Art 2: The auction business model

View study sheet (PDF) View SQL cheat sheet (PDF)

This session turns from artwork pricing (Art 1) to the business that runs the saleroom: the auction house. An auction house is a two-sided intermediary, paid by both the seller who consigns a lot and the buyer who wins it, and its health as a business is only partly visible in any single sale. Students learn the two revenue streams that fund a saleroom, the mechanics that shift risk between house, seller, and buyer before a single hammer falls, and the competitive reality that the real contest among houses happens before the sale, in winning good consignments, not during it. The session then returns to documented fields already met in Art 1, usd_price_decimal, sale_estimates_high_usd_price, status, and vendor, to build the two measures a business analyst can actually compute from public auction records: how well a house's estimates track what buyers pay, and how concentrated or spread out a house's revenue is across its lots. The session builds toward a single question: what does it take to run a profitable saleroom, and what can transaction data tell an outside analyst about whether a given house is succeeding at it.

Target course(s) and level

Art business, auction studies, or arts administration course, and a natural extension for a business or finance elective examining intermediary and marketplace business models. Suitable for graduate students in art market or arts management programs, advanced undergraduates pairing art history or economics with business, and professional or continuing-education students entering the trade. No finance background is assumed; students who have completed Art 1 will move faster through the data sections.

Learning objectives

By the end of this session, students will be able to:

1. Explain the auction house's role as a two-sided intermediary and describe the buyer's premium and the seller's commission as the two revenue streams that fund the business.
2. Describe how a pre-sale estimate is set, why estimate accuracy matters commercially to a house, and compute the ratio of realized price to high estimate from documented fields.
3. Explain guarantees and irrevocable bids as mechanisms that shift financial risk between house, consignor, and a third-party guarantor, and how each changes the incentives around a sale.
4. Compute average lot value by house and contrast it with lot volume, and explain why these describe two different kinds of auction businesses.
5. Measure value concentration among a house's top lots and explain why a single lot or a single consignor can dominate a season's results.
6. Identify the limits of auction data for studying the auction business itself, including that it captures the secondary market only, that no field records buyer's premium, hammer price, or commission, and that the newest periods are still being ingested and should not be read as trends.
7. (Extension) Retrieve one house's sold lots programmatically via the production API and compute estimate accuracy or average lot value client-side after paginating the results.

Prerequisites

Art 1 (art market fundamentals) or equivalent familiarity with the fine art data table, the vendor and designer fields, and the vocabulary of estimate, realized price, and primary vs secondary market. Students who have not taken Art 1 should read its session outline before class. No prior exposure to SQL or finance is assumed beyond that. The optional code extension assumes basic Python familiarity.

Materials and access needed

• Sandbox access at sandbox.altfndata.com, self-registered with a work or school email, auto-approved.
• Projector or screen share for the instructor demo.
• The coverage browser tab, used before the demo to confirm how a chosen auction house appears in the vendor field.
• For the optional extension only: an instructor class API key requested from info@altfndata.com, and the downloadable Python client (altfndata_client.py) or tutorials notebook (altfndata_tutorials.ipynb) at docs.altfndata.com.

Session outline (90 minutes)

• 0 to 15 min: The auction house as a business. Introduce the buyer's premium and seller's commission as the two revenue streams, and explain why winning consignments, not running the sale itself, is the real competitive contest among houses.
• 15 to 25 min: Estimates, guarantees, and irrevocable bids. Explain how a pre-sale estimate is set, why the house has a commercial stake in getting it right, and how a guarantee or an irrevocable bid shifts risk before the sale even opens.
• 25 to 45 min: Sandbox orientation and guided demo, estimate accuracy. Confirm the documented fields on the fine art table and build the ratio of realized price to high estimate for a chosen house.
• 45 to 60 min: Guided demo, average lot value versus lot volume, contrasting a high-value house against a high-volume house.
• 60 to 75 min: Small-group exercise, students pick a different house and compute its value concentration among its own top lots.
• 75 to 85 min: Class discussion, groups report what estimate accuracy, average lot value, and concentration together suggest about the kind of business their house runs.
• 85 to 90 min: Wrap-up and homework assignment.

