Post-Auction Loss Review¶
Retrospective review — instantiates Winner-Conditioned Valuation Correction
Logs what you bid, whether you won, and how the asset actually performed across many contests, then reads the pattern of wins, losses, and regrets to reveal whether you are shading too little or too much.
Every other mechanism in this archetype acts before a bid; this one closes the loop after, across many bids at once. Post-Auction Loss Review is a standing retrospective that logs each contest — the bid, the win-or-loss, and, once it is knowable, how the asset actually performed — and reads the aggregate pattern to calibrate the whole bidding discipline. Its one defining move is taking the losses seriously. A lost contest looks like nothing happened, but if the winners of the deals you lost went on to regret them, your losses were correct — evidence your shading is well-calibrated, not a string of misses to be embarrassed about. Conversely, a high win rate paired with winners' remorse says you are winning the wrong contests: shading too little and catching the curse. By separating good losses from bad and comparing win rate against realized value, the review turns a scatter of individual outcomes into a signal about whether the correction upstream is set too aggressively or not aggressively enough — and it reconciles cumulative overpayment against a pre-set portfolio loss budget so the program can't quietly bleed.
Example¶
A demand-side platform bids into millions of real-time ad auctions a day, and each impression's true worth only reveals itself later as conversions land. The Post-Auction Loss Review aggregates a quarter. It wins ≈30% of the auctions it enters — but the impressions it wins convert ≈15% below the value its pre-bid model assigned them, while a sample of the auctions it lost went to rivals at clearing prices that, judged by later conversion rates, were overpayments. Read together, the verdict is uncomfortable and clear: the platform is winning the wrong auctions. Its bids aren't shaded enough, and the curse is showing up as systematically disappointing wins.
The review also tallies the quarter's realized overpayment against the program's ≈$2M loss budget and finds it roughly ≈80% consumed with weeks to go. Two corrections follow: deepen the winner's-curse shading in the bid model, and trip the budget alarm to throttle bidding before the cap is breached. Neither would have been visible from any single auction — the signal only exists in the pattern.
How it works¶
- Log every contest. Record the bid, the win/loss, and the later-realized value for each, building a standing record rather than a one-off post-mortem.
- Separate good losses from bad. Distinguish losses where the winner later regretted (correct discipline) from losses that were genuine missed value, so a healthy walk-away rate isn't mistaken for failure.
- Compare win rate against realized value. Set the frequency of winning beside how wins actually performed; a high win rate with poor realized value is the fingerprint of under-shading.
- Reconcile against the loss budget. Tally cumulative realized overpayment against the pre-committed portfolio budget and raise the alarm as the cap approaches.
- Feed the calibrated signal forward. Return the finding to the upstream correction as a "shade more / shade less" adjustment.
Tuning parameters¶
- Review cadence — per-contest, quarterly, or rolling. Frequent review reacts faster but on thinner, noisier samples; slower review is stabler but lags a drifting bias.
- Outcome maturation window — how long to wait for realized value before scoring a contest. Waiting longer buys accuracy at the cost of slower learning.
- Loss-budget size and alarm threshold — how large an aggregate overpayment the program tolerates and how early the alarm trips.
- Attribution granularity — whether calibration is read in aggregate or sliced by sector, field size, or deal size, localizing where the shading is mis-set.
When it helps, and when it misleads¶
Its strength is that it is the only mechanism here that can tell you the correction is wrong. Without a retrospective, over-shading is invisible — it shows up only as contests you silently didn't win — and under-shading gets excused as bad luck one deal at a time. Calibration requires realized feedback[1], and this review is where that feedback is gathered and turned into a directional adjustment to the shading.
Its failure modes come from the asymmetry of the data. Wins have observable outcomes; losses' counterfactual value must be inferred, so a review that only scores its wins suffers survivorship bias and will happily conclude the discipline is fine. Slow-maturing outcomes delay every lesson, and the review is easy to run backwards — narrating a bad bid as an unlucky draw rather than a mis-calibrated one. The discipline that guards against this is to track losses' inferred outcomes alongside wins, and to pre-commit the loss budget so the review reconciles against a fixed cap instead of rationalizing each breach after the fact.
How it implements the components¶
This review realizes the feedback-and-accountability side of the archetype — the components that only a retrospective across many contests can fill:
outcome_calibration_record— it maintains the standing record of bids, wins and losses, and realized outcomes, and reads it for whether the shading is calibrated.portfolio_level_loss_budget— it reconciles cumulative realized overpayment against a pre-set aggregate budget and raises the alarm as the cap is approached.
It looks backward and sets no pre-bid number: the winning-conditional correction is Winner's-Curse-Adjusted Bid Model's and the outside-view estimate is Reference-Class Bid Review's. This review only produces the realized-outcome record those forward-looking mechanisms should learn from.
Related¶
- Instantiates: Winner-Conditioned Valuation Correction — this review is the backward-looking loop that calibrates the whole discipline against realized outcomes.
- Sibling mechanisms: Reference-Class Bid Review consumes the outcome record this review builds · Winner's-Curse-Adjusted Bid Model · Reserve Price or Walkaway Limit · Sealed-Bid Premortem · Bid/No-Bid Gate
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: Post-Auction Loss Review operates as a bounded evaluation of existing evidence or work that produces a finding or disposition because it logs what you bid, whether you won, and how the asset actually performed across many contests, then reads the pattern of wins, losses, and regrets to reveal whether you are shading too little or too much.
Independent corroboration: The frozen evidence defines Post-Auction Loss Review as 'Logs what you bid, whether you won, and how the asset actually performed across many contests, then reads the pattern of wins, losses, and regrets to reveal whether you are shading too little or too much', so its operative form is Assessment, Review & Assurance.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Economics & Finance
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Specialized
Rationale: Reviewing bids against subsequent value arises from auction theory and the economics of winner-conditioned valuation.
Related originating lineages:
- Behavioral Economics — Behavioral economics contributes analysis of regret, overbidding, and systematic judgment bias across contests.
Review resolution: Light authoritative-source research resolves the primary-origin disagreement in favor of economics finance. American Economic Association: Anomalies—The Winner's Curse directly documents the defining practice or theory described in the selected origin rationale. Other domains are retained only where the blind reviews identify material co-development or translation; broad application is recorded separately as domain_reach=specialized, while origin_mode=cross_disciplinary_synthesis describes the relationship among origin lineages.
Attribution caveat: The boundary with behavioral economics is substantive because that tradition materially developed or translated part of the mechanism; the cited provenance places the defining form in economics finance.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
Notes¶
The calibration record this review produces is the input that closes the archetype's loop: Reference-Class Bid Review reads it to build its reference classes, and the bid model reads it to reset how far to shade. The loss review sits at the end of one contest and feeds the front of the next — which is why it is worth running as a standing program rather than a one-off post-mortem after a painful win.
References¶
[1] Kagel, J. H., & Levin, D. "The Winner’s Curse and Public Information in Common Value Auctions". American Economic Review 76(5), 894–920 (1986). Shows that calibration in repeated common-value auctions depends on realized outcome feedback. registry ↩