Don’t Do Accounting in dbt Cloud
Andreessen Horowitz and their VC cartel are setting up Pickett’s Charge of cloud data workloads. It’s up to you to hold the high ground.
In June, 1863, Confederates under the command of General Robert E. Lee invaded Pennsylvania.
Two years of the Confederate Army of Northern Virginia (AWS us-east-1) exchanging territory back and forth across Virginia with the Union Army of the Potomac had left both sides weary. Confederate supply lines had been bombed out. Confederate heavy industry was not set up for a war this long. Any meaningful intervention from the French or other European powers in support of the Confederacy was unlikely after Lee’s previous invasion of Northern territory was stopped mere miles into Maryland at Antietam the previous September. Lee was facing pressure from Confederate President Jefferson Davis to reallocate resources into the Western Theater, where two upstart Union generals, Ulysses S. Grant and William Tecumseh Sherman, were threatening to take the Mississippi River.
Lee made a gambit.
Coming off a significant victory at Chancellorsville, though it saw him lose his right-hand man General Thomas “Stonewall” Jackson, Lee sought to destroy Northern popular support for the war and to gain supplies and threaten Philadelphia, Baltimore, and Washington, DC.
Under the smokescreen of General J.E.B. Stuart’s cavalry acting as a buffer, Lee moved his army through the Blue Ridge Mountains and Shenandoah Valley into Union territory.
Within the month over 70,000 armed rebels were in Pennsylvania.
The Invasion of Finance by dbt Labs
Like Robert E. Lee, dbt Labs CEO Tristan Handy and the rest of the “Modern Data Stack” CEOs need to take land, and fast.
They and their VCs need data engineers, data scientists, analytics engineers, analysts, ML engineers, and whatever other titles are floating out there to move workloads into the Modern Data Stack. Every single one of these Modern Data Stack CEOs is on the list, all the way up to Snowflake CEO Frank Slootman.
In fact, Slootman’s got the most pressure on him. He’s going to get taken out back behind the shed if he can’t improve Snowflake’s NRR, which is not only declining but declining at an increasing rate QoQ.
What are some of the best workloads for land grabs? Marketing is pretty good. Product is pretty good. Both tend to have large data sets of behavioral actions, IoT, etc. that make for good consumption and good consumption attribution.
But the best workloads are in finance.
If Tristan Handy can convince enough people to use dbt products and a cloud data warehouse like Snowflake as an accounting solution, then they’ve got even stickier use cases even though dbt Cloud is in no way, shape, or form an appropriate solution for accounting software or an ERP system. He can sell more dbt Cloud and boy does he need it.
At dbt Labs’ Coalesce conference this year, there are multiple featured speaking sessions about using dbt Labs’ products for accounting.
It’s obvious that Snowflake and dbt Labs themselves are going to have people go on stage and make stuff up about whatever they believe to be strategic finance and accounting. Snowflake and dbt Labs need the workloads.
However, perhaps the most dangerous presentation is that Rocket Companies (NYSE:RKT) is presenting on how they rolled their own Quote-to-Cash system in dbt Cloud.
dbt Labs is now positioning themselves as an accounting and ERP solution, and they are using one of the worst possible examples, Rocket Companies, to demonstrate how a publicly traded company may use dbt Cloud for an accounting process that is ostensibly SOX compliant.
Let’s get into it.
Rocket Companies is a rollup of several companies, most notably Rocket Mortgage (fka Quicken Loans) which carries most of the business. The business was founded by Dan Gilbert in the 1980’s, during the dotcom boom of the 1990’s he put mortgage loan origination on the internet and then IPO’d in 1998, sold his business to Intuit for $532 million in 1999, then bought it back 3 years later for $64 million.
If you’re a basketball fan, you may know Dan Gilbert as the owner of the NBA’s Cleveland Cavaliers and as the guy who threw a tantrum and egged on riots and burnings of LeBron James jerseys in the streets of Cleveland when James left the Cavaliers for the first time in 2010.
If you’ve ever been to Michigan and especially Detroit the past few years, it’s hard to avoid Dan Gilbert’s presence. He is the wealthiest person in Michigan. He’s got Detroit in his pocket and Rocket Companies companies together are the city’s largest employer. Since 2009 he’s built up some urban renewal projects in Detroit, which has led to equity value in homes going up, which leads to more turn, and who do you think these people go to when they want a mortgage loan or wish to refinance? It’s very geographically circular, a system that feeds itself not totally unlike what happened with Silicon Valley Bank earlier this year.
