Run a valuation
How dbot picks a valuation shape per company: a single DCF, a segment-driven DCF, a sum of the parts, a dividend model, or a bank FCFE model.
Most companies get a single discounted cash flow. But a bank is not a software company, and an industrial group with a car business and a lending arm is not one business at all. The adaptive pipeline lets dbot fit the valuation to the company in front of it instead of forcing every company through the same model.
Every analysis runs the adaptive pipeline, so there is nothing to pick when you start one. dbot researches the company the usual way, then a plan step decides how to value it: as a single business, segment by segment, as a sum of its parts, on its dividends, or, for a bank, on the regulatory capital its balance sheet has to carry. The rest of the run follows the shape it chose. Every node in the progress graph on the report page is one of these steps, so you can click a node to jump to that step here.
The five shapes share the same research at the front and the same pressure-testing and write-up at the back. What changes in the middle is how the company is modeled. The shared steps are described once below, then each shape's own steps, then the finishing steps they all have in common.
dbot pulls recent news in a few buckets: the company itself, earnings, strategy moves like launches and acquisitions, what competitors are doing, and where the industry is heading. It also runs searches aimed straight at your thesis, looking for anything that backs it up or argues against it, then sorts what it finds by date and source and keeps facts separate from speculation.
dbot reads the company's latest 10-K, 10-Q, and 8-K filings and answers each of your framing questions with evidence pulled straight from the disclosures: the risk factors, the MD&A, the financial statements, and management's guidance. It pays particular attention to the segment disclosures, the tables where a company breaks its revenue and profit out by business line, since those decide whether a segment-driven or sum-of-the-parts valuation is even possible. Every finding is cited by filing, date, and section.
For each framing question, dbot looks for evidence on both sides. It goes after the bear case and the competitive threats instead of just confirming the bullish view, and it weighs how good, how recent, and how reliable each piece of evidence is.
Alongside the research, dbot pulls the live market data the valuation needs: the current share price and market context, the financial history behind a cash-flow model, and the company's reported business segments. That last piece, the revenue and profit each business line reports, is what makes a segment-driven or sum-of-the-parts valuation possible. If a company does not break itself into segments, dbot values it as a single business. A failed fetch here (for example, the cash-flow history for a bank headed to a dividend or bank FCFE model) never stops the run; dbot works with what it has.
dbot pulls the research streams (news, filings, and thesis evidence) into one brief. You get an executive summary, a read on the competitive landscape, an answer to each framing question, what it all means for the valuation, and a list of what is still unknown.
This is the step that makes the adaptive pipeline adaptive. dbot looks at what the research turned up and at how the company reports itself, and it decides how to value it. A planning step proposes a shape; then a deterministic check validates that proposal against the company's actual reported numbers before anything is valued.
| Single-business DCF | One story, one set of drivers, one discounted cash flow. The default for a focused company. |
|---|---|
| Segment-driven DCF | Still one cash flow, but built up business line by business line. Growth and margins are set per segment and summed into the company’s operating story. Fits a company with a few clearly reported segments that still trade as one. |
| Sum of the parts | Each business valued on its own, with the method that suits it, then added together. Fits a company whose parts are so different that one model would distort them, like an industrial with a captive lending arm. |
| Bank FCFE | For a bank, where the binding constraint is regulatory capital rather than dividend policy. Free cash flow to equity is net income minus the capital the bank must add as its risk-adjusted assets grow, discounted at the cost of equity. |
| Dividend discount model | For dividend-driven regulated financials (insurers and reinsurers) and for REITs, whose statutory payout makes dividends the distributable cash flow. Value rests on dividends, return on equity, and the cost of equity. It is also the fallback for a bank whose Tier 1 capital and risk-adjusted assets cannot be sourced. |
Before dbot values anything, a deterministic resolver checks the proposed shape against the company's reported segments and a set of rules. If the plan holds up, it is locked in. If it does not (say a sum-of-the-parts plan that would quietly drop a meaningful slice of revenue it cannot cleanly map, or segments that do not add back up to reported revenue), the resolver simplifies it to a shape it can stand behind, usually a single DCF, and records why. You always see the shape that actually ran, and the reason for any downgrade.
Parts within parts. Inside a sum of the parts, one part can itself be a small business made of several segments valued together. dbot calls this a composite (or hybrid) part: it runs a full segment-driven DCF just for that part, then feeds the result in alongside the simpler parts. A carmaker is the classic case. Its whole vehicle business (several nameplates) becomes one composite part valued as a segment-driven DCF, while its finance arm is valued separately on the returns it earns on its book of loans. A plan with no composite part is called compact.
