Main Street Index

What is in the dataset

• 75,000+ distinct businesses for sale from businesses-for-sale marketplaces and the brokers who post on them, in 100+ countries.
• Each listing classified into one of 130+ industries across 13 sectors, with asking price, revenue, earnings, employees and year founded where the listing discloses them. Missing fields are missing, never zero.
• 4,000+ brokerage firms, recorded at firm level only with the number of listings each posts. Never a person or a contact. Some firm names are derived from a URL slug, so capitalisation can be off.
• Per-industry benchmarks, a Software Gap Score per industry, the stated reason for selling, and 30+ product theses that survived an adversarial review.

The public overview is on the Main Street Index landing page. The app itself is at /main-street-index and is included on every paid plan.

SDE versus net profit

Seller's discretionary earnings (SDE) adds the owner's salary and perks back into profit. It is how US and Canadian listings quote earnings. UK and Australian listings quote net profit, which leaves the owner's pay as a cost. The same business has a higher SDE than net profit, so an SDE multiple and a net profit multiple are not comparable.

Benchmarks and multiples

• A multiple is the asking price divided by the disclosed earnings. Asking prices usually sit above what a business closes at, so read a median multiple as the opening position, not the result.
• Benchmarks are split by currency, earnings basis, region and size band.
• A cell with fewer than 30 listings is marked indicative and shows no percentiles. Do not compute your own percentiles over a thin cell.

The Software Gap Score

One score from 0 to 100 per industry, built from two equally weighted inputs. Operator density is how many businesses in the industry are for sale, capped at the 75th percentile across industries. Software scarcity is how few software products serve the industry, matched from the G2, Capterra and Stripe Index corpora.

• 70 to 100: wide open. Many operators, little software built for them.
• 50 to 69: underserved. Software exists but is thin for the number of businesses.
• Below 50: served.
• Unknown: the software side could not be read, or the industry has too few listings. A gap we cannot see is never scored as a gap.

The score finds markets, not winners. Pair it with demand from the complaint corpus and with the Stripe Index, which counts companies already selling into a category.

The stated reason for selling

When a listing gives a reason for selling, it is classified into a category such as retirement, relocation, health or other business interests. This is the stated reason: what the seller chose to write. It is not a verified motive and should never be read as why owners actually sell.

• Rates exclude non-answers and listings that are not an owner exit.
• The classifier was accuracy-tested on US listings. Other countries share the same head of the distribution but were not tested separately.

For buyers: the facts each listing states

Every listing carries the buyer facts its site printed as labels, read without any model. The rule for every field: silence is never no. A field the listing does not mention reads Not stated, and every published rate is the share of listings that state something, never the share that have it.

• Seller financing: offered, not offered or not stated, from the site's financing label or the BizBuySell search card. The financing text is kept verbatim, with a flag when it names a bank, lender or SBA, and an SBA eligible or not eligible label where the site prints one.
• Management model: owner operated, absentee or employee owned, where the site has a management label.
• Real estate: leased, owned, or lease or purchase; whether it is included in the asking price; and the real estate price where listed.
• Rent: the amount and its period (month, year, week or per square foot), annualised where possible. Lease expiry date on US sources only, because other sites write day-first dates.
• Inventory and FF&E: the value and whether each is included in the asking price.
• Training: whether it is offered, for how many weeks, and its cost.
• Home based, relocatable, franchise resale (with the brand on BizBuySell), living accommodation, asset sale and distressed flags.
• Price reduced: BizBuySell only, as the search card stated it on the capture date. A point-in-time flag, not a price history.

Deal vs industry, yield and payback

Each listing with a disclosed price and earnings gets a multiple (kept only between 0.1x and 50x) and a comparison against its industry's median asking multiple for the same currency, earnings basis and asking-price band. When that band has fewer than 30 listings, the industry's all-sizes cell is used instead; when that is also under 30, there is no comparison. A cell in another currency or basis is never used.

• Well below: under 0.60 times the industry median.
• Below: 0.60 to under 0.85.
• In line: 0.85 to 1.15.
• Above: over 1.15 up to 1.50.
• Well above: over 1.50.
• Earnings yield is earnings divided by the asking price. Payback years is the asking price divided by earnings. Both are before debt service, tax and capex.

AI-read buyer fields and their gate

Some facts only the description states. An AI model (gpt-6-luna) reads the listing description alone, never the labels, and records owner involvement (absentee, semi-absentee, owner operator, manager run), owner hours per week, whether a manager is in place, seller financing and SBA mentions in the text, lease years remaining, renewal options and assignability, equipment condition, revenue model, customer concentration, licences required, visa eligibility (E-2, EB-5), growth levers, and its own opinion of who the business suits. Every field defaults to Not stated and carries the verbatim quote it was read from.