The Gilbert Pavilion at Michigan State University houses a fan experience center and stands next to the basketball stadium. In 2021 it was announced that Michigan State University Spartans Men’s Basketball became known, if only in written title and mostly ceremonial, as The MSU Spartans, Presented by Rocket Mortgage.
If basketball isn’t your thing and you prefer American football, each Saturday throughout the fall you can watch the head coach of Michigan State Mel Tucker yell through his Rocket Mortgage headset.
If golf is your thing, you can swing by the Rocket Mortgage Classic on the PGA Tour, which began play in 2019. If you win the Rocket Mortgage Classic, you can hoist a trophy which is quite literally a giant Rocket Mortgage logo, only the black bottom part of the logo is a golf tee.
Michigan’s wealthiest man who owns Detroit’s largest employer spreading money around buying sponsorships for everything in Michigan may sound all nice and dandy to you.
But here’s where it gets stupid.
Rocket Companies IPO’d in August 2020, after about 18 years since Dan Gilbert and co. repurchased the company from Intuit. With the Fed dropping interest rates to the floor in 2020, there was no time better to float public stock for Rocket Companies.
It very quickly became a meme stock.
Hedge funds jumped on and began shorting the stock as soon as they could. They know Dan Gilbert. They know the whole point of Rocket Companies’ IPO after about 18 years as a private business was done for one reason — to dump shares after the lockup period ended after 180 days.
A few months in after the IPO, Jim Cramer gets involved. He does a segment on CNBC egging on Reddit and the meme stock crowd to consider buying the stock to bust the short sellers. The price jumps after Cramer’s segment.
As expected, Dan Gilbert dumps shares at the end of March, 2021, after the lock-up period had ended but before the quarterly earnings call.
This “philanthropic gesture” is, of course, one big load of baloney.
Any money that actually gets out to cover property tax delinquencies in Detroit goes right back into the game. If delinquincies go down, vacancies go down. Fewer defaults occur. The city government likes that Dan Gilbert covers their costs of evicting people and him covering taxes essentially goes right back to the government. All of this increases home values in Detroit, and it means more people get more mortgages and/or refinance when home values go up. Who owns the game in Detroit? Where do these people turn to for refinancing or mortgages? They go to Dan Gilbert. It goes right back into his game as volume on loan origination.
Meanwhile, who holds the bag? Well, anyone who bought and held stock between the IPO and Dan Gilbert paying himself $500M holds the bag.
Who had a bag? Among others, The Detroit Police and Fire Retirement Fund — the fund representing the pensioners in Detroit who worked in public services.
They sued on the grounds of insider trading, then months later the suit vanished.
Remember: Dan Gilbert owns Detroit. He’s the largest private employer. He does “urban renewal” to pump house values. He runs a “philanthropy” to help people behind on their property taxes, which is really just a way to pass money to the city of Detroit without being an outright bribe.
It’s hysterical.
Even as recently as April of 2023 there is another class action lawsuit proceeding related to the same Dan Gilbert sale of stock via Rock Holdings.
There is an entire Reddit forum dedicated to taking long positions on Rocket Companies’ stock.
Even better, in May this year, Rocket Mortgages announced a new product called ONE+ which offers low-income individuals the chance to put 1% down on a home, Rocket Mortgages covers another 2%, and borrowers become locked into high monthly payments. If home prices fall like they are in much of the country then the borrower is likely to go underwater and defaults will start happening.
It’s a predatory product.
But Rocket Mortgages has to do it. They need volume because refinancings are down and very few people are refinancing with rates up. They either already refinanced in 2020–2021 or they are holding off.
It’s alllllll bad.
Of course dbt Cloud is being used for accounting at a company like this, where the IPO was unnecessary and is just a scam.
Of course a presentation about this nonsense gets selected and put on stage at the venture capital party known as the dbt Coalesce conference.
dbt Labs CEO Tristan Handy needs more use cases for dbt Cloud, more examples of workloads, more examples of any use cases out there in the market that can be trotted out on stage.
He needs the workloads and he needs to convince enough suckers and inexperienced management that using a SQL manager hosted on the cloud for running a quote-to-cash and revenue recognition process is somehow more appropriate than using actual accounting and ERP software, of which there are hundreds of options available on the market.
Not only this, but passing a SOX audit with a custom-rolled SQL management system instead of accounting software tells me that whomever was running this SOX audit has no idea what they are doing.
A SOX audit is not armed officers of the FBI and SEC coming into a business and watching the accountants put together invoices. All a SOX audit of IT systems means is that the business hired an outside consulting firm to come in and check boxes.
It doesn’t mean it’s good. It doesn’t mean it’s bad. It just means that a consultant was paid money to give a thumbs up because the law says that you need to hire an outside party to come in and check IT systems that feed into finance.