Expansion options. A plan can also carry an expansion option: a business the company has not built yet, with no revenue today, that is worth something because of the option to build it. dbot values it as an expected payoff rather than as a going concern, shows it as a dashed amber sliver in the composition bar, and never lets it count toward the number of parts or segments a shape needs to qualify. It is upside, kept separate and kept honest.
What you see. While dbot is deciding, the report page shows a "choosing the valuation approach" placeholder. Once the plan is locked in, the plan card fills in: the chosen shape, a bar showing how revenue splits across the parts or segments, a tile for each one with its valuation method and industry, a short note on why this shape was chosen, and a "downgraded" flag if the resolver simplified the plan. When the report finishes, the same summary is saved on it as "How This Was Valued."
Every shape starts by turning the research into a company story: the narrative of where the business is going that the numbers then have to honor. In a single DCF it is one story. In a segment-driven or sum-of-the-parts run it is the company-level story that the individual segment or part stories hang off. In a dividend model it frames how much the company can afford to pay out and still grow. In a bank FCFE run it is written around the franchise and the capital position: how sticky the deposits and the funding are, what sits inside the risk-adjusted assets, where the Tier 1 ratio stands against the regulatory minimum and management's own buffer, and where that ratio is headed. From here the run does the work its shape needs before it can put numbers to the story.
For a segment-driven DCF, dbot researches each reported business line on its own: how fast it is growing, how profitable it is, and what it is up against. It leans on the segment disclosures in the filings.
dbot writes a short story for each segment, so every line of business gets its own view of growth and margins rather than a single company-wide average.
dbot sets the value drivers, growth, margin, and reinvestment, for each segment. These add up into the company's overall operating numbers, which feed one cash flow.
For a sum of the parts, dbot researches each part as if it were its own company, because for valuation purposes it is. Each part gets its own read on growth, profitability, and competition.
dbot records what each part's method needs. A part valued by compact DCF needs operating drivers (growth, margin, reinvestment). A finance arm valued on the returns it earns needs its book equity, its return on equity, and the debt that belongs to it rather than to the rest of the company. Keeping that debt with the arm it belongs to is what stops a lending business from dragging down the value of the operating business next to it.
When one part is a composite (several segments valued together), dbot values that part first, on its own. It researches, writes a story, and sets drivers for each of the part's segments, runs a full segment-driven DCF for just that part, and carries the result into the sum. On the report page this shows up as a composite-part step, and the composition bar breaks the composite down into its pieces. A sum-of-the-parts plan with no composite part (a compact plan) skips this step.
For a dividend model, dbot pulls the dividend history and the numbers behind it: the payout, the return on equity, and the book value the model reasons from.
For a bank FCFE run, dbot pulls the bank's statement facts (net income to common and the book equity behind it, on the same contract the dividend model uses) and then goes after the two numbers no statement table carries: the risk-adjusted assets and the Tier 1 capital held against them. Those live in the capital-adequacy disclosures, so dbot sources them from the filings and has to cite each one. Plausibility checks bounce a figure back when the units slip or the pair disagrees with the capital ratio the bank reports. Everything downstream turns on those two numbers, because Tier 1 capital against risk-adjusted assets is the constraint that decides how much of its earnings a bank can actually hand over.
With the story set and each shape's inputs in place, dbot runs the valuation. The engine underneath is the same Damodaran framework in every case; what differs is what goes into it.
The drivers are the handful of numbers the story has to cash out into: how fast revenue grows, the operating margin the business earns its way to, how much capital it has to put back in to grow, the growth it settles into for good, and the rate all of it is discounted at. Reinvestment is expressed as sales to capital, the revenue each dollar of invested capital carries. A segment-driven run sets these per segment and consolidates them; a sum of the parts sets them per part. Each driver is recorded with the reasoning behind it, so you can read the sentence of story that produced the number.
These are shown on the report page under "Assumptions & Drivers", in the units the engine recorded them in: rates in percent, sales to capital in turns, revenue in millions.
The discount rate is built up rather than assumed. dbot works out a bottom-up cost of capital for the business it is actually valuing, off the risk profile of the industry it sits in and a mature-market equity risk premium, instead of applying one house rate to every company. A separate terminal anchor takes over as the company matures, because a business in steady state should not carry the risk profile of one still growing into its market. The report shows both, alongside the industry cost of capital and industry beta they were struck against, so you can see how far the company sits from its own industry and decide whether that gap is earned.
