Posted Quantities, Not Estimates

The posting is the fact. The national total is a model.

A word on terms. A scheduled quantity is a nomination that cleared the cycle, not metered flow; measured “actuals” post days later. Every daily interstate flow series in the market, ours included, is built from scheduled quantities. That is not the difference. The difference is what happens next.

Derived national supply and demand figures are a sample of postings, that are then classified into regions, and then multiplied by a factor to scale-up. The result is a useful number, and on a settled gas day it is probably close. But it is a model output, and a model is only as good as its last recalibration. A parser that breaks, a meter that gets reassigned, a plant that switches suppliers: each of those moves the multiplier, and none of them shows up in the published figure as anything other than a market move.

We tested this on our own data before deciding not to do it. Scaling our scraped totals to match a national estimate produced a number that was, on most days, simply the national estimate with our noise added. On the flat series, “same as yesterday” beat it outright. It told a trader nothing the estimate would not tell them a few hours later, and it hid the one thing our data is actually good at.

What the posted number is good at

A change. When a pipeline’s scheduled delivery into an LNG terminal drops 400,000 Dth between the evening cycle and ID1, that is not a model adjusting. It is a pipeline telling the market, in a filing, that less gas is flowing there today. Enercast shows you that change at the meter where it happened, on the cycle it happened, from the posting it came from. You can click through to the source. You cannot click through to a multiplier.

Three things follow from that.

• Timing. We publish the day’s first figures shortly after the timely and evening cycles close and update as intraday cycles land, ahead of the settled national estimates. On weekends and holidays, when those estimates do not update at all, the postings still do, and so do we.
• Granularity. National estimates arrive as national estimates. Our figures are built from named points on named pipelines, so the same data that gives you a national power-burn change also tells you it was Florida and the Southeast, not the Northeast, and which interconnects carried it.
• Honesty about what moved. Every figure we store is stamped with a fingerprint of the meter registry that produced it. When we add coverage, the fingerprint changes, and the site refuses to print a day-over-day change across that boundary. We would rather show you a dash and a note than a fake 20% move that was really us adding a pipeline. If you have used flow data for long, you have been burned by the other approach.

What we do not cover, and what we are doing about it

Postings only exist where posting is required, so our totals are a sample of the interstate system, not the whole country. We say this on every page rather than papering over it with a multiplier. The main gaps are known and named:

• Texas intrastate pipelines carry most of the gas to Mexico and a large share of Gulf Coast production, and they have no federal posting obligation. No one measures those flows daily — vendors included; their Mexico figures are inferred too.
• Behind the citygate. Gas delivered to a local distribution company is metered once, at the citygate. The power plants and industrial customers behind that gate are invisible to everyone reading pipeline postings.
• California. The two large California utilities operate their own transmission systems and do not post in the same way, so California demand is under-represented in our power and citygate figures.

We are actively closing the part of that gap that can be closed. Every week we are mapping additional meters on the pipelines we already scrape and adding new pipelines, and each addition goes through the same three-day validation and mirror test before it counts. Our coverage of the interstate system is materially higher than it was a month ago and will keep rising. What we will not do is invent the rest. If a number on Enercast changes, it is because a pipeline changed what it posted.

How a meter earns its place

Summing postings is only as good as the list of points being summed. Interconnects post on both sides, two-way points net, and a nomination revises across cycles. Every point in our registry clears the same gate before it counts:

• Mirror test. A candidate whose series matches an existing point to the Dth is the other side of a meter already counted. It is recorded as a mirror and excluded, never summed twice.
• Three gas days. A point has to post real, non-zero volume on at least three consecutive settled days before it enters a total. A single-day match proves nothing; we learned that the expensive way.
• Same cycle, same basis. Day-over-day changes compare like with like: the same cycle on both days, and the same registry. If the registry changed in between, the change is withheld and the cell says why.
• Daily drift check. An automated pass reads every registered point against the database each day and flags sign contradictions, duplicated meters and points that have stopped posting.

A category total is still a sample, and a reroute from a pipeline we cover to one we do not will show as a decline. That is why every total opens into its meters: the point-level detail is the product, and the total is the index to it.

How to use it

Read Enercast for direction and location: what moved, where, how much, and when it was posted. If you need a single national level for a settled day, an estimate is the right tool, and there are good ones. If you need to know at 6am that something changed at a specific point on a specific pipeline, and to see the filing that says so, that is what we built.

Back to the Daily Supply & Demand Summary →