Notable fact: 4 out of 10 BVFLS professionals don’t prepare DCF projections if the client has not prepared them. (So I imagine a DCF model is also not prepared.)

The question above is one I ask during the course of NACVA’s 5-day Report Writing, Review & Analysis webinar.

I’ve conducted 26 of these webinars between January 2019 and last week ... asking hundreds of attendees this question.

When a DCF model is the most appropriate valuation method, projections are required. And there is nothing in our professional standards that says we don’t have to prepare projections if client-prepared projections aren’t available.

And whether the projections are developed by the client or by us with management collaboration, there is a prompt we can use to stress-test them.

An excellent AI for BV use case!

We use the DCF method when a company’s historical financials are not representative of how it is expected to perform in the future.

This can result from a whole host of reasons, including that the company is starting on or in the midst of:

1/ Realigning its current supply chain.
2/ Embarking on a capital expenditure campaign.
3/ Expanding into new geographical sales territories.
4/ Adding to its current portfolio of products or services.
5/ Reducing headcount for automation and/or AI efficiencies.

And if the company does not have a set of financial projections modeling the new state of affairs, it’s incumbent upon us to prepare them ... usually with management’s assistance.

But regardless of whether management has already prepared them or we get their assistance in preparing them, we must challenge the prospective financial information.

How might we use AI to assist with that?
Here's an example ... 

The prompt

You are a business valuation expert preparing a discounted cash flow analysis. Your client company operates in the [3D printing] industry for which you possess extensive knowledge. Your client has made recent improvements to its supply chain and expects to be able to produce its products for 10% less. Your client has decided to pass on those cost savings by reducing the price it charges to its customers. Your client has prepared a financial projection that reflects this, as well as their belief that this will lead to capturing increased market share and generating more top-line revenue. Your task is to challenge the assumptions of this decision that may be incorrect. List them as bullet points of 1-3 sentences each.

ChatGPT-4o result

I tried this prompt with the ChatGPT-4o model (the old, gold standard you select from the model picker). This is what I got ...

Here’s a list of key assumptions in your client’s decision that are worth challenging. Each of these has the potential to materially impact the reliability of the discounted cash flow (DCF) model and the business valuation conclusions that depend on it.

Assumptions to challenge

  • Price elasticity is favorable
    The client assumes that lowering prices will significantly increase demand. But this only holds if customers are highly price sensitive, which may not be true for industrial buyers of specialized 3D printing products.

  • Cost savings are sustainable
    Supply chain improvements might be temporary or reliant on favorable supplier terms. If those reverse (e.g., due to input cost inflation or geopolitical events), margins could compress further under the new, lower pricing structure.

  • Competitors won’t follow suit
    The client expects to gain market share, assuming competitors won’t also cut prices. But in a price-sensitive market, aggressive pricing may trigger a race to the bottom, eroding industry-wide profitability.

  • Brand or perceived quality won’t suffer
    A price cut may inadvertently signal lower quality or a commoditization of the product—particularly risky in an industry that often values innovation and reliability.

  • Capacity can absorb increased demand
    Even if lower pricing drives demand, the company must have the operational capacity to meet it—without compromising lead times, quality, or customer support.

  • Increased volume will offset lower margins
    The projections assume that the volume increase will more than compensate for the reduced unit margins. This depends on execution, market dynamics, and the fixed vs. variable cost structure.

  • New customers will be profitable
    Attracting market share doesn’t guarantee profit. Some customers may be less profitable to serve (e.g., smaller order sizes, higher service requirements, longer receivables cycles).

  • Market growth is not overstated
    Gaining share assumes the total addressable market is large enough or growing fast enough. If the market is stagnant or shrinking, growth via pricing will be harder to sustain.

  • Switching costs for customers are low
    Implicit in the plan is that buyers will quickly change suppliers based on price. But if switching costs are high—due to integration, qualification processes, or service relationships—the expected migration may not materialize.

  • R&D and innovation won’t be compromised
    Sustained lower pricing may leave less cash to reinvest in product development, which is critical in the rapidly evolving 3D printing industry.

Uncertainty map

  • Least confident: Actual customer price sensitivity in the client’s specific niche—elasticity can vary widely across segments.

  • Possible oversimplification: The idea that competitors will remain static; market reactions are complex and interdependent.

