Knowledge Base/How should I go about reporting a possible Massive data discrepancy?

How should I go about reporting a possible Massive data discrepancy?

Start a chat at massive.com/contact and paste the filled-in template below. Without it, support spends the first reply asking for the request URL, the request_id and the timestamps.

Start a chat at massive.com/contact and fill in the template below. Every line in it is there so we can do something specific with your report: rerun your exact request, find that request in our logs, or rule out the handful of causes that explain most reports. Send those details and the first reply is usually the answer. Send a description without them and the first reply is a list of questions.

Applies to

  • Plans: all plans.
  • Endpoints: any REST endpoint, WebSocket channel or flat file. The template has a section for each, because they identify a request differently.
  • Asset classes: all.

The template

The template is deliberately in sections, because REST, the WebSocket and the flat files identify a request in three different ways. Copy it into the chat, fill in what applies and delete the rest:

WHAT I SAW
  Ticker(s):
  Timestamp or date range (UTC):
  What the API returned:
  What I expected:
  Where my expectation came from:

HOW I ASKED (REST)
  Request URL (key replaced with apiKey=YOUR_API_KEY):
  request_id, or the x-request-id response header:
  adjusted= value, if this was an aggregates call:
  Which timestamp field I filtered on:

HOW I ASKED (WebSocket)
  Cluster URL:
  Exact subscribe message:
  UTC time of the events involved:
  A few of the raw messages received:

HOW I ASKED (flat files)
  Full object key, e.g. us_stocks_sip/trades_v1/2024/03/2024-03-07.csv.gz:
  The line or lines in question:

CONTEXT
  Plan and asset class:
  Does the same request still return the same thing right now?
  First noticed:

What each field lets us do

Nothing here is bureaucracy. This is what each line is for:

FieldWhat it lets us do
Request URL, key redactedRerun your exact request against the same endpoint. Most reports are settled here
request_id, or the x-request-id headerFind that single request in our own logs: which service answered it and what it returned
Ticker and UTC timestampsNarrow to the window. Many discrepancies are minutes wide, so a date alone is not enough
What you expected, and its sourceTurn "the data is wrong" into something testable. Another vendor, an exchange page, your own earlier pull
adjusted=Rule out the single most common false positive, a split-adjusted bar compared against an unadjusted source
Which timestamp field you filtered onRule out the second most common one. A trade carries participant, SIP and TRF timestamps, and they differ
Plan and asset classReproduce with the same entitlements you have. Results genuinely differ by plan
Still reproduces?Separate a live problem from one already corrected. Late prints and corrections arrive after the fact
Full object key, for flat filesFetch the same file. Flat files have no request_id

Where to find the request_id

Most REST responses carry request_id in the body, including error responses:

curl -X GET "https://api.massive.com/v2/aggs/ticker/AAPL/range/1/day/2026-09-10/2026-09-10?apiKey=YOUR_API_KEY"

Response

{
  "ticker": "AAPL",
  "queryCount": 1,
  "resultsCount": 1,
  "adjusted": true,
  "results": [
    { "v": 70011913.06091, "vw": 322.9591, "o": 316.67, "c": 326.57,
      "h": 326.74, "l": 316.51, "t": 1789012800000, "n": 1226854 }
  ],
  "status": "OK",
  "request_id": "e24b806121fda70c1633fb4bd32e8c6d",
  "count": 1
}

Two things in that block are worth naming. adjusted is echoed back, so the response tells you which of the two price series you actually got without your having to remember what you sent. And t is the bar's start in Unix milliseconds: 1789012800000 is 2026-09-10 00:00 Eastern, not 09:30, because daily bars are anchored to the start of the day.

The header is the one to rely on. x-request-id comes back on every call and carries the same value, while the body field is common rather than universal: /v1/open-close, /v1/marketstatus/now, /v1/marketstatus/upcoming and /v2/reference/markets all answer without it. So take the id from the headers with curl -D - and you never have to know which endpoint you are on:

curl -D - -o /dev/null -s "https://api.massive.com/v1/marketstatus/now?apiKey=YOUR_API_KEY" | grep -i x-request-id

Response

x-request-id: 851353d1a8d27cbea9c010ed0b05852d

There is no request_id on the WebSocket or on flat files. That is why the template asks for the subscribe message and the object key instead: those identify the request in the only way those two surfaces can.

Four checks that resolve most reports before you send one

Each of these takes a minute and each one accounts for a large share of reports that turn out not to be defects:

  1. Was the market open? For a past date, call grouped daily (/v2/aggs/grouped/locale/us/market/stocks/{date}): resultsCount 0 means no session. /v1/marketstatus/upcoming lists future holidays only. A short or absent session is the most common cause of data that looks missing.
  2. Are the prices adjusted the same way on both sides? Aggregate bars are split-adjusted by default. Add &adjusted=false to compare against an unadjusted source.
  3. Was the trade eligible? Not every trade updates a bar. A bar that disagrees with your own sum of the trades is usually sale conditions, not arithmetic.
  4. Has it been corrected since? A bar you stored live can be superseded. REST holds the current version, so compare against a fresh pull rather than your stored copy.

If one of those explains it, there is nothing to report. If none does, you now have most of the template filled in as a side effect.

Why we ask for the request_id rather than a screenshot

We ask for the request_id because it identifies the single request you made, including which service answered it and what it returned. A screenshot shows the symptom; the id shows the request. With it, a data question that would take a day of back and forth is often one lookup.

The same logic runs through the rest of the template: we are trying to get to the point where we can run what you ran, with the entitlements you have, against the window you meant, and see what you saw. Everything the template asks for is in service of that.

If you see an error

If a request returns a request_id you can quote, quote it, even for an error rather than a discrepancy.

Never paste a live API key into a chat, a ticket or a screenshot. Replace it with apiKey=YOUR_API_KEY. If you already sent one, delete it from your dashboard and make a new one.

You can also open an issue on our GitHub organization for problems with the client libraries. Data questions are better in the chat at massive.com/contact, where we can see your account.

Related


Was this article helpful?

Modernizing Wall St.

Reimagining financial market data for the 21st century.

Schedule a demo

Talk with our market data experts

Recommended

Can't find the answer you're looking for? Contact our team.
  • Title
    Are all indices available through the basic subscription tier?
    Data Coverage Category

    Data Coverage

    Indices Category

    Indices

    Description
    No, and the ones it withholds are the famous ones: SPX, DJI, RUT and VIX all need a paid Indices tier. Basic still reaches more than 10,000 of the 13,000-odd indices Massive publishes, and NDX and the Nasdaq Composite are among them.
  • Title
    Does Massive offer dark pool data?
    Data Coverage Category

    Data Coverage

    Stocks Category

    Stocks

    Description
    Yes. Off-exchange prints, which include dark pool executions, arrive on the ordinary trades feed and are identified by an exchange ID of 4. There is no separate dark pool endpoint, and no way to tell which individual venue a print came from.
  • Title
    Does Massive offer pre-market and after-hours data?
    FAQ Category

    FAQ

    Stocks Category

    Stocks

    Description
    Yes. Every extended-hours trade and quote is published, from 04:00 to 20:00 Eastern. There is no session parameter, so you select a session with timestamps, and expect far fewer aggregate bars outside regular hours because many extended-hours trades are odd lots that cannot set a price.