Tuesday, November 10, 2015

Flotek: a plea for accuracy

This post had a material impact on the stock: the company admitted to most the important allegations in this post and the stock halved. My view is that there is much more to do. Details in the post script. 

Warning: this is a really gonzo story. We tried to check the data and use the main marketing tool for a billion dollar company. 
None of it worked out. The data looks like it was made up or at least systematically rigged. The marketing tool (an iPad app) simply crashes and is non-functional. 
We can't tell whether the whole company is this suspect - but this is as non-sensical as anything we have seen on Wall Street. The closest comparable on this blog was Universal Travel Group, a stock which is now delisted and where the principals settled civil fraud charges.
If you own the stock you must work this to the end. Try to reproduce our results. If this is the whole story this stock will trade with a low single digit handle or lower.

John W. Chisholm is a fan of accuracy. Demands it, even. As the CEO of Flotek Industries, Inc., (NYSE:FTK) he requires factual, mathematical accuracy.

How do I know? Because Chisholm devoted a bit of time on the recent 3Q conference call to express his apparent displeasure with a certain unnamed analyst who, it seems, had chosen a wayward path:

Before I turn the call over to Rob Schmitz to provide additional financial review, I want to take a minute to discuss an issue of interest to all Flotek investors of which all of us around this table today are. We appreciate the thoughtful work and effort that many of the analysts who cover Flotek provide and we understand how difficult the economic analysis of a company like Flotek will be, especially in an environment with such high levels of uncertainty. That said, the challenges in the analysis business also create challenges for companies like Flotek, especially when one erroneous estimate can have such a disproportionate impact on consensus numbers which is what happened in the third quarter.

Removing just that one estimate from the mix provides a more cogent view of consensus thinking and makes Flotek's third quarter results not only at or near the top of the service sector, but slightly ahead of mainstream consensus. Please understand, we appreciate the effort that goes into complete and accurate coverage of our Company and have no interest in a debate over valuation and rating calls that are based on factual and accurate information. That is what makes the US public market so robust.

However, accurate-based mathematics, factual statements that don't involve blind conjecture or guesses as to material facts such as customers or key chemistry performance and correct statements of fact about items such as the current status of credit agreements and the like that can easily be located in our public filings seem like basic tenants of objective, thorough securities research. While we have little control over what is published in today's world, I would say we welcome firms and analysts that have an interest in factual, constructive evaluation and we'll feel compelled going forward to mention those who appear to potentially [have] agendas other than objective securities analysis and critique.

We share Mr Chisholm's interest in “factual, constructive evaluation.”

Flotek, the Houston-based company he heads, supplies chemicals and production equipment to the energy industry. The market for those products looks a little bleak at the moment. However Flotek claim exception due to special products and special technology. They have grown regardless. Flotek is a beacon in these harsh times.

Flotek is keen to tell you how they did it. On 11 September in New York it gave the company's fifth investor presentation since mid-August, and again it gave the story. 

The attraction of the stock (and the miracle of Flotek's business resilience) comes from Flotek’s citrus-based “complex nano-fluid” (“CnF”) surfactants. These, they say, boost the production of oil and gas from hydraulic fracturing of existing and newly-completed wells. 

Here’s John Chisholm describing it to Jim Cramer in February 2015:

In recent quarters (and in the video) the company has highlighted FracMax, an iPad app created as a sales tool to demonstrate the complex nano-fluids’ capabilities.

Chisholm has been lavish in his praise for FracMax, assigning it almost sole credit for the Company’s industry-downturn-defying increases in chemical sales. At a Barclays conference on 10 September, Chisholm stated (p. 3):

While like other companies with an energetic focus, we are not immune to the activity swings resulting from powerful commodity price cycles. Our energy chemistry business, for example continues to grow, even as completion activity weakens. This anomalous and unprecedented growth is the result of continued market penetration and the power of FracMax, our proprietary data analysis and visualization software which shows, using producers' own empirical data, the compelling economic benefits of our customized chemistry.

