The structural test is workflow ordering
The structural test for whether a model is exploring or defending is workflow ordering. An exploring model builds inputs from named, dated, sourced evidence BEFORE the analyst writes down a target price. A defending model starts from the target — often anchored to the current price plus a familiar 15-20% upside — and reverse-engineers the assumptions. Both produce a number; only one of them is useful for decisions.
How exploring and defending models behave
| Behavior | Exploring model | Defending model |
|---|---|---|
| Assumption sources | Each input has a named source (filing line, peer mean, industry report) | Inputs are "around peer level" or "in line with management" |
| Sensitivity table | Shows where the answer breaks; analyst notices and reports the break | Set up after the fact to show the answer is "robust" |
| When evidence contradicts an input | Input gets updated; output moves; analyst writes down the move | Input gets reframed ("that data point is noisy") to preserve the output |
| Round-trip with current price | Compared after the model is built, not before | Drives the build from the start |
| What the analyst reports if model says SELL | Reports SELL or explains why the model is wrong about the situation | Tunes inputs until model says HOLD or BUY |
Worked example: two analysts model Brindlewood Networks
Worked example — two analysts model Brindlewood Networks. Analyst A pulls last-three-year revenue growth (8.2% mean), peer terminal-growth medians (2.5%), and a CAPM cost-of-equity build with named beta and risk-free rate. She runs the model and gets $43. Current price is $54, so she reports SELL with a $43 base case and writes a one-page note on why peers may be over-paying for the segment. Analyst B sees the stock at $54, opens the same model, and types in terminal growth of 3.5% ("justifiable" he says), revenue growth of 11% ("sell-side is too low"), and discount rate of 7.5% ("low-beta name"). He gets $61 and reports BUY with a $61 target. The outputs differ by 40%. The model is identical. The process is not.
Write your assumptions before opening the spreadsheet
Practical defense for the analyst-on-themselves: write the assumptions in a separate document BEFORE opening the spreadsheet. Stamp each assumption with the source. Build the model. Then look at current price. If the answer needs to move materially to "justify" the output, do not nudge — investigate what evidence would have to be different for the answer to move that far, and check whether that evidence exists. Most of the time it does not, and the right answer is to accept the model's number or kill the idea.
The right instinct when the model says overpriced
Track the models you nudged after seeing price
Sit with the ideas.
An analyst's DCF on Brindlewood Networks produces a $58 intrinsic value. The stock trades at $54. The analyst's terminal-growth assumption is 3.0% and the discount rate is 8.5%. Which of the following BEST diagnoses whether this is an exploring model or a defending model?