How to Defend a Site Selection Forecast to Your Board

A site selection forecast gets defended by making it traceable. Name each input, show the reading it produced for this site, and show which direction it pushed the number, so anyone in the room can follow the forecast back to its evidence. Committees approve analysis they can take apart. They defer analysis they cannot.

Most real estate teams prepare for the wrong test. They spend the week before committee polishing the recommendation and pressure-testing the revenue figure, on the assumption that the meeting is a referendum on the number. It usually is not. The number is the thing being pointed at. What is being tested is whether you understand where it came from well enough to keep the room from having to trust you blindly.

That distinction decides how the meeting goes. This piece covers what a committee is really testing, why a closed score fails that test, what a defensible forecast has to be able to show, and why franchisees ask a harder version of the same question than any board will.

What a committee is actually testing

A committee is testing your method, not your conclusion. Members are deciding whether the way you evaluate sites is consistent enough to keep funding one deal at a time. A specific question about a demand number is usually a proxy for a broader question about whether every site gets held to the same standard.

You can hear this in the questions that get asked most often. "How does this compare to the last ten we approved?" is a question about consistency. "What has to be true for the base case to hold?" is a question about whether you have looked at the downside. "Where did that number come from?" is a question about whether there is a source underneath the analysis or just a tool.

None of those are questions about the site. They are questions about you. A forecast that answers them without you having to vouch for it personally is worth more than a forecast that is slightly more accurate but arrives as an assertion.

This is also why the deck matters less than teams think. Structure and sequencing help, and there is a real craft to presenting site analysis to a real estate committee. But a well-organized presentation of an untraceable number still leaves the room with nothing to check.

Why a closed score fails in the room

Present a score with no visible inputs and you are in an impossible position. Every follow-up question has the same answer, which is that you would have to go back to the vendor. Do that twice in one meeting and the committee stops evaluating the site and starts evaluating the model, which is a conversation you cannot win from the chair.

The failure is not that closed models are bad at prediction. Some are quite good. The failure is that a number you cannot open is a number the room has to take on faith, and committees exist specifically so that capital decisions do not run on faith.

Look at what changes on the right side of that comparison. The score is identical. What moved is the subject of the conversation. Instead of debating whether the model can be trusted, the committee is now debating the competitor read, which is a question your team can research, answer, and settle. You have converted a credibility argument into a factual one, and factual arguments end.

This is the practical case for transparent scoring, and it has very little to do with methodology purity. Transparent site scoring opens every score into its component lenses with each input traced to its source and each weight set by your team rather than by a vendor. The point of that design is not elegance. It is that the person holding the number in a meeting can answer the next question.

The scrutiny is also proportionate to the bet. Cushman and Wakefield's 2026 U.S. Retail Fit Out Cost Guide puts the average in-line retail fit-out at $157 per square foot across the 15 markets it surveys, and that is hard construction cost alone, before rent, fixtures, or soft costs. Attach a lease that commonly runs five to ten years and the room is approving something it will live with long after the meeting ends.

Banking wrote this rule down first

Real estate committees are relitigating a question another industry settled some time ago. In April 2011 the Federal Reserve and the OCC issued joint supervisory guidance on model risk management, published as SR 11-7 by the Fed and Bulletin 2011-12 by the OCC, governing how banks use models in decisions that move real money.

Its organizing idea is effective challenge, which the guidance describes as "critical analysis by objective, informed parties that can identify model limitations and produce appropriate changes." The guidance ties that challenge directly to documentation: development and validation have to be written down in enough detail that someone who did not build the model can follow how it works, where its limits are, and what it assumes. The agencies replaced the 2011 letter with revised guidance in April 2026, so point at the current version if you ever cite it in a meeting. The phrase is the part worth borrowing.

Read that description again with your committee in mind. Objective, informed parties. Model limitations identified. Detail sufficient for someone who did not build it. A regulator decided, for one entire industry, that a model driving material capital decisions has to be legible to an outsider. Retail real estate has no examiner and no rule number. It has a board asking for the same thing informally, which is a reasonable thing for a board to ask.

Five things a defensible forecast can show

Across the retail, restaurant, and franchise teams we work with, the forecasts that clear committee reliably can produce all five of the following on request. The ones that stall are usually missing two or three.