  • Key questions that could shift this view:

    • Does the client have exclusive IP or a technological advantage?

    • Are the lower production costs truly sustainable?

    • How price-sensitive is the end-user base historically?

    • Has the company tested this pricing strategy in a smaller market?

Let me know if you want to test the DCF model’s sensitivity to these assumptions or need help framing rebuttals in a valuation report.

ChatGPT-5 Thinking result

I also used the same prompt in the ChatGPT-5 Thinking model (the new kid on the block you select from the model picker). It thought for 19 seconds and returned this more detailed response ...

Here are the specific assumptions I would challenge before relying on the DCF that embeds the price-cut strategy: 

  • Demand is sufficiently price elastic.
    The model presumes a meaningful volume lift from a lower ASP. In many 3D printing segments (aerospace, medical), qualification and reliability often trump price.

  • Sales cycles will compress.
    Capital equipment purchases frequently run 6–18 months with budget gating. A price cut next quarter won’t instantly convert pipeline to revenue.

  • Switching costs are low.
    Customers that have validated printers/materials/software rarely switch on price alone. Re-validation and operator training can kill the expected share gains.

  • Competitors won’t retaliate.
    A price move may trigger matching discounts or rebates within weeks. The share “win” disappears while industry margins reset lower.

  • Channel partners will cooperate.
    Distributors holding higher-priced inventory will demand make-goods or slow orders. Expect channel friction, delayed sell-through, and one-time write-downs.

  • Brand positioning won’t erode.
    In professional/industrial 3DP, price often signals capability and reliability. A broad discount can unintentionally reposition the product as commodity.

  • Volume scale won’t stress operations.
    More units mean more installs, training, and field service. If service headcount lags, customer experience degrades and warranty accruals rise.

  • The 10% cost reduction is durable.
    Supplier concessions, freight rates, and powder/resin inputs can swing back. Temporary savings don’t justify permanent list-price cuts.

  • Cost savings hold at higher volume.
    Scrap, rework, and yield drift with throughput and mix. The “10%” measured at current run-rate may vanish under load.

  • Mix effects are benign.
    Cheaper printers may pull demand into lower-margin SKUs. If materials or service attach doesn’t scale, gross margin dollars fall despite unit growth.

  • Installed-base economics are correctly modeled.
    Lower printer price may grow installed base, but consumables usage is not linear with units; utilization varies widely by customer. Don’t assume a fixed materials attach rate.

  • All incremental customers are profitable.
    Smaller shops often require more hand-holding, financing, and longer terms. Contribution margin per customer can compress even as revenue grows.

  • Working capital won’t expand materially.
    Higher volume + lower ASP typically grows AR and inventory. DCF must reflect cash tied up in WC, not just the income statement.

  • Capacity and lead times are adequate.
    If key subsystems (lasers, motion control) are constrained, orders slip and revenue timing shifts right. Deferred installs push out service and materials pull-through.

  • Capex requirements are minimal.
    More units may require additional test cells, QC equipment, and service tooling. Understated growth capex inflates free cash flow.

  • Price reductions don’t trigger MFN/most-favored clauses.
    Some enterprise contracts automatically reset when list prices drop. Hidden revenue leakage is common after broad repricing.

  • No conflict with MAP or reseller agreements.
    Minimum advertised price provisions can limit promotional flexibility. Violations invite penalties or channel churn.

  • Aftermarket/service pricing holds.
    Customers anchor on the headline discount and push for service and materials concessions. Margin dilution spreads beyond hardware.

  • Geography and FX are neutral.
    If costs fell in one currency but sales are in another, FX can unwind the gain. Hedging assumptions should be explicit.

  • Regulatory/qualification timelines won’t delay adoption.
    In regulated end-markets, new materials or parameter changes may require re-qualification. Adoption lags undercut near-term projections.

  • Cannibalization is limited.
    A lower price can pull forward replacements or displace higher-end models in your own line. Net share gain may simply be internal mix shift.

  • Price integrity is preserved post-promotion.
    Once buyers learn there’s room to discount, future deals start lower. The terminal margin in the DCF should reflect a tougher pricing environment.

  • Customer value is defined by price, not outcomes.
    Many buyers optimize for throughput, tolerance, or validated workflows. A value-based pricing lens may outperform blanket cuts.

