A sudden plunge in a price chart could be a bad record, a change in units or a corporate action—or a genuine market move. Treat an unusual observation as a reason to investigate, not as proof that the data is wrong. Preserve the original series, check the record and its context, and change or exclude a value only when you can document why.
How do I clean stock price data? Start with a review-first workflow
Because the asset, venue and sampling interval matter, there is no single cleaning rule for every financial time series. A trade-level feed, daily closing-price history and reference-rate calculation may need different checks. The consistent starting point is to retain the source data and distinguish flagging a value for review from changing the published series.
As an Amazon Associate I earn from qualifying purchases.
- Preserve the raw input. Keep an unchanged copy of the received records, along with the source and instrument identifiers. Work on a separate version so a later reviewer can trace what changed.
- Check basic record validity. Look for missing or non-numeric prices, values that are invalid for the instrument, and timestamps that are absent or implausible. The CFTC-filed reference-rate method described below rejects malformed messages, absent, non-numeric or non-positive prices, and future-dated execution times. Those are checks in that particular method, not universal rules for every instrument or data type.
- Verify what the record represents. Confirm the symbol or instrument identifier, currency, price scale, venue, timestamp convention and sampling interval. A symbol change, unit mismatch or timestamp interpretation can produce an apparent jump even when the underlying market did not move that way.
- Check adjustment status and events. Establish whether prices are raw or adjusted, and look for a relevant corporate action before interpreting a discontinuity as a crash.
- Flag and investigate unusual observations. Examine nearby records and compare with an independent source where available. A statistical alert is a prompt for this review, not a verdict.
- Correct or exclude only with a reason. Record the original value, the change or exclusion, the rationale, the method and the version of the resulting dataset.
The SEC’s Final Data Quality Assurance Guidelines emphasize assessing data collection and analysis and identifying supporting sources where appropriate. They call for disclosing data sources, quantitative methods and assumptions when needed to make influential information reproducible.
Recommended Free Tools
What should you check when a price looks wrong?
Work through these checks in order. The first asks whether the record is usable at all; the later checks ask whether the apparent anomaly makes sense in context and what action, if any, is defensible.
#1 Best Overall
Structural checks
- Is the price present and numeric?
- Is the value valid for the instrument and the field it occupies?
- Is the timestamp present and plausible for the feed’s convention and time zone?
- Do the instrument identifier, currency, price scale, venue and interval match the rest of the series?
These are examples of review questions, not a universal validation specification. For example, the CFTC-filed method rejects future-dated execution times for its calculation; a different dataset may define its own timestamp fields and validity rules.
Continuity and event context
- Inspect observations immediately before and after the flagged point. Check whether the change persists, reverses, or appears only in one record.
- Check for changes to the symbol, instrument identity, units, venue coverage or time convention that could create a discontinuity.
- Look for corporate actions relevant to the security and date. NYSE Regulation lists reverse splits and reorganizations among events affecting listed securities.
- Identify whether the series represents raw prices, split-adjusted prices, another adjusted price series, or returns. Also determine whether the analysis calls for price return or total return.
NYSE’s corporate actions and market watch information covers listed-security event notices. The appropriate treatment depends on the event and the intended analysis; the available guidance here does not establish one adjustment formula for all assets.
Rank #2
- Comes with secure packaging
- Easy to read text
- It can be a gift option
Cross-source checks
- Compare the observation with a separate source that covers the same instrument, venue or market and time interval, if one is available.
- Check that the comparison is genuinely like-for-like: different venues, timestamps, currencies or price definitions can explain differences without either source being wrong.
- Keep the identity of each source with the records so a comparison can be repeated and its limits understood.
In the UK, FCA rules require consolidated tape providers to identify incomplete or likely erroneous reports, use automated equity price and volume alerts, provide a way for users to flag concerns, and periodically reconcile received and published reports. These are requirements for UK consolidated tape providers, not general obligations on every analyst. The relevant provision is FCA Handbook MAR 9.2B; the cited requirements show effective dates of 31 July 2026.
Action logging
- Keep the raw record and the flagged or cleaned version distinct.