Moreover, we do not expect that growth to slow. Even in this more challenging environment, as data supporting the efficacy of our technologies provide ample opportunities for progress as exploration and production companies understand that the optimal well results only when they come and put Flotek in their wells.

The Company’s SEC filings have been similarly kind to FracMax. Last year’s 10-K noted (p.24-25):

Energy Chemical Technologies revenue for the year ended December 31, 2014 increased…33.8%...primarily due to the increased sales of stimulation chemical additives, largely the result of the introduction of the Company’s proprietary, patent-pending FracMax™ software which statistically demonstrates the positive production and economic impact of using Flotek’s CnF® chemistries in unconventional well completions. The FracMaxTM software has led to a record number of new validation projects and accelerated commercial acceptance of the Company’s CnF® completion chemistries.

During the 11 September presentation, Chisholm showed a video (seen below) that proclaimed:

It's more impactful and convenient than ever to experience the real-world benefit of CnF.  An innovation that's revolutionizing an industry by truly making a difference. You can touch it, you can see it, and now, you can believe it. Your data has become intelligent in a way never imagined. The future is here.

Have a look:


Put simply, FracMax matters to Flotek. And precision matters to FracMax.

And as any true fan of mathematical precision knows, the proof is in the showing, not the telling. So Chisholm demonstrates the analysis Flotek hopes potential customers will perform using FracMax. 

Let's go through one of his presentations and check the data.

Chisholm compares four wells in Texas: one that used CnF in its completion (“Molnoskey 1H”) and three nearby wells that did not (“Targac 1H,” “Gillespie 1H,” and “Berger 1H”).

Digging into the Data

According to Flotek, the FracMax app visualizes data derived from public sources, such as FracFocus.org (“FracFocus”) and state oil and gas production records.  Slide #50 from February’s investor day presentation gives a bit more detail (highlighting added):

FracFocus.org is a centralized database that collects self-reported information on the chemicals used by energy companies in hydraulic fracturing (“fracking”). FracFocus’s data is available in a machine-readable format, so with effort anybody—not just Flotek—can reproduce FracMax-style analysis, provided that the chemical data is joined to production data by a common industry identifier such as an API number.

For the wells Chisholm highlighted in the presentation, production data comes from the Texas Railroad Commission (“RRC”), which collects information on Texas energy production. Chisholm emphasized the independence of the data he presented on 11 September, stressing that it had been reported directly to FracFocus and the Texas RRC and was “un-adjusted” (Slide 34, highlighting added).

Chisholm could hardly have declaimed the data’s provenance more loudly. For instance, Chisholm took a shot at critics at 46:35 in the replay:

It was reported three or four weeks ago in a comment that we claim performance of CnF. That's not right. We report the performance, and there's a big difference. We don't create the data, we allow the visualization of the data in a way that people have never seen before. [emphasis in original]

Chisholm again proclaimed the data’s bona fides (Source: 1:01:53 in linked video):

That's why the data backchecks and validates. This is data coming right from the Texas Railroad Commission. Forget about FracMax. Just forget about it completely. That's data that comes right from the Texas Railroad Commission.

Let’s have a look.

The four proximate wells Chisholm highlighted were:

Slide #   
TX Lease#
Well Name
58, 59
Targac 1H
Sabine Oil & Gas LLC
60, 61
Gillespie 1H
Sabine Oil & Gas LLC
62, 63
Molnoskey 1H (CnF)
Sabine Oil & Gas LLC
64, 65
Berger Unit 1H
Devon Energy Corp

Targac 1H

Targac 1H from Sabine Oil & Gas (API#: 42-285-33654) was selected for its location within four miles of the CnF-using Molnoskey 1H (see map on slide 58). Targac, said Chisholm, is an example of a large, expensive well that used substantial amounts of water.

Slide 59 shows production and additive data from FracMax:

Chisholm narrated at 1:07:20 in the replay

So that particular well there, here's what we want you to focus on. 11 million gallons were pumped in that frack job. Here's the initial production: eight thousand eight hundred barrels in the first month, no CnF in that well. 'Kay? 11 million gallons, 8800 barrels.