In-class demo (sandbox-first, no code)

1. Open sandbox.altfndata.com, sign in, and select the fine art data table from the SQL editor dropdown.
2. Open the coverage browser tab and confirm how a major auction house is written in the vendor field. Spelling matters for exact filters.
3. Before running any query, walk the class through the two revenue streams conceptually: the buyer's premium, a fee the winning bidder pays on top of the hammer price, and the seller's commission, a fee the house deducts from what the consignor receives. Note that neither figure appears in the data students are about to query, only the final realized price the buyer paid.
4. Run the first guided query, estimate accuracy for a chosen house, and explain that the ratio of usd_price_decimal to sale_estimates_high_usd_price tells students how often the house's own pre-sale number tracked what buyers actually paid, which is a direct read on the skill of the specialists who set that estimate.
5. Explain guarantees and irrevocable bids conceptually at this point: a guarantee promises the consignor a minimum price regardless of the room's bidding, and an irrevocable bid commits a third party to buy the lot at a set price if no one else bids higher. Both remove downside risk from the consignor before the sale, in exchange for giving up some upside, and both are a house's tool for winning a prized consignment away from a competitor.
6. Run the second guided query, average lot value by house, and explain that dividing total realized value by the count of sold lots separates a house that competes on a handful of very expensive lots from a house that competes on a large volume of moderately priced ones. Ask the class which model looks more resilient to a single lot failing to sell.
7. Run the third guided query, value concentration among a house's own top lots, and explain that a season's results can be dominated by one exceptional consignment, which is why the consignment pipeline, not the sale night itself, is where a house's business is really won or lost.
8. Explain the two caveats before anyone reads the figures as settled fact: the data captures the secondary market only, so it says nothing about private sales, gallery activity, or a house's fee schedule, and the newest periods are still being ingested, so the most recent quarter should never be read as evidence of a trend.

Datasets and queries used

Dataset: fine art data (documented fields: designer, model, item_title, sale_date, usd_price_decimal, sale_estimates_high_usd_price, status, vendor, stock_ticker). In the fine art data the designer field holds the artist or maker name, and the vendor field holds the auction house. Students who want to repeat any of these queries on a second table can run the same pattern against all_works_of_art_data, which shares the same documented fields.

Query 1, estimate accuracy for a chosen house:

SELECT vendor,
       approx_percentile(usd_price_decimal / sale_estimates_high_usd_price, 0.5) AS estimate_accuracy,
       COUNT(*) AS sold_lots
FROM all_fine_art_data
WHERE status = 'sold'
  AND sale_estimates_high_usd_price > 0
  AND vendor LIKE '%Christie%'
GROUP BY vendor;

Query 2, average lot value versus lot volume by house:

SELECT vendor,
       COUNT(*) AS sold_lots,
       SUM(usd_price_decimal) AS total_realized_usd,
       SUM(usd_price_decimal) / COUNT(*) AS avg_lot_value_usd
FROM all_fine_art_data
WHERE status = 'sold'
GROUP BY vendor
HAVING COUNT(*) >= 200
ORDER BY avg_lot_value_usd DESC
LIMIT 25;

Query 3, value concentration among a house's top lots:

WITH house_lots AS (
  SELECT usd_price_decimal
  FROM all_fine_art_data
  WHERE status = 'sold'
    AND vendor LIKE '%Sotheby%'
),
ranked AS (
  SELECT usd_price_decimal,
         ROW_NUMBER() OVER (ORDER BY usd_price_decimal DESC) AS rnk,
         COUNT(*) OVER () AS total_lots
  FROM house_lots
)
SELECT SUM(usd_price_decimal) FILTER (WHERE rnk <= 20) * 1.0
         / SUM(usd_price_decimal) AS top_20_share_of_value,
       MAX(total_lots) AS sold_lots
FROM ranked;

Optional API extension, retrieve one house's sold lots to compute estimate accuracy or average lot value client-side (POST /v1/tables/all_fine_art_data/query, header X-API-Key). The API returns rows rather than server-side aggregates, so the ratios and totals are computed client-side after paginating on offset:

{
  "fields": ["vendor", "designer", "item_title", "sale_date", "usd_price_decimal", "sale_estimates_high_usd_price"],
  "filters": [
    {"field": "vendor", "op": "eq", "value": "Christie's"},
    {"field": "status", "op": "eq", "value": "sold"}
  ],
  "sort": [{"field": "sale_date", "direction": "desc"}],
  "limit": 1000
}

Discussion questions

1. The buyer's premium and the seller's commission are the two revenue streams that fund an auction house, yet neither figure appears in the transaction data. What can an outside analyst still infer about a house's revenue without ever seeing its fee schedule?
2. Why does estimate accuracy matter commercially to a house, beyond simply making the specialists look good on sale night?
3. A guarantee removes the consignor's downside risk before the sale. Who absorbs that risk instead, and what does the house gain by offering one?
4. An irrevocable bid and a guarantee both shift risk before a sale opens. How do they differ in who is taking on the exposure, and why might a house prefer one over the other for a given lot?
5. Two houses show similar total realized value in a season, but one has a much higher average lot value and the other a much higher lot count. What different businesses are these two houses actually running?
6. If one house's top twenty lots account for most of its realized value in a season, what does that concentration imply about the risk in its business, and what would you want to know about those twenty lots before drawing a conclusion?
7. The text describes winning good consignments as the real competition among houses, more than the sale itself. What evidence in estimate accuracy, average lot value, or concentration would support or undercut that claim for a given house?
8. Auction data captures the secondary market only and has no field for buyer's premium, hammer price, commission, or guarantees. What would a complete picture of a house's business require that this dataset cannot provide?