Every forecast starts from somewhere. dbot records the base-year facts the model was built off: the industry it placed the company in, its market cap, its operating margin both as reported and after capitalizing R&D as the investment it really is, the invested capital behind the business, and the sales to capital it actually achieved. These are the anchors the drivers get judged against. A margin forecast that ignores where the company starts from, or a reinvestment assumption nowhere near the capital efficiency it has ever managed, is exactly what the quality critique is looking for.
dbot runs the valuation in reverse to read what the current price is really assuming. It nudges the driver that matters most until its own value meets the market price, then tells you the growth, margin, or return the price implies. It stops once it lands within tolerance or hits its iteration cap.
dbot runs a battery of scenarios, changing one key driver at a time, and works out how much each one moves the value. You get a fair-value range and a ranking of which assumptions matter most. On the report page these show up in the live activity feed as named tests, each with a value per share and an upside or downside. If dbot revises the valuation later (see Revision), it re-bases these tests against the new numbers so the range always reflects the final model.
dbot pulls the sell-side consensus: price targets (low, average, high), EPS and revenue estimates, and the mix of buy, hold, and sell ratings. It lines those up against its own fair value and growth assumptions and explains anywhere the two really disagree. If a company has no analyst coverage, dbot notes the gap and moves on.
dbot grabs data on the peer group and works out the multiples that matter for the company (the usual P/E, P/S, and EV/EBITDA for an operating business, price-to-book and return on equity for a financial). It checks the company's own implied multiples against its peers to see whether a premium or discount actually makes sense given the fundamentals.
This is a Professor-Damodaran-style review of the base valuation, and it ends with a clear approve or reject.
The terminal-value check is life-cycle aware. A young, fast-growing company that is still burning cash is expected to hold most of its value in the future, and the critic allows for that instead of rejecting it out of hand.
This step runs only when the critic rejects the first pass. dbot takes the critique to heart, adjusts the drivers or inputs it flagged (and, where the critique calls for it, the discount rate), and re-runs the valuation. It then re-bases the sensitivity tests off the revised numbers so nothing downstream still points at the rejected version. If the critic approves the first time, this step never appears in the graph.
For the two cash-flow shapes, single-business and segment-driven, dbot runs a Monte Carlo simulation: it varies the key drivers across plausible ranges thousands of times and shows the distribution of outcomes, not just a single point estimate. Sum-of-the-parts, dividend, and bank FCFE runs skip this step and lean on the sensitivity battery instead.
dbot pulls everything together (the research, the valuation, the convergence work, the sensitivity range, the consensus comparison, the comparables, and the critic's verdict) into a tight brief: what the research found, what the valuation says, the assumptions it leans on, how the pieces fit, where it lands against Damodaran's principles, and the risks worth watching.
dbot renders the charts and tables for the shape that ran. A single or segment-driven run gets a valuation bridge, a cash-flow projection, a peer comparison, and a sensitivity heatmap, and a segmented run adds a revenue stack and a value-by-segment chart. A sum of the parts gets a value-by-part bar chart and a parts summary table. A dividend run gets a dividend projection and a price-to-book-versus-return chart. A bank FCFE run gets its own value bridge and a projection of the free cash flow to equity the capital plan leaves behind, plus a one-page summary table. Anything a given plan does not have, such as a parts chart on a single DCF, is simply left out.
dbot fills in a spreadsheet model you can download and push on yourself. A single or segment-driven run gets a Ginzu-style DCF workbook; a dividend run gets Damodaran's dividend spreadsheet; a bank FCFE run fills Damodaran's own bank workbook, the CitiVal2023 spreadsheet behind his published Citi valuation, so the capital path and the FCFE line recalculate exactly as they do in his version. A sum of the parts gets one combined workbook: the parts bridge on the front sheets and, for a composite part, the full segment-driven model behind it, all wired together so the numbers recalculate live when you change an assumption.
dbot writes the long-form, Damodaran-style report. It takes your thesis and framing questions head-on, drops in the charts, links the Excel model, and walks the numbers for the shape that ran, part by part or segment by segment where that applies. It runs from an opening market observation to the valuation thesis, then through the numbers, the counter-arguments, and a balanced close. No buy or sell calls, and it ends with a few specific things to watch.
Finally dbot renders the report to a PDF, saves the workbook and charts alongside it, and delivers everything to your report page.
For the mechanics a single DCF or a dividend run shares with the rest of dbot, see the DCF pipeline and the DDM pipeline. When the run finishes, head to Viewing a report to read it, or Downloading PDF, Excel & charts to grab the files.