The RRC reports production data by RRC ID#, to which well API#s directly correspond for the four wells of concern to us. So first we look up the RRC ID using the API #:

Using the RRC’s query tool, we lookup the RRC production data for lease ID# 10702

Let’s compare the numbers.

The table below shows, by month, the production data for oil and gas from Targac 1H. The first column under each of Oil and Gas headings, labeled “FTK – FracMax,” reproduces the numbers as shown on Chisholm’s slide above that was presented to investors. The middle column under each of the Oil and Gas headings, labeled “Texas RRC” shows the original production data (not disposition) as reported by the Texas RRC and reproduced in full above. The final column under each of the Oil and Gas headings simply divides the FracMax number into the Texas RRC number, stated as a percentage.

Eagle-eyed readers—who no doubt share our commitment to correctness—will note several problems here. 

  • The production data shown on Chisholm’s slide does not match what was reported to the RRC. 

  • The data appears to have been adjusted, as the numbers shown on the FracMax slide total only 60% of the oil production reported to the RRC and 50% of the gas. 

  • The first month’s oil production was not 8900 barrels in December 2013, but 17,761 in November 2013. 

Chisholm it seems understated production data for the well that contained none of Flotek's special "complex nano-fluid". 

We gave John Chisholm the benefit of the doubt. We thought this could be a coding or data entry or a simple copy-paste error. 

Alas we found that this is a pattern rather than isolated error.

Gillespie 1H

Chisholm next presented the Gillespie 1H well (API# 42-285-33673), also from Sabine Oil and Gas, located nearby (see slide 60).

At 1:07:39 in the replay, Chisholm describes Gillespie 1H as a “clone” of the Targac 1H well, as Sabine used a similar set of chemicals:

The operator, Sabine, then went and decided they needed to reduce the size of their jobs. And that's this well right here that they used a clone with. Focus in on this. They reduced the size of the job from 11 million gallons to eight; production went down, 5800 barrels instead of 8800.

Slide 61, reproduced below, shows the production reported by FracMax:

As before, we use the API#, 42-285-33673, to look up the Texas RRC ID #, which is 10811.

After retrieving the production data using ID # 10811, we find that Gillespie has the same undisclosed adjustment as Targac. Using the same table as we did for Targac, we compare the FracMax numbers from Flotek’s slide and the original RRC data.

Once again, the numbers don’t match. Chisholm correctly notes that both water use and production at Gillespie 1H were lower than at Targac 1H. 

But what he didn’t say was that his numbers did not match the official data from the Texas Railroad Commission: production was both a month earlier (May, not June) and substantially higher than he claimed.

Molnoskey 1H

The Molnoskey 1H well, also by Sabine Oil & Gas, is the punchline of Chisholm’s presentation of well comparisons using FracMax screenshots. This is the well that enjoyed the benefit (and pleasing citrus odour) of Flotek’s complex nano-fluids, and the other wells were selected for comparison because they are near Molnoskey (slide 62).

Slide 63 shows the additives and production data as reported by FracMax:

Chisholm noted Flotek's complex nano-fluid sold under the trade name “Super Stim 120,” and highlighted one ingredient, citrus terpenes. Chisholm then made the key, direct comparison between Molnoskey and Targac, the first well presented. At 1:08:04 in the replay, he stated:

They [Sabine] then came to us and said, we're going to reduce it even further, in fact they called these wells “smurf” wells. And we're gonna run complex nano-fluid.

Focus here on this. Volume six million gallons versus the original one [Targac] of over 11. Production 12.5 thousand gallons in the first month versus eight, so the production's up 50%, the fluid's down 50%, here's complex nano-fluid, in fact it's Super Stim 120, remember FDP? Super-stim in this case they happen to put citrus terpene in there.

This is the key to Flotek’s claims of the value of CnF and its FracMax app. Given our experience so far, we should probably double check the numbers. We use API #42-285-33756 to retrieve the Texas RRC ID# of 10920.

