Trying To Get More Out Of Expert Calls (Without Ruining Them)

I’ve been stuck on a simple but uncomfortable question:

I spend a big chunk of my time on expert calls, and they’re genuinely my favorite part of the job. The question I keep coming back to is: how do I take what I hear on those calls and turn it into actual numbers in the model?

Not “use” in the vague sense of “it informed my thinking,” but in the literal sense: what number do I plug into the cell for cents per gallon (CPG), long-term volumes, or normalized margins after I hang up?

I’m not writing this as someone who has the answer. This is me trying to move my own process in a better direction, and I’d genuinely love feedback from people who’ve been doing this longer and at a higher level.

The tension: prodding vs keeping the relationship

On almost every expert call, there’s a moment where I could push a little harder.

• “When you say ‘margins are pretty healthy’… what does that actually mean in cents per gallon?”

• “When you say ‘down a touch’… is that 1–2%, 5–10%, or something else?”

• “When you say ‘single digits’… do you mean 1–3% or 7–9%?”

But there’s always a trade-off in the back of my mind:

• If I push too hard for precision, do I come off as transactional or annoying?

• Will this person be less willing to talk next time?

• Am I turning a good, open conversation into an interrogation for the sake of a model input that isn’t even “my” number?

So I sometimes back off a little earlier than I should. I walk away with a sense of the direction and some rough boundaries, but not always with numbers I feel comfortable tying directly to a cell.

That’s the core tension I’m trying to work through: how to get (and ultimately use) more output from a call without ruining the relationship or the vibe of the conversation.

The “false precision” trap (and what to do instead)

One understandable reaction is: “Don’t overthink it. Put in 40 cents per gallon and move on. The model is approximate anyway.”

I get that. Fake precision is dangerous.

But if I’m going to do the work and make the calls, it feels like I should aim a little higher than “eh, 40 cents seems fine.” The point isn’t to pretend I know the exact right number. It’s to use what I’ve learned to narrow the range on the 2–5 variables that actually drive the outcome.

In practice, that probably means:

• Actually tallying what I’ve heard from people who live this every day

• Putting some structure around the ranges for CPG, volumes, and margins

• Letting those ranges show up explicitly in the model, instead of just in my head

If I do twenty calls on an industry and hear a bunch of data points on CPG, volumes, and store productivity, it seems odd for all of that to collapse into a single “0.40” that’s mostly based on vibe and conservatism.

A better goal might be something closer to how superforecasters handle key variables: carefully collecting base rates and ranges, tracking how views differ by segment or cycle, and then updating as new information comes in. The output doesn’t have to be a single magic number; it can be a range with a clear center of gravity and a few explicit scenarios.

That’s the direction I’m trying to move in: still humble about what I don’t know, but a little more intentional about using what I do learn, instead of hiding behind “false precision” as a reason not to decide.

Moving toward a “superforecaster” approach on 2–5 key variables

I don’t think every line in a model deserves this treatment. Most don’t.

But for the 2–5 variables that actually matter, I’m starting to think they do:

• Cents per gallon over the cycle

• Long-term volume trends

• Normalized inside-store margins

• Realistic cadence for unit growth or closures

• Maybe one or two company-specific levers

The idea is to treat each of those like its own mini-forecasting problem:

1. Build an input tab just for that variable.
   Not a mess of assumptions, but a simple place where all the relevant data points live: every range, quote, or directional statement from management, competitors, and experts.

2. Tag the sources and context.

   • “Single-store operator — Midwest — said 25–35 CPG ‘through the cycle’.”

   • “Jobber, Southeast, said recent margins feel ‘elevated versus history’ and might normalize lower.”

   • “Public comp: disclosed X CPG in 2018–2024, with Y volatility.”

3. Make the range explicit.
   Instead of pretending there’s one “true” number, actually write down (making up numbers):

   • Base case: 30–35

   • Bull: 35–40

   • Bear: 20–25

   And link those ranges into the scenarios rather than a single “best guess” hard-coded into the model.

4. Let the model reference the range.
   If I’ve done the work and the weighted average of everything I’ve heard points to 30, then I shouldn’t plug in 40 just because management and consensus said so.

This is less about worshiping precision and more about respecting the information I’ve chosen to spend my time collecting.

Actually using the call: comments, ranges, and receipts

One very practical change I may experiment with is simple:

Where the key input lives in the model, I add a comment that lists:

• The 5–10 experts I’ve spoken with

• What range they gave (or what their qualitative language implies)

• Any public data that cross-checks it

Something like:

“CPG assumption (30–35):
– Operator #1 (TX, 20+ yrs): ‘Used to be low-20s, now 30+ feels normal.’
– Jobber #2 (Southeast): ‘Mid-30s right now, don’t count on it lasting.’
– Public comp A: 2018–24 avg ~28, last 2 yrs ~33.
– Public comp B: similar pattern.”

Is it perfect? No. It isn’t perfect but it forces me to actually use the work and be rooted in the data.

It also has two side benefits:

• Future-me can see it. I won’t remember in two years why I picked 32 vs 35 unless I write it down.

• It’s shareable. If someone else on the team opens the model for the first time, they can easily see where my numbers came from.

How to ask for numbers

On the call-side, I’m trying to get a little more intentional about how I push for numbers so it feels less like a cross-examination and more like a collaboration.

A few things that may help:

• Offering ranges in the question.
  “When you say ‘pretty healthy’, is that more like 20–25 CPG or 35–40?”
  People seem much more comfortable reacting to a range than producing a point estimate from scratch.

• Anchoring to history.
  “Versus, say, 2018–2019, are margins today 5 CPG higher, 10 higher, or about the same?”
  It’s often easier for operators to compare than to level-set.

• Being explicit about why I’m asking.
  “I’m not trying to pin you down to the penny, I’m just trying to avoid making up a number on my end that’s way off from reality.”
  That would probably disarm people a bit.

I don’t think this eliminates the risk of annoying someone, but it feels more honest: I’m telling them what I’m trying to do and why.

Where I’m still unsure

Here’s where I’d especially love pushback from more experienced people:

• How far do you go before you’re overfitting?
  At what point does all this become pseudo-scientific, dressing up a fundamentally uncertain world in too much structure?

• How much weight do you give experts versus your own prior?
  If my prior is 40 CPG and ten smart people all cluster around 30, how aggressive should I be in moving?

• Do you formalize this (model features, checklists, templates, etc.), or just do it in your head?
  I’m tempted to build a more explicit “superforecaster-style” tab for the 2–5 big drivers in every name. Not sure if that’s a good discipline or busywork.

Closing

The direction I’m trying to move in is simple:

If I’m going to invest time in expert calls, I want the work to show up where it matters: in the few cells that actually drive the investment.

That means:

• Asking slightly more targeted questions on the call

• Aggregating what I hear in a structured way

• And letting that structure influence the model more than my mood or my desire to feel “conservative”

I’m not there yet. This is very much a work in progress.

If you’re a PM or a senior analyst who’s wrestled with this, I’d genuinely love to hear how you handle it: what’s worked, what hasn’t, and any traps I should watch out for as I try to push my process in this direction.

Disclosure: none of this is investment advice, I/we may own positions in securities mentioned.