Start with the decision, the options and the baseline

A business case is a written argument that an investment is worth making. It answers what we spend, what we get, when, how sure we are, and why this option beats the others. Before any arithmetic, write three things.

The decision. "Approve $756,000 to replace the stores' handheld scanners, in three waves, starting in January." A case with no clear ask gets a polite discussion and no decision.

The baseline. Doing nothing is rarely free. Mangrove's roughly 1,000 handheld scanners are about six years old. Batteries die mid-shift, about one in ten is out for repair at any time, and staff queue for a working device while a supplier's driver waits at the dock. The case compares "replace" with "keep repairing and keep waiting", not with zero. Only incremental cash counts: money that changes because of the decision.

The options. Keep repairing; replace all 1,300 at once; replace in waves; lease the devices as a service; let staff use their own phones. Say why you rejected each one you rejected. Leasing spreads the cost but costs more over three years; staff phones fail on security, glare, gloves and drops.

Hardware-plus-software cases have costs that are easy to miss. List them all:

Finance will talk about capex (assets such as scanners, recorded on the balance sheet and depreciated over their useful life) and opex (running costs such as licences). The business case works in cash, month by month, because cash timing decides payback and NPV. Learn both words; the accounts and the case answer different questions.

Three numbers: payback, ROI and NPV

Here is the base case for the scanners. The costs come from the scanner vendor's quote and the Store app team's estimate; the benefits come from a four-week pilot, explained in the next section.

One-off, month 0
  1,300 scanners x $420                       546,000
  Store app changes and device setup          150,000
  Training and rollout                         60,000
  Investment                                  756,000

Monthly at full rate (stores go live over months 1-6)
  Hours saved   320 stores x 40 h x $3         38,400
  Shrink saved  320 stores x $90               28,800
  Support desk                                 -6,000
  Device management  1,300 x $4                -5,200
  Net per month                                56,000

Three years (36 months)
  Total net benefit                         1,848,000
  Payback   month 17 (cumulative cash back above zero)
  ROI       (1,848,000 - 756,000) / 756,000 = 144%
  NPV at 12% a year                           781,000

Each number answers one question and hides something:

• Payback answers "when do we get the money back?" It is easy to explain, and it ignores everything after that month: a case that pays back in month 17 and then stops looks the same as one that keeps paying for years.
• ROI answers "how much do we get for each dollar?" Here (total net benefit − investment) ÷ investment, over a stated horizon. It ignores timing: $1 in year 3 counts the same as $1 next month. Always state the horizon, because ROI grows with it.
• NPV answers "is this worth more than the company's other uses for the money?" Each month's cash is divided by (1 + r)^t, where r is the discount rate finance sets and t is how far away the cash is. A positive NPV means the investment beats that rate.

Why discount at all? $100 next year is worth less than $100 today, because today's $100 could earn a return elsewhere, and because the future is uncertain. Mangrove's finance team uses 12% a year for projects like this. Governments use lower rates for public projects: the UK Treasury's Green Book (2026 edition) sets 3.5% in real terms for the first 30 years. The purpose differs, so never borrow a rate from another organisation; ask your finance team. Finance may also quote the IRR, the rate at which NPV is exactly zero. For the scanners it is about 83% a year, which mostly says the payback is fast.

Hard benefits, soft benefits and the haircut

Finance reads the benefits line first and trusts it least. Sort every benefit into one of two groups:

• Hard benefits change cash: repair bills that stop, overtime and temporary staff not paid, stock no longer written off, licences cancelled.
• Soft benefits are real but don't change cash by themselves: minutes saved inside a shift that is paid anyway, faster answers, happier staff.

Time saved turns into money only in two ways. Either fewer paid hours (less overtime, fewer temporary staff at festival peaks, slower hiring), or the freed time goes to work that earns money, such as refilling shelves that would otherwise be empty. Name the person who will make that happen and get their agreement in writing. In the scanner case, the head of store operations commits to cutting temporary staff and overtime by about 30 hours per store per month once a store is live.

Where do 40 hours come from? The pilot in Ama's store and three others measured receiving time per delivery falling from 51 to 38 minutes, because staff no longer wait for a working device or re-key notes from paper. With about 4 deliveries a day, that is 13 × 4 × 30 ≈ 26 hours a month. Faster stock counts and price checks add about 14. Mangrove's staff cost about $4 an hour, but the case counts each saved hour at $3, a 25% haircut, because the store operations commitment covers about three quarters of the hours. Show finance the case three ways: with the haircut (NPV $781k), at the full $4 (about $1.14M), and with no hours counted at all.

That last version is the important one. With hours saved set to zero, shrink savings alone give an NPV of about −$292k and no payback within three years. This case stands or falls on the hours. Saying so out loud is what makes a managing director trust the rest of the document.

Ranges, sensitivity and switching values

Every input is an estimate, so the result is a range. Three tools show it.

Low, base and high cases. Move the uncertain inputs together to a pessimistic and an optimistic set. For the scanners, the low case (25 hours saved, $50 shrink, dearer devices and software) gives an NPV of about −$111k and payback in month 35. The high case (55 hours, $120 shrink, cheaper devices) gives about $1.51M and month 12.

Sensitivity. Change one input at a time by ±20% and sort by how far NPV moves. Draw it as a tornado chart:

The chart tells you what to argue about. A team that spends the meeting negotiating $20 off the scanner price, or debating 11% against 13%, is working on the small bars. The pilot should measure hours saved in more stores, and someone should ask the vendor how long the devices and their batteries really last.

Switching values. The UK Green Book defines them as "the values that a key assumption would need to change to, in order to make an option no longer value for money". For the scanners, holding everything else at base: NPV reaches zero if hours saved fall to about 11 a store a month, if the devices last only about 18 months, or if each scanner cost about $1,020 instead of $420. A managing director can judge "is 11 hours plausible?" much faster than a spreadsheet.

Finally, assume you are optimistic, because almost everyone is. In the database behind Bent Flyvbjerg and Dan Gardner's How Big Things Get Done (2023), about 48% of big projects finished on budget, 8.5% on budget and on time, and only 0.5% on budget, on time and with the benefits promised. The Green Book calls this optimism bias, "the proven tendency for appraisals to be over-optimistic about key assumptions", and tells appraisers to raise cost and time estimates and lower benefits. Compare your rollout plan with what Mangrove's last device rollout actually achieved, not with the vendor's brochure.

Read it the way a managing director reads it

A managing director reads a business case in about two minutes, top to bottom, and stops when a question goes unanswered. Put the answer first and expect these questions, roughly in this order:

The strongest answer to "what if you are wrong?" is often a staged decision. Ask for wave 1 now and the rest after a gate with a measured threshold. Mangrove's case already assumes stores go live over six months, so staging costs little, and it turns the riskiest assumption into something you measure before spending most of the money. For hardware, ask the vendor to hold the price for the later waves.

Hardware brings two more questions. "How long will it last?" needs the vendor's end-of-support date and the battery life, because the horizon is the biggest bar. "Who supports it at 6 am?" needs the support model: spares in each region, a repair turnaround, and device management so a lost scanner can be locked remotely. For a server, the matching questions are "how busy will it be?" and "what happens when models outgrow it?". Presenting the case in the room is covered in pm-52.