Composable CDPs: Holding Little Round Top from Johnny Reb
Day 2 at Gettysburg saw some of the most intense fighting as Confederate troops tried to bust the Union left flank in the heavily wooded area around Little Round Top.
Little Round Top was a strategic location on the southern end of the Union line. If the Confederates were able to take it, they would have been able to flank the Union army and potentially win the battle.
The 20th Maine was positioned along the crest of the hill, with their left flank anchored on a rocky outcropping known as Devil’s Den. Their line was thin, with under 400 men present, but they were bolstered by the fact that they were fighting on high ground.
The Confederates sent wave after wave of troops up the hill, taking heavy casualties along the way but wearing down the Union lines.
The 20th Maine was led by Colonel Joshua Chamberlain, a professor at Bowdoin College and a former Sunday school teacher who spoke several languages. Unlike many officers on both sides, he was not a career military man, only joining the Union Army in 1862 once the war was underway. He was offered the position of Colonel upon joining but declined as he didn’t feel he was ready yet, so he took the next position down of Lieutenant Colonel.
The Union Army’s left flank was held by an honest nerd.
Wave after wave of Confederates charged the Union lines throughout the afternoon of July 2. Running low on ammunition from fighting off wave after wave of Confederate advances, Chamberlain ordered a bayonet charge. The 20th Maine held the line.
The last vestiges of an assault from Composable CDPs (which is a way to sell you 4–8 tools instead of 1 CDP, plus you have to work out all the id resolution in dbt yourself) are coming from Andreessen Horowitz.
We’ve got to keep holding Little Round Top and we’ve got to fix bayonets whenever we can.
For over a year Johnny Reb, led by Hightouch Reverse ETL CEO Tejas Manohar, has hit Little Round Top with wave after wave after wave of Confederate content about the Composable CDP.
There is probably nobody out there more skilled at burning VC money to a crisp than Hightouch CEO Tejas Manohar. All Hightouch has done is take tens of millions of dollars and confuse the market.
Further, most of the arguments Andreessen Horowitz, Hightouch, and others have made are all rug pulls to get around the fact that dbt is not a great solution to put in the middle of any of this.
Here are all the reasons why Andreessen Horowitz’s charge up Little Round Top fails and why Andreessen Horowitz, Amplify Partners, Census, Hightouch, and everyone else is about to get smoked by the 20th Maine of Fuggin’ Reality.
First, zero copy is not a real argument. Data will be copied to end applications like ad networks, email tools, etc. This is a nonsense buzzword from Andreessen Horowitz and Amplify Partners and co.
Now, minimizing data copying and reprocessing could be a good argument, as unnecessary copying and reprocessing of data can be very expensive.
However, one of the worst offenders of unnecessary copying and reprocessing of data is dbt Labs, which has promoted for several years the idea of doing full refreshes. The idea of “incremental” loads and incremental processing is deemed an advanced task. For example, you can read this piece from dbt Labs called “How we shaved 90 minutes off our longest running model” which makes no sense. They are burning money for no reason.
Second, lock-in to a data warehouse/lake is secondary to lock-in caused by dbt. There is very little reason why id resolution and other CDP-related business logic need to be expressed in SQL.
Third, the no-code interface and act of pushing down SQL to the data warehouse is simply another way of selling more dbt and compute.
Fourth, the idea of “define once, use anywhere” is not unique to schemas and metrics defined in the cloud data warehouse. In fact, many mature businesses define this in middleware like an ETL tool or in code before the data warehouse and then propagate out to multiple target systems.
Most of these Composable CDP use cases around the market seem to be at companies with 1–5M customers. You’re not really seeing large companies actually executing on a Composable CDP, and if they are they are probably using something like MessageGears for “Reverse ETL” and message delivery.
There is unlikely to be much lift derived from a Composable CDP. It’s just a way to move compute to Snowflake or Databricks (an a16z investment) via several other a16z investments.
The Lost Cause of Andreessen Horowitz and How to Avoid ‘Going Down to Dixie’
The Lost Cause of the Confederacy is the myth that developed in the years after the Civil War that secession and the war against the United States was a just cause and actually about states’ rights and not about an aristocratic class of Southerners owning human slaves.
Sure, you can argue that it was about states’ rights. It was about the right of a state to decide whether it would allow slavery.
We call this game an “abstraction layer” in tech. You layer on a GUI or some features not offered in the underlying service or product. You layer on a narrative and lots of content marketing. Voila. Now the Civil War is not about slavery, it’s about states’ rights.