But this time the RRC production data for ID# 10920 holds a surprise: while the first month was again omitted, it hasn’t otherwise been adjusted at all. In other words, only for the well with CnF did Flotek report (mostly) the original, unadjusted RRC numbers:

John Chisholm fails to meet his own "accuracy" standard

Not only were Flotek and Chisholm’s repeated insistences that the data is “as reported by the operator” incorrect, but the adjustments we have seen ARE consistently favorable to the Company’s CnF well and detrimental to the non-CnF wells. 

Mistakes happen, though they undermine the many declarations of FracMax’s statistical robustness. 

Given this pattern of adjustments we can’t help but wonder if these divergences are, in fact, mere mistakes. It is just too suspicious to consistently understate the results from wells that did not use their product. 

Starker problems

Comparing the adjusted and unadjusted data for Molnoskey (CnF) and Targac (no CnF) reveals a starker problem: for at least some months, when the oil production is reported without Flotek’s alterations, the CnF-containing well performed worse, not better.

The table below compares monthly oil production in barrels, starting with the first month, of Targac and Molnoskey, the two wells Chisholm compared above. Under the heading for each well, there are three columns: Flotek’s numbers from its FracMax slide, the Texas RRC numbers, and Flotek’s numbers divided by the Texas #s shown as a percentage. (This data is the same as shown in each well’s respective comparison tables above.)

As Chisholm’s claim was that Molnoskey produced more oil than Targac with less water—thanks, he implies, to CnF—the final two columns check that claim using the original data. 

Under the heading “Diff in Production (Mol. – Targ.),” we subtract the two sets of reported numbers. First, we show Flotek’s numbers for Molnoskey – Targac, for the same month of production. These show an increase, as Chisholm stated. But when we use the original Texas RRC numbers, in the rightmost column below, we find that, for at least some months, Targac (no CnF) outperformed Molnoskey (CnF).

Sure many factors affect production (eg geology, how much water and sand and maybe acid are pumped into wells and maybe even the surfactants use). 

Our aim is not to prove that CnF makes no difference. To do that we would need a large sample set. Rather we point out that on John Chisholm's data (after taking out the errors) it is impossible to conclude that Flotek's complex nano-fluids do anything.

We believe in "accuracy" but there ain't none here.

Like the wayward analyst (with apologies to James 1:3), the testing of our faith in accuracy produces perseverance.  

So let’s look at the final well.

Berger 1H

The final well Chisholm highlighted was the Berger 1H Unit from Devon Energy, a nearby well (slide 64) that he told us had no CnF. A screenshot from FracMax reports Berger’s additives and production (slide 65):

Chisholm narrated at 1:08:43 in the replay:

Wanted to show you the well right next door to that CnF well was with...by Devon.  Uh...amount of fluid 5.9 million, less than half the production. Right next door. To the well that had CnF in it, with a similar amount of fluid in the frack job. This type of data analysis was absolutely [sic] no way Flotek had access to it 'till we invented FracMax. Absolutely no one else in the industry has it.

Using the API#42-285-33802, we retrieve its RRC Lease ID#10972:

As above, we look up the production data for Berger from the RRC using ID# 10972. Once again, Flotek’s reported production data from FracMax does not match the RRC’s, as the by-now-familiar table below shows:

Where before Flotek omitted the first month of production (usually the largest), in this case they have omitted the first two months of September and October 2014, as well as the most recently reported month, June 2015. And we find again the same undisclosed alteration of the RRC’s data, as FracMax displays only 60% of the RRC’s number for oil and 50% for gas.

Chisholm compared Berger and Molnoskey directly, as the two wells are nearby and used a similar amount of water, but only Molnoskey contained CnF. Berger, Chisholm said, had “less than half the production,” presumably intending for his investor audience to conclude that CnF made the difference. While Molnoskey clearly produced more oil in its first 10 months, the difference is far less stark once the full RRC production numbers—including the missing months—are taken into account.