Homework assignment

Each student selects one auction house represented in the fine art data and writes a two-page business brief on that house, as if preparing a competitive analysis for a rival saleroom's management. The brief must report the house's estimate accuracy, its average lot value against its lot volume, and its value concentration among its own top lots, and must argue, using those three figures together, what kind of auction business the house runs and where its consignment strategy appears to be working or under strain. The brief must include the SQL queries used as an appendix, at least one exported table or chart, and an explicit limitations section that states the data is secondary market only, notes that no field records buyer's premium, hammer price, commission, or guarantees so none of those can be computed, and does not draw conclusions from the most recent quarter given ongoing ingestion. Grading criteria: correct computation of estimate accuracy with the required sale_estimates_high_usd_price > 0 guard (25 percent), correct computation and contrast of average lot value and concentration (25 percent), quality of the business argument connecting the three figures to a coherent read of the house's strategy (30 percent), and honest treatment of the analysis's limitations (20 percent).

Going deeper

Key terms

• Buyer's premium: the fee a winning bidder pays on top of the hammer price, one of the two revenue streams that fund an auction house. Not present as a field in the data.
• Seller's commission: the fee a house deducts from the proceeds paid to a consignor, the second of the two revenue streams. Not present as a field in the data.
• Pre-sale estimate: the auction house's own published price range for a lot before the sale, recorded here only at its high end, in sale_estimates_high_usd_price.
• Estimate accuracy: the ratio of realized price to the high estimate across sold lots, a measure of how closely a house's pre-sale expectation tracked what buyers actually paid.
• Guarantee: a house's or a third party's promise to pay a consignor a minimum price for a lot regardless of the room's bidding, which removes the consignor's downside risk before the sale.
• Irrevocable bid: a commitment from a third party to buy a lot at a set price if no higher bid emerges, functioning as a form of guarantee negotiated ahead of the sale.
• Private sale: a sale negotiated directly between a house and a buyer outside the public auction room, a second channel alongside public auction that this dataset does not capture.
• Consignment pipeline: the ongoing process by which a house sources lots from sellers, widely considered the real point of competition among houses rather than the sale night itself.
• Average lot value: total realized value divided by the count of sold lots for a house, distinguishing a business built on a few expensive lots from one built on high volume.
• Value concentration: the share of a house's or a season's total realized value that comes from a small number of top lots or top consignors.

Common pitfalls

1. Writing or expecting a query that computes buyer's premium, hammer price, commission, or a guarantee amount directly from the data. No such field exists on the fine art table, and usd_price_decimal is the realized price the buyer paid, not a component fee.
2. Computing estimate accuracy without the sale_estimates_high_usd_price > 0 guard, letting zero or null estimates distort the ratio.
3. Treating a high average lot value alone as evidence of a stronger business than a high-volume house, without considering how concentrated that value is among a few lots.
4. Reading a season dominated by one exceptional lot as the house's normal, ongoing level of business, rather than checking whether that concentration repeats across seasons.
5. Reading the newest quarter's totals as evidence of a rising or falling trend, when the newest periods are still being ingested.
6. Carrying the fine art table's designer-equals-artist, vendor-equals-auction-house convention into another table, where designer and vendor mean something different.
7. Ranking a house on too few sold lots, producing an estimate accuracy or concentration figure too noisy to support a real business conclusion.

Additional queries to explore

1. Estimate accuracy over time for one house, grouped by year extracted from sale_date, to see whether a house's pre-sale expectations have tracked realized prices more or less closely across different market conditions, while still treating the newest year with the recency caveat.
2. Sell-through (sold lots over offered lots) alongside estimate accuracy for the same house, to see whether a house that prices estimates conservatively also clears a higher share of its offered lots.
3. Value concentration by top consignor proxy, comparing the share of total value held by a house's top fifty lots against its top twenty, to see how quickly concentration falls off outside the very highest lots.

Extension activities

1. Have a group repeat the API extension for two different houses and compare their estimate-accuracy and average-lot-value figures as retrieved programmatically, rather than through the SQL editor.
2. Ask students to sketch a one-page saleroom business memo combining estimate accuracy, average lot value, and concentration into a single view, as a preview of Data 2 (visualization).
3. Have students research one house's publicly disclosed buyer's premium schedule or guarantee policy, where available in trade press, to compare qualitative fee and risk practices against what the transaction data alone can show.

Connections to other modules

Previous in this track: Art 1 supplies the price-factor and estimate-vs-realized vocabulary this session builds on. Next: Art 3, the business of the art market, widens from one house's mechanics to league tables and artist-level demand across the trade. Art 4 (auction theory) explains why estimates anchor bidding; Art 5 extends estimate-accuracy and concentration into formal research methods. Cross-track: Data 2 for presenting a saleroom memo as a chart.