Most people counting business cases fail for one of two reasons: they claim benefits that cannot be traced to anything, or they omit costs that everyone in the room already knows about. Either way, the finance team stops reading.
A credible case is smaller and more boring than the first draft. It puts the honest range on each line, names the assumption behind it, and states clearly which benefits are excluded because you cannot prove them yet.
Start by separating benefits you can measure from benefits you believe
Dividing the claimed value into two categories changes the conversation from advocacy to analysis.
Measurable within one or two quarters, given the system:
- Labor hours aligned to real traffic instead of fixed schedules.
- Reduced overtime at genuinely busy periods, or reduced overstaffing at quiet ones.
- Traffic and conversion reporting that replaces manual counts and guesswork.
- A common basis for comparing stores, replacing anecdote with an index.
Plausible but harder to attribute:
- Conversion improvement from better staffing, layout, or merchandising decisions.
- Marketing and promotion evaluation, by comparing traffic and conversion before and during a campaign.
- Rent or lease negotiation informed by measured footfall.
- Store opening, relocation, and closure decisions based on traffic potential rather than intuition.
Build the base case from the first group. Present the second group as upside with an explicit note that attribution is shared with many other factors.
The cost side, complete
Understating costs is the fastest way to have a model rejected. Include:
- Hardware. Cameras or sensors per store, plus spares held at regional level, because a failed unit that stays failed destroys the store's traffic history.
- Installation. Mounting, cabling or power provision, and calibration. Sites with difficult access to power or ceiling voids cost more, and every estate has a few.
- Software and platform. Licence or subscription per store or per device, plus any one-time setup.
- Support and maintenance. Firmware updates, RMA handling, and the technician time to swap units.
- Network. Additional connectivity where a store's link cannot carry what the design needs, though edge processing reduces this materially.
- POS integration, if conversion rate is in scope. This is the most commonly under-budgeted line in the entire project.
- Project and internal time. Store visits, coordination with regional managers, data validation, and reporting setup. Typically weeks of one person's time.
- Ongoing operations ownership. Someone at head office is accountable for the data being used. That is a real, recurring cost — and skipping it is what makes the rest of the investment worth nothing.
A model that leaves out support, spares, or internal time will show an attractive payback that does not survive contact with the second year.
Quantifying the labor benefit honestly
Labor scheduling is usually the largest defensible line, because the traffic data directly answers a question the schedule is currently guessing at.
The method is straightforward:
- Take a representative period of hourly traffic per store.
- Define the service standard — for example, customers per staff member per hour, or maximum queue length at peak.
- Compare the current roster against a roster built on that standard.
- Value the difference at fully loaded hourly labor cost, including statutory contributions where they apply.
- Report the result as a range, not a single number, because the service standard is a business decision, not a measurement.
Two cautions. First, savings only appear if roster changes are actually implemented — a chain that measures the imbalance and changes nothing gets the cost and none of the benefit. Second, some chains do not reduce total hours; they redeploy them from quiet mornings to busy evenings. That is a genuine service improvement, but it may not appear as a cash saving, and the model should say so rather than overstating it.
Quantifying conversion improvement
Conversion rate is transactions divided by traffic. It is easy to compute and difficult to attribute.
If traffic per store is known, a conversion improvement of small magnitude can produce a meaningful sales number, which is exactly why this line gets inflated in business cases. Keep it disciplined:
- Use your own historical data to show how much conversion usually varies between stores in the same peer group. If the spread is wide, there is theoretical room for improvement; if all stores sit close together, there may be very little.
- Attribute only a fraction to the project. Staffing and layout changes, not the camera, produce the improvement; the camera only makes the decision visible.
- Give a timeline. Decisions take time to implement and to show up in sales, which means the benefit curve is slower at the start.
- Show the break-even version. State how small the conversion improvement would have to be for the project to break even. That number is often surprisingly low, and it is more persuasive than an ambitious projection.