The size of the claimed Flotek data base

Ultimately, of the 9,800 wells Chisholm claims have used CnF (replay at 45:48), Flotek picked these to compare. They were chosen deliberately. If this is management’s “best foot forward” we wonder how FracMax will persuade sophisticated reservoir engineers who must build models incorporating production cost, geological complexity, etc., while maintaining statistical robustness.

It’s worth quoting Chisholm’s statement at 45:48 in the replay in full: 

“[W]hat you'll see today is on a FracMax instance that has 98 thousand wells. Not 98. Not 980. Not 9000. Ninety-eight thousand wells in the United States that we have all categorized based on what went in and what the production is. And 10% of it has CnF in it. And any statistician will tell you that's enough of a database that you can make algorithmic, statistic inferences as to what the economic impact is on a dataset that big.” [emphasis in original]
So we thought we might test the whole data base. And it was all meant to be in their FracMax app all ready for your ipad.

So we downloaded an app claiming to be FracMax.

We just can’t make it work and we tried with a brand new ipad just to test compatability.

Trying the FracMax App (?)

A version of what claims to be FracMax is available in the Apple iTunes Store under the name “Flotek Max.” Released on 17 February 2015, the description of the app fits the public descriptions of FracMax:

 A quick but important caveat: we cannot know for certain if this is the app Chisholm demonstrated or the app used by Flotek’s sales force. It is entirely possible—likely, even—that another version of FracMax exists. That said, our reading Apple's policies make it unlikely that a parallel private app exists. 

With that caveat out of the way, we do what we can to try to use the app. 

We find the same Flotek application pictured in the consumer app store above:

Clicking on the link leads us to the same page as before, except with “Volume Purchase Program” as the header:

(Note the section on “Managed Distribution” on the webpage above.  It indicates that free apps are only available in bulk when using a “Mobile Data Management” solution, which implies control over the end user’s iPad.  This form of distribution would be more common within a single enterprise whose mobile devices are managed centrally.)

Anyway, we downloaded the iTunes Store application—the first one shown, from the consumer app store—and tried to run it on two separate iPads, each running the latest version of iOS (one an iPad 4th Generation the other an iPad Mini 4).

Here’s what happened when we tried it, from the initial download onwards - note that this is a very boring video showing you what happened right to the point where the ipad crashed:

On neither iPad we tried did the application work. First, it asked for access to the camera, for what purpose we do not know. Next, it simply displayed a black screen with a set of red bubbles, much like the screenshot of the business card in the app store images above—except without any discernable image in the background. No amount of tapping or dragging or swiping or pinching could persuade it to do anything.

We would love if other readers could try to load the app. At last testing it crashed both our ipads. 

This is curious as Flotek tells us that this app is responsible for the fantastic recent sales performance. (Also curious is the claim in the app store description that the data is updated weekly; as the images show, the latest version of the app was released on 17 February 2015.)

As we can’t run FracMax—or an app from Flotek claiming to be FracMax, anyway—accuracy-minded investors like us can’t check other wells. It would be fascinating if Mr Chisholm's faulty data is repeated over the whole United States. That would be deeply "curious".

And we like a good stock puzzle - especially one where the management seem to make it up.


And a side note to Mr Chisholm. 

Adlai Stevenson suggests that “Accuracy is to a newspaper what virtue is to a lady, but a newspaper can always print a retraction.”

Can we expect a retraction or is Flotek truly without virtue?


Post script:

The company put out a press release that admitted to most the key issues in this blog post. The press release is repeated here.

