Developing A Model For U.S. Shale Oil Production

| About: The United (USO)
This article is now exclusive for PRO subscribers.


The project life cycles for shale oil are extremely short, and the projects can be rolled out in a way typical of an assembly line.

This lends itself to constant improve in productivity, cost, and efficiency that will lead to ever decreasing costs.

Changes in rig efficiency, success rates, and production rates have been steadily improving from all datasets.

Finally, a model is developed for the Bakken field to estimate future production based on current rig data.

I have a firm belief that shale oil (NYSEARCA:USO) has changed the oil industry forever. The project life cycles are extremely short, and the projects can be rolled out in a way typical of an assembly line. As a mechanical engineer, I can't help but see the similarities of shale to the manufacturing business. In working through the data, it is evident that shale is improving in efficiency and cost effectiveness at an impressive rate.

In this article, I am going to explore the production from Bakken field since 2007 and attempt to develop a model for the play. The intended result will be to be able to analyze weekly rig counts and publicly available data to forecast forward production in a weekly/bi-weekly segment, but also to understand the history of development in the shale, in general, a little better.

I'll work to develop the model in a way that is easily replicated by someone else if they are desired. While the EIA already publishes a report that likely uses a similar method, I find it useful to have my own to explore what-if scenarios.

1 - Production History

The EIA provides a dataset of weekly rig count, initial rates per rig, legacy loss of production, and total production for each region. Each month the most recent month is added, and each item has been averaged over the month. Simple enough to get started, and we can use this as a baseline for our model. By using averages, we should also be able to get a good idea of what is happening across the entire formation without having to take into account out of sample wells.

One of the first things we notice is the rapid increase of well productivity as technology improved, companies moved to high-grade their reservoirs and costs dropped.

2 - Drilling Rig Efficiency

As prices have dropped, rigs have been placed on standby by the hundreds. In a Darwinian manner, the most inefficient rigs were idled, while the most efficient rigs remained in service. Furthermore, new drilling techniques are allowing those same rigs to drill faster, deeper and farther. While the number of rigs has dropped substantially, the number of wells drilled will have been reduced by a much thinner margin. Drilling techniques allow drilling of multiple wells from a single pad and multiple laterals from a single well bore. Advanced completions allow those same wells to produce at much higher rates.

Source : Rystad Energy

In the Bakken, we have seen a steady increase in the number of wells drilled per year from 12 to just above 20. By merging the data with the number of rigs from section 1, we can estimate the number of wells drilled per year. These figures only start from 2007, and I utilized the initial production from the Bakken in January 2007 as a starting point.

As of March the DMR has said that there are 13,024 operating wells in North Dakota. The inactive well count sat at 1,523 and the backlog sat at 920 bringing a total of 15,482, close to the models estimate of 16,197.

Developing the Base Model

To build the model, we need to get an idea of how many wells are producing. The Department Of Mineral Resources for the North Dakota state government gives us a separate dataset. There are also several thousand additional conventional wells and as of

Both datasets converge through the years, but I will be using the EIA statistics as my baseline model. I'll use a simple decline rate of 2%/month for the EIA production before the model begins.

Well Success Rates

Estimates for successful drilling percentages in the Bakken range from high (9 out of 10 wells), to very high (99 out of 100 wells).

"Makes you wonder... Is it even gambling if you win every hand?" - Energy And Capital

The truth likely lies somewhere in the middle. An EIA paper from 2010 titled " Quantifying Rig Efficiency" gives an overview up until that time. It appears the dry holes stagnated around the 10% mark, and likely suffered under the laws of diminishing returns thereafter. I'll be using a progression from 14% dry holes in 2007 to 8% dry holes today in the model.

Drilled Uncompleted Wells

To match the model, we need to remove DUCs. The DMR has made a post once a month regarding a backlog of wells. From the early days of reporting, there is about a 300 well inventory count; any number exceeding that I have assumed as a DUC. This estimate has lined up differently than the data provided by Bloomberg within a reasonable margin of error, it is not perfect, but the DMR data is direct from the source. The biggest anomaly occurs in September 2015, I haven't worked out the explanation as yet, but hopefully this can be sorted in the coming weeks.

Type Curve

I based the type curve off of a paper provided by the Post Carbon Institute. The article is dated Feb 2013, so the data is a bit dated, but I've based my curve for on this for now. I've been doing some research into how different vintages of wells (i.e., wells drilled in 2009,2011,2013,2015), etc. have changed over time to add some accuracy. However, for now, the main data points are the most recent ones.

The Result

Using the rig efficiencies, rig numbers, type curve, success rates, and accounting for DUCs I was able to simulate every well produced by the model over the 9 year period. Each well has its own decline curve and the cumulative sum provided the following output:

It performs reasonably well; the most recent six months has been anomalous. But I have a feeling that the EIA typically smooth the numbers out over time. There was a big spike in DUC's going into September 2015 which definitely should have had a more substantial impact on fluid rates than the EIA has shown.


I'm relatively comfortable with where the model sits about actual production, small adjustments over time will bring it closer to reality. I'll be working with the model to look at a few scenarios to estimate the supply response coming from shale. Please post below in the comments or IM if you have anything to add.

If you find this article interesting and would like updates in the future on these and other commodity related writings, head to the top of the article and hit the follow button next to my name. Thanks for reading.

Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.

I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.