• Every field has an accuracy gate: 85% on a fresh, hand-adjudicated sample of the full run. On the October 2026 run all 14 fields passed on a 100-listing sample, with revenue model lowest at 86%. Owner hours, SBA status and visa mentions are rare, so their pass rests on only a handful of stated cases. A field that fails a later run goes back to labelled context with its quote and stops being a filter.
• A field that fails its gate stays labelled text-sourced context, never filtered or counted.
• Who the business suits is model opinion and is always labelled that way.
• Where a site label and the description disagree, the label is shown first and the AI field as context.

The SBA 7(a) debt-service calculator

Each listing with a price and earnings has a debt-service calculator. It runs in your browser and stores nothing. It works out the loan (price less your down payment), the annual payment on monthly amortisation, and the debt service coverage ratio: stated earnings divided by annual debt service, against the 1.25 rule of thumb lenders use.

• Down payment, interest rate and term are all visible and editable. The defaults (10% down, 10.5% interest, 10-year term) are assumptions, not a quote from a lender.
• Stated SDE is before a replacement manager's salary, tax and capex, so a lender will haircut it. A ratio that only just clears 1.25 on stated SDE usually fails underwriting.
• No verdict is stored. The ratio is a check you run with your own assumptions.
• SBA 7(a) is a US programme. On listings from other countries, read it as a generic debt check.

Buyer Fit: best businesses to buy

Buyer Fit ranks industries for buyers on Best businesses to buy. The weights were written down before any score was computed, so they could not be tuned toward the industries they happen to favour (buyer scoring version 1.0.0).

• Earnings yield, median earnings over price, higher is better: 0.25.
• Affordability, median asking price, lower is better: 0.15.
• Margin, median earnings over revenue: 0.15.
• Durability, median years in operation: 0.10.
• Healthy exit, the share of stated reasons that are retirement, health, relocation and similar, not distress or burnout: 0.10.
• Supply, listings for sale: 0.10.
• Productivity, median revenue per employee: 0.05.
• Seller financing, the share of listings that state it is offered: 0.05.
• Real estate, the share that state owned real estate: 0.05.

Each input is a percentile rank among the industries in the same currency and earnings basis, so pounds are never pooled with dollars and SDE is never pooled with net profit. An industry is rankable when 30 or more of its listings disclose price and earnings. Yield, affordability and margin are required; the optional inputs that are missing are dropped and the rest reweighted, and every row lists the inputs it rests on.

Accuracy gates and coverage limits

• Industry classification passed its accuracy gate for the US, UK, Australia, Canada, France, Spain and South Africa sites and for BizBuySell. Per-industry figures for those are publishable.
• The UAE site failed its gate. Its per-industry figures are indicative only.
• Around 20 smaller country sites have too few listings to test. Their per-industry figures are indicative only.
• Businesses cross-posted on two marketplaces are counted once. Both listings stay in the data, but every count and benchmark uses one.
• Delistings are not tracked yet. A listing that was sold or withdrawn after we collected it may still read as active, so do not treat the active count as a live inventory.
• Price history is only days deep, so time on market is not published.
• Prices are shown in the listing's own currency. Several international sites quote some listings in USD, AED or ZAR.

Use it from the chat or the MCP server

Ask the home chat about buying a business, an industry's multiples or its gap score and it answers from this dataset. The same data is on the BigIdeasDB MCP server as eight tools, for Claude, ChatGPT, Cursor and any other MCP client. MCP access is part of Pro.

• search_mainstreet_listings: filter by keyword, industry, country, state, currency, asking price, earnings and stated reason, plus the buyer filters (seller financing, real estate, management model, training offered, visa mention (AI-read, gate passed), franchise, home based, relocatable, price reduced, earnings yield, deal vs industry). Paged with a cursor, each business counted once.
• semantic_search_mainstreet: describe a business in plain English and get the closest listings by meaning.
• get_mainstreet_listing: one business in full by slug, every scraped and analyzed field, with the daily asking-price history, the buyer fields, the deal metrics and the AI-read fields with their quotes.
• get_mainstreet_industry: benchmarks split by currency and earnings basis, the Software Gap Score, the Buyer Fit block and the stated reasons, each with n and disclosure rates.
• search_mainstreet_theses: the build theses that survived the adversarial kill pass.
• get_mainstreet_brokers: brokerage firms at firm level only.
• get_mainstreet_coverage: coverage per listing site and country, disclosure rates and accuracy gate status.
• rank_mainstreet_industries_for_buyers: industries ranked by Buyer Fit within one currency and earnings basis, every input with its n, filterable by budget and sector.

Setup and example prompts are in the Main Street Index MCP tools guide. Every tool argument is in the MCP tools reference.

Export

Browsing, search, benchmarks, gap scores and theses are on every paid plan. CSV export of listings is part of Pro Lifetime, or available on any plan with the $200 export add-on. See pricing.