1. Every input, by name. Not "demographics" but the specific measures used, the geography they were measured over, and the reading for this site. A committee member who can see the input list can tell you what you left out, which is a far more productive meeting than one where they can only ask whether you thought of everything.
2. A source and a vintage for each reading. Data has an age, and the age matters. Trade area figures pulled before a major local employer announcement describe a market that no longer exists. Knowing your sources by name and date is a small thing that buys a surprising amount of credibility.
3. The weight each input carried. Two brands can look at the same site and reach opposite conclusions because one lives on daytime traffic and the other on evening household density. If your weights are set by your own operating history, say so. If they came out of a box, you are defending someone else's assumptions about your business.
4. A range, not a point. A single figure invites the room to argue about its precision. A range with a named assumption moves the argument to what has to be true, which your analysis can actually address. It also gives finance what it needs to run occupancy cost against the low end instead of the middle.
5. The cannibalization math, quantified. Not a statement that overlap was considered. The stores inside the draw, the share of the trade area they share, and the estimated transfer. This is the single most common place a strong recommendation gets deferred, and there is a full treatment in our retail cannibalization analysis guide.

None of these require a better model. They require a model that keeps its work.

Franchisees ask the harder version

A board is spending the company's money. A franchisee is spending their own, usually with a personal guarantee attached to a lease that commonly runs five to ten years. That difference changes what you have to prove.

The pattern we see most often in franchise systems is that corporate development brings a well-supported recommendation to a franchisee who found their own site and is being told, in effect, that their instinct was wrong. If only the corporate pick was ever scored, that conversation goes badly regardless of how good the corporate pick is. Running both sites through the same criteria and showing both results changes the dynamic completely, even when the answer does not change. The franchisee is no longer being overruled. They are being shown the comparison.

The second pattern is local knowledge. An operator who has run a market for years knows things your data does not contain, including which corner floods, which landlord is difficult, and which competitor is quietly failing. A model that cannot take a local correction and re-run is telling that operator their experience is not admissible. Most will not sign under those terms, and they are not wrong to refuse.

What is striking is how little the rulebook helps here. The Franchise Rule requires disclosure of the mechanics: Item 12, at 16 C.F.R. § 436.5(l), covers whether the franchise is for a specific location or one the franchisor will approve, and Item 11, at § 436.5(k), covers what help the franchisor gives with locating a site and negotiating the lease. Both require you to say what you do. Neither requires you to show the analysis behind a site you recommend.

The industry's own guidance stops short in the same place. The IFA Statement of Guiding Principles says franchising carries a risk of failure like any business model, and that neither the franchisor nor the franchisee is promised economic success. It does not address site selection or location risk anywhere in the document. Nothing in the disclosure regime or the trade association's principles settles who carries the downside on a bad location. That gap is the reason a franchisee reads your forecast the way they do, and the reason showing your work matters more here than in any board meeting.

Territory questions sit right alongside this. A franchisee evaluating a new corporate site is also asking what it does to the boundary they were promised, which is why franchise territory mapping and site defensibility tend to surface in the same meeting. The franchise site selection guide covers the approval mechanics in more depth.

What to do when you cannot trace a number

Sometimes the input is genuinely unavailable, the vintage is unknown, or the calculation happened inside a system you do not control. The instinct is to smooth over it. Do not.

Say what you know, say what you do not, and commit to a date. "That figure comes out of the model and I have not confirmed the underlying vintage. I will have it to you Thursday" costs you very little in the room. A confident answer that turns out to be wrong costs you the next three meetings, because the committee will start checking your work instead of reading it.

The longer-term fix is to reduce how often this happens. Every untraceable number in your packet is a place where the meeting can stall, and the count of those is a decent proxy for how defensible your process is overall.

Defensibility gets built before the meeting

The forecasts that survive interrogation are not assembled the night before. They come out of a process where the evaluation criteria were set once, written down, and applied the same way to every site.

That is the real work. Decide what your brand actually cares about, set the weights from your own store performance rather than from a vendor default, and hold every site to that standard, including the ones you like. Store site selection criteria that live in a document instead of in someone's head are what make the tenth site as defensible as the first.

Teams that do this find committee meetings get shorter. The questions do not stop. The answers are already sitting in the analysis, and the room can see that for itself, which is a better outcome than winning an argument about a number. If your forecast tracking is thin, measuring forecast accuracy against actual results is where the credibility for the next meeting gets built.

Frequently Asked Questions about Defending a Site Selection Forecast