- Log the specific field or observation affected, the action taken and the reason.
- Document the sources, filters, assumptions and method version used to produce the output.
- Make clear whether a value was flagged, corrected, excluded or left unchanged after review.
The SEC guidelines describe the value of supporting sources, methods and assumptions for reproducibility. A clear record lets another analyst understand the transformation instead of having to infer it from the final chart.
Rank #3
- Ideal for Gifting
- Ideal for a bookworm
- Comes with Proper Binding
How can I tell whether a stock price crash was a bad tick?
No single visual pattern proves that a price is erroneous. Start by checking whether the record is structurally valid, then investigate nearby observations, instrument identity, units, timestamp conventions, venue and any event that could explain the move. A comparison with an independent, like-for-like source can add evidence, but a disagreement alone does not establish which feed is correct.
A dramatic move may be genuine. Deleting it solely because it is extreme can erase real market information; leaving a clearly malformed record in place can distort calculations. The defensible decision depends on the evidence and the purpose of the series. If the evidence is inconclusive, preserve the value and flag the uncertainty rather than silently removing it.
Why can generic outlier thresholds mislead?
A threshold that works for one kind of data can miss errors or flag legitimate moves in another. The right validation approach depends on what is being measured and how the result will be used.
Free tools Windows power users keep installed
One-click scans. No signup required.
- Data type and interval: trades, quotes, daily bars and reference rates have different structures and timing.
- Rule purpose: a check for malformed records is not the same as a statistical test for extremeness or a comparison across venues.
- Instrument behavior: sensitivity to legitimate jumps and illiquid instruments should be considered when calibrating any alert.
- Comparison quality: cross-source checks depend on whether a sufficiently independent and comparable feed exists.
- Action: a rule that flags a point for review carries a different risk from one that automatically changes or deletes it.
- Reproducibility: the rule and each resulting transformation should be explainable and repeatable.
There is no general, authoritative prevalence rate for bad ticks or false price observations established here, and no universal threshold is supported. Choose and document checks for the instrument, frequency and calculation objective rather than treating one cutoff as a market-wide standard.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What does the CFTC-filed reference-rate method illustrate?
A methodology filed with the CFTC for a specific bitcoin and ether reference-rate calculation combines record-validity checks with anomaly controls and a fallback. It is useful as an example of a defined process, not as a recommendation for cleaning ordinary stock-price histories.
- It filters invalid transactions, including malformed messages and records with absent, non-numeric or non-positive prices, as well as future-dated execution times.
- Within a partition, it excludes a venue VWAP that deviates by more than 10% from the median of the venue VWAP set.
- For the final reference rate, a deviation greater than 5% from a separate Lukka Global VWAP triggers calculation failure and fallback.
The 10% and 5% values are controls in that methodology, not accepted universal bad-tick thresholds. Its stages also show why anomaly handling is more than choosing a percentage: the calculation defines valid records, compares venue-level values, checks against a separate benchmark and specifies what happens if the final check fails. See the CFTC-hosted filed Data Quality Controls methodology for its exact scope and rules.
Should I adjust historical stock prices for splits?
First decide which series your analysis needs. Raw prices preserve the prices recorded at the time; split-adjusted prices account for a split so that the historical scale is comparable across the event. A price-return series and a total-return series answer different questions. Do not interpret a discontinuity until you know which convention the data uses and whether a corporate action occurred.
NYSE Regulation identifies reverse splits and reorganizations among listed-security corporate actions and provides event information through its market watch page. That establishes why event checks matter, but it is not a complete adjustment recipe for every security or asset class. State the series convention you use and follow a method appropriate to the event and analysis rather than applying an assumed universal formula.
How do you keep a cleaned series defensible?
Make the result traceable from the published series back to its inputs: retain the raw records, identify sources and instrument details, document rules and assumptions, and log every correction or exclusion with its rationale. The SEC’s data-quality guidelines call for supporting sources, quantitative methods and assumptions where needed to facilitate reproducibility. A cleaned dataset is defensible when a reviewer can see what changed and why.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