  • Competitive tech roadmaps are static.
    If a rival launches a platform leap (speed, build volume), your price cut won’t protect share. The model should allow for technology shocks.

  • One-time transition costs are negligible.
    Price list updates, ERP changes, rebate true-ups, and channel credits are real cash costs. They belong in near-term free cash flows.

  • Terminal value assumes margin normalization without evidence.
    If the strategy structurally lowers price floors, the terminal EBIT margin and growth must be reset, or TV will be overstated.

  • Sensitivity to key drivers is modest.
    Small misses on elasticity, competitive response, or service costs can swing value materially. If the DCF isn’t stress-tested, it’s fragile.

Uncertainty map

  • Least confident: True demand elasticity and attach rates for this client’s specific sub-segments; these are empirical and vary widely by application and region.

  • Potential oversimplifications: Treating “3D printing” as one market; dynamics differ for metal vs polymer, hardware vs materials, direct vs channel.

  • What would change my view: Evidence from controlled pricing pilots, cohort-level utilization data (materials/service attach by vertical), competitor pricing intelligence, channel inventory positions, and a capacity/field-service ramp plan with hiring and capex mapped to the volume case. 

The Uncertainty Map

I’ve mentioned this before, and it’s worth reiterating. My Custom Instructions (Your name in the lower left corner > Customize ChatGPT > What traits should ChatGPT have?) ask ChatGPT to:

Add an “Uncertainty Map” at the end of your responses. Here, describe what you're least confident about, what you may be oversimplifying, and what questions or follow-ups would change your opinion.

My take

I think the quality and quantity of the output are pretty cool ... and honestly, I’m not sure I would have thought of all of ChatGPT’s points on my own.

Also, the list of bullet points offers some excellent brainstorming ideas that, along with your other business valuation knowledge and experience, might trigger more management questions and some projection fine-tuning.

Finally, this kind of prompt can also be extremely helpful if you do not have extensive knowledge of the industry in which the client company operates.

That’s my take on this ... what do you think?!

And thanks for reading ... I appreciate you.

👀 FWIW 👀

QUESTIONS: Any questions about today’s newsletter? Email me, and I will respond directly and/or discuss it in the AI for BV Forum. Otherwise, I’ll be back on Wednesday, September 3.

Speaking of the Forum, it is your companion guide to this newsletter. A lot is happening in the AI for BV world ... more than I can share in a twice-monthly newsletter. So, I regularly post short tips, tricks, or hacks that will save you time, money, and tears! Most of you are already members. 🙏 But if you are not a Forum member …

ASSISTANCE: Identifying and implementing the right AI tools for your practice might require time you don’t have ... or even where to start if you did. I can help. Schedule a short call (20 minutes) to discuss your needs and how we might work together on this.

REMINDER: You have full access to the Archive of newsletters that were published before you subscribed. Also, there is the Resource Guide to accelerate your AI IQ. It’s a Featured Post on the AI for BV Home Page.

🖥 COMING TO YOU 🎤

For ASA:
Zoom workshop starting @ 1pm ET
Aug 28: BVFLS 3.0: Harnessing AI to transform your BVFLS practice 

For Business Valuation Resources:
Zoom workshops starting @ 1pm ET
Sep 12: Topic TBD
Oct 15 / Nov 12 / Dec 10: Topics TBD

For Jim Hitchner’s Valuation Products & Services:
Zoom workshop starting at 1pm ET
Oct 1: A deep dive on specific BV use cases and the AI tools to execute them

For TSCPA Forensic and Valuation Conference:
In person, time TBD
Oct 21: How to incorporate AI into your everyday BVFLS systems & processes
             ROI pricing (i.e., value pricing for the brave new world of AI)

For MNCPA BV Conference:
Zoom workshop starting at 310pm ET
Nov 3: BVFLS 3.0: Harnessing AI to transform your BVFLS practice

For FICPA Valuation, Forensic Accounting & Litigation Services Conference
In person, 810 - 900am ET
Nov 18: BVFLS 3.0: Harnessing AI to transform your BVFLS practice

For RICPA BV Conference:
Zoom workshop starting at 12pm ET
Nov 20: BVFLS 3.0: Harnessing AI to transform your BVFLS practice

If you attend, ask the questions you want answers to!

Keep Reading

No posts found