Problem Discovery, ROI Baselines and Go/No-Go Analysis
You can turn a weekly pile of repeated tasks into a baseline cost in SEK, set an error-rate ceiling from the cost of one mistake, and mark a spreadsheet Go and a document add-on No-Go from the year-1 net.
- Level
- Navigator
- Time
- 18 h

Jump to
After this course you can
The student can calculate baseline workflow costs, error-rate thresholds, and break-even ROI in SEK across 15 supplied business case dossiers and correctly classify which cases should be rejected for AI.
A demo and a counted bottleneck
A demo is one reply. You paste a single order confirmation into the chat service, and the reply names a delivery week. That shows a sentence can be produced from one document. The bottleneck is the pile that returns every week: how many confirmations, how many minutes each one takes, what one hour of pay costs, and how often the typed week is wrong.
At the parts counter in this case, the clerk handles 100 confirmations a week and spends 3 minutes on each at 400 SEK an hour. One confirmation costs 3/60 × 400 = 20 SEK of labor. The week costs 100 × 20 = 2,000 SEK of labor, which is 5 hours. A strong reply on one confirmation leaves that 2,000 SEK uncounted. Write four numbers before you compare tools: the weekly count, the minutes, the hourly rate, and the size of the mistake sample. A missing number keeps the case a demo.
Hours times the hourly rate
Weekly hours = count × minutes / 60. Weekly labor in SEK = weekly hours × the hourly rate. Here 100 × 3 / 60 = 5 hours, and 5 × 400 = 2,000 SEK. Across 50 working weeks the labor baseline is 2,000 × 50 = 100,000 SEK. Keep the week count written beside the product, because a case with 46 weeks would change the year.
Then split the pile by what is already stored. 80 confirmations already hold an order id, a delivery week, and a customer id in the shop system. A spreadsheet formula can copy those three fields. The other 20 arrive as email with no order id, so the formula has nothing to match, and the clerk keeps them. Those 20 cost 20 × 20 = 400 SEK a week, and 400 × 50 = 20,000 SEK a year. Labor the copy removes is 100,000 − 20,000 = 80,000 SEK a year.
Building the sheet takes 5 hours at the same 400 SEK, so setup is 2,000 SEK. Setup sits on its own line. The sheet has no license, so its annual running cost is 0. Weekly labor after the sheet is 400 SEK, and the 2,000 SEK setup is earned back from the weekly saving, which the later arithmetic divides out.
The error rate and the one-tenth ceiling
The baseline error rate = mistakes in the sample / sample size. A second person checked 40 retyped confirmations and found 4 wrong delivery weeks. The rate is 4/40 = 0.10, or 10%. On a week of 100 retyped confirmations that is 0.10 × 100 = 10 mistakes. Each mistake is a phone call of 30 minutes: 30/60 × 400 = 200 SEK. Weekly error cost is 10 × 200 = 2,000 SEK. Annual error cost is 2,000 × 50 = 100,000 SEK.
Weekly baseline = labor + error cost = 2,000 + 2,000 = 4,000 SEK. Annual baseline = 4,000 × 50 = 200,000 SEK. The sample was taken while every confirmation was still retyped. The 20 emails stay retyped, so this case keeps the same 10% on them until a separate email sample replaces it. Then 0.10 × 20 = 2 mistakes a week, 2 × 200 = 400 SEK a week, and 400 × 50 = 20,000 SEK a year. Error cost the copy removes is 100,000 − 20,000 = 80,000 SEK a year. The 80 stored rows pick up no new typo from a formula, because the week in the sheet is the same cell the shop system already stores. That copy sits at 0 new mistakes.
The one-tenth rule in this case sets the ceiling for a row that nobody checks. The expected cost of a new mistake on one row may be at most one tenth of the labor on that row. Labor is 20 SEK, so the allowance is 2 SEK. A mistake costs 200 SEK, so the ceiling is 2/200 = 0.01, or 1%. The formula copy is under that ceiling. A demo that brings no sample has no rate to place under it, so the rows it would touch stay with a person until a sample is counted.