Flotek has carefully reviewed a recent report regarding the validity of the data from the Company's FracMax® database by analyzing data from a small subset of wells in FracMax®.  The analysis suggests the production data presented by Flotek – for three of the wells analyzed – were misinterpreted by Flotek and understated the production of those wells. 
"We take any contention of errors in our data, processes and analysis very seriously," said John Chisholm, Chairman, President and Chief Executive Officer of Flotek. "We appreciate the thorough analysis provided in the report and are using this critique as well as others to improve our FracMax® application to ensure both the validity and reliability of the underlying data as well as the accuracy of the analytical processes." 
As a result of our initial review of the report we have concluded that the FracMax® database – partly a result of third party data used by Flotek – identified the three wells in question to be contained in units with multiple wells (the state of Texas organizes production reporting by units, or leases, that report total production in the aggregate).  The FracMax® application uses algorithms to assign production to individual wells within multiple well units. In this case, the report contends that the wells in question were on single-well units and, as a result, 100% of the production from those units should be assigned to the identified wells. After review of the report, data from the Texas Railroad Commission as well as other third-party data providers, it appears that the wells in question are single-well units. 
While the adjustment does impact the magnitude of the outperformance of the well completed using CnF® when compared to the non-CnF® treated wells, it does not, the Company believes, change the conclusion that the CnF® well outperformed the non-CnF® wells, especially when normalized for the length of the lateral completion zone.
"While we are concerned by the unintentional data and processing error that led to this unitization miscalculation and are taking aggressive steps to ensure this process is immediately corrected, our analysis of the wells in question concludes the use of CnF® improved productivity when compared to the neighboring wells that did not use CnF® in the completion process," said Chisholm. "Moreover, Sabine Oil & Gas – the operator of three of the wells in question – continues to use CnF® on its completions, an indication that Sabine's internal data show compelling benefits from the use of Flotek's Complex nano-Fluids® suite of completion chemistries." 
The Company has conducted an initial review of the FracMax® database and has determined that this data and process error does not impact the vast majority of wells in the database which appear to be unitized appropriately by the software application. "In fact, there were several other wells in the investor presentation that were unitized and reported correctly, showing the benefits of CnF® that were not discussed or referred to in the report," added Chisholm. 
"While FracMax® is an important tool in the development of new markets for CnF®, the most important determinant of the success of CnF® is the actual performance of the Company's completion chemistry in the wells of new and existing clients," said Chisholm. "The growth in validations that result in new commercial customers is the ultimate proof of the efficacy of our completion chemistries. Our clients and their experiences – as seen through their own proprietary production data – are by far the best evidence of the performance-enhancing nature of CnF®." 
The number of production companies using CnF® in the completion process continues to grow with a consistent flow of validation projects for operators of all sizes.

A previous press release took issue with my assertion that the FracMax iPad app did not work. To quote:
"In addition, as we have noted consistently over the past year, FracMax® is currently used exclusively by the Flotek team to analyze well and performance data for our clients," added Chisholm. "The application is not available outside of Flotek and, as a result, cannot be downloaded without appropriate credentials."
I have yet to see any working FracMax app. Moreover I have not seen anyone (such as a client) using said app. The company now says that the app is entirely an internal tool. However this is not how it was represented in the past. Here is a link to the app's website on Archive.org.

And here is a picture.

This is the "innovative mobile app where the industry revolutionising impact of Flotek's CnF technology is on full display" and we should "prepare to experience what it can do for yourself".

This is very different from the current assertion that FrackMax is an internal sales tool only.

Here is a simple challenge to Flotek management: find a friendly Wall Street Journal reporter who will report - even in a brief column - that they have experienced it for themselves.

I do not think it can be done. Without further proof I remain convinced that the management are simply making it all up.


General disclaimer

The content contained in this blog represents the opinions of Mr. Hempton. You should assume Mr. Hempton and his affiliates have positions in the securities discussed in this blog, and such beneficial ownership can create a conflict of interest regarding the objectivity of this blog. Statements in the blog are not guarantees of future performance and are subject to certain risks, uncertainties and other factors. Certain information in this blog concerning economic trends and performance is based on or derived from information provided by third-party sources. Mr. Hempton does not guarantee the accuracy of such information and has not independently verified the accuracy or completeness of such information or the assumptions on which such information is based. Such information may change after it is posted and Mr. Hempton is not obligated to, and may not, update it. The commentary in this blog in no way constitutes a solicitation of business, an offer of a security or a solicitation to purchase a security, or investment advice. In fact, it should not be relied upon in making investment decisions, ever. It is intended solely for the entertainment of the reader, and the author. In particular this blog is not directed for investment purposes at US Persons.