping cost
Worked. Delivery time fell from five days to three.
“Give store staff a bonus for every online order picked up in store.”Jordan, in ChatGPT
In-store pickup orders
No clear change. Pickup orders moved, but within their usual ups and downs.
“Put real customer photos on product pages.”Lena, in Claude
Product page visits that end in a sale
Worked. Sales from product pages rose from 2.1% of visits to 2.6%.
“Switch returns to a printable label instead of a form.”Diego, in Slack
Returns handled per week, and complaints about returns
Mixed. Complaints fell by half, but returns rose 5%.
“Let loyalty members return items for 60 days instead of 30.”Hana, in ChatGPT
Repeat orders from loyalty members
Didn't work. Repeat orders didn't move, and returns cost $4,000 more a month.
Learn which changes to pricing and staffing raise your margin.
A firm decides how to price, staff and run client work in partner meetings, email threads and documents. ZeroWidth reads those decisions and checks each one against proposals won, hours billed and profit per project. Partners see which policies paid off before setting the next one, and an agent drafting a proposal sees what happened the last time fees changed.
- Decisions from
- Meetings, email, Google Docs and Word, Slack, and ChatGPT and Claude.
- Checked against
- Proposals won, hours billed, profit per project and people staying.
- What the team sees
- Which ways of pricing and staffing paid off, by kind of client.
Acme's decisions
“Let's pilot the 4-day turnaround with our top 10 accounts first.”Jamie, in ChatGPT
Repeat work from those clients
Worked. Repeat work from those accounts rose 15%.
“Quote fixed fees for audits instead of hourly rates.”Rosa, in Claude
Proposals won
Worked. The win rate went from 30% to 41%.
“Use one shared template for every status report.”Ben, in Slack
Hours spent on reporting
Didn't work. Hours on reporting rose from 6 a week to 8. Teams filled in the template and kept their old formats.
“Hire two senior people instead of four juniors.”Nora, in Claude
Profit on client work
Mixed. Margin rose 4 points, but two projects slipped during hiring.
“Pair every new consultant with a partner for their first two projects.”Chris, in Claude
New consultants staying past one year
Worked. New consultants staying past one year went from 70% to 85%.
“Publish one short case study a month instead of a newsletter.”Maya, in ChatGPT
Inbound requests for proposals
No clear change. Requests for proposals went from six a quarter to eight, too few to tell.
“Track time daily in the project tool instead of weekly in a spreadsheet.”Sean, in Slack
Hours billed
Mixed. Hours billed rose 7%, but people spent an extra hour a week entering time.
“Stop taking projects under $10,000.”Priti, in Claude
Profit per project
Mixed. Profit per project rose 18%, but total profit fell 5% with fewer projects.
See which changes to scheduling and the front desk help patients and the budget.
A clinic's managers decide in staff meetings, Slack and their AI tools: when to send reminders, how to book follow-ups, which clinics stay open late. ZeroWidth checks each change against missed appointments, wait times and cost per patient, across every clinic. The numbers are counts, like missed appointments a week, not patient records.
- Decisions from
- Meetings, Slack, email, and ChatGPT and Claude.
- Checked against
- Missed appointments, wait times, follow-ups kept and cost per patient.
- What the team sees
- Which changes worked at one clinic, before rolling them out to the rest.
Acme's decisions
“Text appointment reminders 48 hours ahead, not 24.”Luis, in ChatGPT
Missed appointments
Worked. Missed appointments fell from 11% to 7%.
“Move the lab contract to a price per test.”Grace, in Claude
Lab cost per patient
Worked. Lab cost per patient fell 12%.
“Put intake forms online before the visit.”Omar, in Slack
Time in the waiting room
Worked. The average wait fell from 22 minutes to 14.
“Open evening hours at three clinics.”Hannah, in ChatGPT
New patients
No clear change. New patients moved, but within their usual ups and downs. Evening slots went to existing patients.
“Let front-desk staff book follow-ups before the patient leaves.”Daniel, in ChatGPT
Follow-up visits kept
Worked. Follow-up visits kept went from 64% to 81%.
“Ask patients for a review by text the day after a visit.”Yara, in Claude
New patients from search
Mixed. New patients from search rose 10%, but twice as many patients turned off text messages.
“Use one scheduling system across all five clinics.”Felix, in Slack
Open slots left unfilled
Worked. Unfilled slots fell from 14% to 8%.
“Offer video visits for medication check-ins.”Amara, in ChatGPT
Time doctors spend per check-in
Didn't work. Time per check-in rose from 15 minutes to 19. Some patients still came in after the call.
Tie changes on the floor and in the supply chain to output, defects and cost.
A plant's decisions are made at shift meetings, in supplier emails and in plans written with AI tools. ZeroWidth reads those decisions and checks each one against units per shift, defects, downtime and cost per unit, compared with how those numbers usually swing from week to week. The next line to try the same change sees the result first.
- Decisions from
- Meetings, email, Slack, documents, and ChatGPT and Claude.
- Checked against
- Units per shift, defects, unplanned downtime and cost per unit.
- What the team sees
- Which changes held up on the floor, and what each one cost.
Acme's decisions
“Start the line-two shift 30 minutes earlier.”Kofi, in ChatGPT
Units per shift
Worked. Units per shift rose 6%.
“Buy 30% of our steel from a second supplier.”Elena, in Claude
Late deliveries and material cost
Mixed. Late deliveries halved, but material cost rose 8%.
“Add a quality check after welding, not only at the end.”Raj, in Slack
Defects
Worked. Defects fell from 2.4% to 1.1%.
“Automate the packing line.”Ines, in Claude
Cost per unit
Didn't work. Cost per unit rose 6% after six months, with downtime during the changeover.
“Train every operator on two machines, not one.”Marta, in Claude
Hours a line stands idle for lack of staff
Worked. Idle hours fell from 30 a month to 8.
“Show at two trade fairs this year instead of five.”Paolo, in ChatGPT
New distributors signed, and marketing spend
Mixed. Spend fell 40%, but new distributors fell from nine to six.
“Put sensors on the three oldest presses to warn before a breakdown.”Yuki, in Slack
Unplanned downtime
Worked. Unplanned downtime fell 45%.
“Ask large customers for a three-month order forecast.”Sipho, in Claude
Rush orders and overtime
No clear change. Overtime moved, but within the usual ups and downs. Forecasts came in for two customers of twelve.
Put your team's decisions to use.
Have a signup code? Start with your code