Building the payback picture
Once costs and benefits are valued, present the investment in the form the finance function uses:
- Simple payback, in months, on the base case — with the conservative benefit set, not the optimistic one.
- Annual recurring cost broken out separately from one-time cost, so the comparison to a subscription or a staff budget is clean.
- Sensitivity table, showing payback at low, medium, and high assumptions on the two largest variables, typically labor benefit and conversion attribution.
- Excluded benefits stated explicitly, so nobody accuses the model of hiding upside or of double counting.
If the project only works at the optimistic end of both variables, say so plainly. That is a decision about risk appetite, and the model should surface it rather than bury it.
What actually decides the return
Three factors determine whether the investment pays off in practice, and none of them are in the spreadsheet:
Who acts on the data. A chain with one named owner at head office, a weekly reporting cadence, and a roster process that consumes the hourly curve will realise more value than a chain that installs the same hardware and files the dashboards.
Whether the counting basis is consistent. If stores use different technologies, cover different entrances, or count staff in some locations and not others, cross-store comparison — the source of most decision value — collapses. Consistency is worth more than a marginal accuracy edge.
Whether the project scope stays small enough to finish. Chains that try to deliver traffic analytics, queue management, zone heatmaps, and full POS integration in the first release often deliver none of them on schedule. Sequence the scope and bank the first benefit.
A worked structure you can copy
Rather than hunting for a template, use this structure and fill in your own numbers:
- Scope. Number of stores, entrances per store, one-line description of what is measured.
- Costs, one-time. Hardware, installation, integration, project time.
- Costs, recurring, annual. Platform, support, connectivity, internal owner time.
- Base-case benefits, annual. Labor alignment and reporting efficiency, each with its assumption stated in a footnote.
- Upside, unquantified. Conversion improvement, marketing evaluation, property decisions — described, not valued.
- Payback and sensitivity. Two variables, three scenarios each.
- What we will measure to prove it. The specific metrics reviewed at 90 days and 12 months, and who owns them.
Point 7 is the one that distinguishes a serious proposal. Committing in advance to what would prove the investment wrong is what makes the rest of the numbers credible.
FAQ
How long does a people counting deployment usually take to pay back? There is no defensible universal figure, because payback depends on store count, labor cost, installation conditions, and whether roster changes are actually implemented. The useful approach is to model your own costs and benefits with a sensitivity range rather than adopt a quoted payback period — a number taken from another market is not evidence about yours.
Can I claim conversion improvement as a guaranteed benefit? No. Conversion is influenced by pricing, assortment, stock availability, competition, and seasonality alongside staffing. Treat conversion upside as an unquantified benefit or attribute only a small, clearly stated fraction of it, and show what improvement would be needed to break even.
Is it better to pilot in a few stores before committing the full budget? Yes, and the pilot should be scoped so it can also produce the missing number in your model: the true installation cost per store, including the difficult sites. A pilot that only proves the camera works leaves the cost side of the business case unvalidated.
What is the strongest argument for a single-store retailer to invest? For one store, the labor-alignment line is usually weak and the reporting line is thin. The stronger case is decision quality: knowing whether a bad month was fewer visitors or fewer purchases, and being able to test a change — layout, promotion, opening hours — with an actual before-and-after measurement rather than an impression.
Should I buy hardware outright or use a subscription model? It depends on your cash position and how you fund technology. Outright purchase shifts cost to capital expenditure and is usually cheaper over a long horizon; subscription spreads cost and often bundles support and updates. Either way, model the full recurring cost including spares and internal ownership time, which exist in both options.
---
HooViz builds the sensing layer for retail traffic analytics: 2D AI and 3D people counting cameras, fisheye panoramic thermal cameras for zone analytics, 4G solar-powered cameras for sites without infrastructure, and AI NVR systems for local aggregation. The solutions section shows where each fits in a retail or mall deployment, and the product center lists the current families. If you want to discuss a project scope and the cost drivers that will shape your model, contact our team with your store count and entrance layout.