Price the sheet first, then Go or No-Go
Annual saving from the sheet = labor saved + error cost saved = 80,000 + 80,000 = 160,000 SEK. Running cost is 0 and setup is 2,000, so the year-1 net is 160,000 − 0 − 2,000 = 158,000 SEK. The year-1 net is the ROI figure used here, in SEK: annual saving minus annual running cost minus one-time setup. Weekly saving is 160,000 / 50 = 3,200 SEK. Break-even is setup divided by weekly saving: 2,000 / 3,200 = 0.625 weeks.
The document add-on in the quote is 2,400 SEK a month, so 2,400 × 12 = 28,800 SEK a year, or 28,800 / 50 = 576 SEK a week. The quote still has a person check each confirmation for 1.5 minutes, because a wrong week means a late delivery. On all 100 rows, 100 × 1.5 = 150 minutes = 2.5 hours, and 2.5 × 400 = 1,000 SEK a week, which is 50,000 SEK a year. The quote treats that check as the moment corrections happen, so the remaining error cost is 0. Connecting the export takes 4 hours at 800 SEK, a setup of 3,200 SEK. Year-1 cost on all 100 rows is 50,000 + 28,800 + 3,200 = 82,000 SEK. Year-1 net against the 200,000 baseline is 200,000 − 82,000 = 118,000 SEK.
On the 20 emails alone, with the sheet already covering the 80, the check is 20 × 1.5 = 30 minutes = 0.5 hours × 400 = 200 SEK a week, or 10,000 SEK a year. The clerk costs 20,000 SEK a year on those emails, so labor saved is 10,000. Error cost saved is the 20,000 above. Gross gain is 10,000 + 20,000 = 30,000 SEK. The license stays 28,800 because the quote is a flat year price, and the setup stays 3,200. Year-1 net of the add-on on those 20 emails is 30,000 − 28,800 − 3,200 = −2,000 SEK.
A Go needs all four checks. The weekly cost is written as hours × SEK and as mistakes × SEK, with the sample size beside the rate. The rows that already have fields are priced as a formula before any add-on is priced, and the path with the higher year-1 net is the one you keep. The year-1 net is above zero. Unattended rows are allowed only at or under the written ceiling, from a sample. The sheet meets all four: the counts are written, the 80 rows are the formula path, 158,000 is above zero, and the copy adds 0 new mistakes under the 1% ceiling, while the 20 emails stay with the clerk. The add-on on all 100 rows has a year-1 net of 118,000, which is above zero and still below the sheet's 158,000, so the sheet is the path you keep. The add-on on the 20 emails has a year-1 net of −2,000. The add-on is No-Go. The sheet is Go.
The parts counter, counted through
Inputs for the parts counter: 100 confirmations a week, 3 minutes each, 400 SEK an hour, 50 working weeks. A second person found 4 wrong delivery weeks in 40 retyped confirmations. One wrong week costs a 30-minute call. 80 rows already store order id, delivery week, and customer id. 20 rows are emails. The sheet takes 5 hours at 400 SEK. The add-on quote is 2,400 SEK a month, plus 1.5 minutes of checking on each row it touches, plus 4 hours at 800 SEK to connect the export. The quote sets the remaining error cost to 0 once that check is in place.
Labor per confirmation is 3/60 × 400 = 20 SEK. Weekly labor is 100 × 20 = 2,000 SEK, from 5 hours. Annual labor is 2,000 × 50 = 100,000 SEK. The error rate is 4/40 = 0.10. Weekly mistakes are 10, weekly error cost is 10 × 200 = 2,000 SEK, and annual error cost is 100,000 SEK. The weekly baseline is 4,000 SEK and the annual baseline is 200,000 SEK. The unattended allowance is 20/10 = 2 SEK, and the ceiling is 2/200 = 0.01, or 1%.