The fact of the matter is that 99% of the people in the Confederate Army did not own slaves. The rank and file were mostly a bunch of poor farm boys pulled into a political power struggle by landed aristocrats trying to capture higher margins by enslaving human beings as their means of production.
If you pay people nothing and you own them as property, you get better margins.
If you can convince enough people who have never been more than 20 miles from their homes to join your cause, you can raise an army to help you protect your margins while taking on less personal risk.
It’s alllllll bad.
Don’t Pickett’s Charge the US Government and Don’t Be Stupid
If you’re a Confederate soldier, you have to ask yourself who gets the good end of the deal.
You went out into the battlefield. You risked your life, your limbs, and you risked widowing your wife and leaving your family without. If you win, who benefits?
The colonel of your division benefits. The generals benefit. The rest of the landed elites back in Georgia or Alabama or Virginia benefit. They all grew up with slaves, they own slaves, and they’re using you to help keep their margins fat. That’s all it is to them — human life as their margins. They want free labor.
Sure, you may get a medal. You may have a song written about you. But at what cost? You think your colonel is going to hire you and give you a cushy job when you get back home after you beat the Union? No. Why would he give you a job? He has people working for him for free. All of his buddies have people working for them for free.
You’re in a cult. You’re in a “community” of likeminded people. You are the cannon fodder.
You take all the risk with very little reward. These people telling you where to stand, whether to shoot or fix your bayonets, which Yankees to charge at, they all take lower risk and get more of the reward.
Pickett’s Charge was a major event that took place on the third and final day of the Battle of Gettysburg. After getting turned back at Little Round Top the day prior and unable to flank the Union, the Confederates sought to take the middle of the Union Army lines through brute force.
Pickett’s Charge is often cited as one of the best examples of the futility of war due to the sheer loss of human life sacrificed to obtain land.
The charge involved over 12,000 Confederate soldiers ordered to march across an open field for three-quarters of a mile under heavy Union fire. The charge was preceded by a massive artillery bombardment that was meant to soften up the Union defenses, but it failed to do so. As a result, the Confederates were met with a hail of lead that cut their numbers down significantly.
Today, if you go to Gettysburg Battlefield you’ll see several markings on the appropriately named Cemetery Ridge.
The High Water Mark at Gettysburg, quite literally the furthest point the Confederates got trying to take Cemetery Ridge, portrays a soldier of the 72nd Pennsylvania Division with his rifle ready to club any Confederates coming over the wall.
YOU ARE NOT GETTING PAST THIS GUY. EVEN IF HE RUNS OUT OF BULLETS HE’S JUST GOING TO USE HIS GUN AS A CLUB AND BEAT YOU UNTIL ONE OF YOU IS DEAD.
Of the 12,000-plus men who charged the open battlefield against the United States government, over 8,800 were killed, wounded, or captured.
The Lost Cause of the Confederacy and “analytics engineering” your way into 5,000+ dbt models built on cloud compute and using dbt for accounting are examples of being stupid.
It doesn’t matter whether you’re a CFO or a Director of Data Engineering or a first-year analyst, or anyone in between, you have to fight against being stupid.
Joining the Confederate cause as a non-slaveowning poor, rural farm boy is stupid. You get almost no upside and you take all the risk trying to shoot people who are shooting back at you because you declared war on them.
Let the officers charge the Union on their own.
Let Tristan Handy and the rest of his cult charge forward in the open battlefield promoting dbt Cloud as an accounting and ERP tool.
You shouldn’t be doing accounting and financial process management in dbt Cloud.
They need the workloads and they are out of ideas.
Just like the Confederates, Tristan Handy and Andreessen Horowitz are going up against the United States government.
Don’t get caught up in his nonsense of doing accounting in dbt Cloud and dbt Labs trotting out someone from Rocket Companies as an example of a healthy business doing proper things. There is probably not a worse example of a publicly traded company that could have been used, which is both hysterical and sad.
Throw down your weapons. Stop listening to Tristan Handy and the rest of the Andreessen Horowitz and Amplify Partners influencers, and go home. Stop retweeting and parroting their nonsense.
Let finance run their own processes. Let marketing run their own processes.
Then, if you want to bring the data into dbt and a cloud warehouse or analytics database after it’s already been verified, do that, but don’t reconcile finances or try and roll your own CDP. It’s nothing but a waste of time at best and at worst it can have serious legal and financial consequences for both you and your organization.
Let the cult keep singing “Dixie” with their content marketing and paid influencers.
On the side of the righteous we’ll be singing “The Battle Hymn of the Republic” and we’ll be getting real work done on lean teams.