After the sheet, 20 emails remain at 20 SEK each: 400 SEK a week and 20,000 SEK a year of labor. At the same 10% retyping rate those emails produce 2 mistakes a week, 400 SEK a week, and 20,000 SEK a year. Labor saved is 80,000 SEK and error cost saved is 80,000 SEK, so annual saving is 160,000 SEK. Setup is 5 × 400 = 2,000 SEK and running cost is 0. The year-1 net is 158,000 SEK. Weekly saving is 3,200 SEK. Break-even is 2,000/3,200 = 0.625 weeks. The 80 copied rows add 0 new mistakes, which is under the 1% ceiling.
On all 100 rows the add-on costs 1,000 SEK a week of checking, 50,000 SEK a year, plus a license of 28,800 SEK and setup of 3,200 SEK. Year-1 cost is 82,000 SEK and the year-1 net is 118,000 SEK, which is 40,000 SEK below the sheet. On the 20 emails alone, checking costs 10,000 SEK a year against 20,000 of clerk labor, and the error saving is 20,000, for a gross gain of 30,000. Subtract 28,800 and 3,200 and the year-1 net is −2,000 SEK. The sheet is Go. The add-on is No-Go.
Check it on paper
Use 60 invoices a week, 5 minutes each, 360 SEK an hour, and 48 weeks. Write the weekly labor and the annual labor. A sample of 30 invoices holds 3 wrong sums, and each wrong sum costs 180 SEK to correct. Write the error rate, the weekly error cost, and the annual baseline, labor plus errors. Then 45 invoices already have supplier id, amount, and invoice id, and a sheet takes 3 hours at 360 SEK. The other 15 stay manual at the same 10% rate. Write the year-1 net of the sheet and the break-even in weeks. Also write the unattended ceiling as a fraction, from an allowance of one tenth of the labor on one invoice, divided by 180. An add-on is 800 SEK a month, with a 2-minute review on every invoice, remaining error cost 0, and no setup in this exercise. Write its annual operating cost and its year-1 net against the baseline, and mark the add-on Go or No-Go.
Weekly labor is 1,800 SEK and annual labor is 86,400 SEK, from 30 SEK an invoice. The error rate is 3/30 = 0.10. Weekly error cost is 6 × 180 = 1,080 SEK. Annual error cost is 51,840 SEK and the annual baseline is 138,240 SEK. The 15 manual invoices cost 21,600 SEK a year of labor and 12,960 SEK a year of errors. Annual saving is 64,800 + 38,880 = 103,680 SEK. Setup is 1,080 SEK, so the year-1 net is 102,600 SEK. Weekly saving is 2,160 SEK and break-even is 1,080/2,160 = 0.5 weeks. The ceiling is 3/180 = 1/60. The add-on review is 720 SEK a week, 34,560 SEK a year, plus a license of 9,600 SEK. Operating cost is 44,160 SEK and the year-1 net is 94,080 SEK, below 102,600, so the add-on is No-Go.
Check questions
In the parts-counter case, what is the weekly baseline in SEK, and which two amounts add up to it?
4,000 SEK, from 2,000 of labor and 2,000 of error cost.
The sample has 4 wrong weeks in 40 confirmations. What is the baseline error rate, and what unattended ceiling does the one-tenth rule set?
10%, and 1%, because the allowance is 2 SEK and 2/200 = 0.01.
What is the spreadsheet year-1 net, and what is the break-even time in weeks?
158,000 SEK, and 0.625 weeks, from 2,000/3,200.
What is the year-1 net of the add-on on the 20 emails alone, and is that add-on Go or No-Go?
−2,000 SEK, so No-Go. The sheet is the path you keep, at a year-1 net of 158,000 SEK.
A weekly baseline is hours times the hourly rate, plus mistakes times the cost of one mistake. Price a formula on the rows that already have fields before you price an add-on. A Go needs a year-1 net above zero and a measured error rate at or under the ceiling you wrote down.
Final task
Pass a 30-item keyed quantitative feasibility and ROI benchmark over 15 frozen business dossiers, plus submit an ROI & break-even calculator script passing 15 fixed unit tests. (Optional Portfolio item: real customer discovery notes with signed consent).
Terms in this course
This course has no terms of its own in the glossary yet. Terms from earlier courses are used here too.
The first lesson is open to read. The rest of the course is being written, and tests and diplomas come later. Join the waitlist