Chief Executive Officer,
Tri-Point Oil & Gas Production Systems
Thank you for joining today’s session on the cash forecasting case study of TRIPOINT and expected benefits with AI. We have with us Jeff Martini, CEO at Oil and Gas Production. Jeff loves helping organizations succeed and finds it essential to focus more on the stories and opportunities behind the numbers than just the financials. And also we have Bernie’s vendor Weldon, a functional consultant at HighRadius, Jeff and Bernice, the stage is all yours.
[0:30] Bernice van der Velden:
Hi, everyone. You’ve already seen my face a lot by this point. My name’s Bernice, and I’m talking today with Jeff. He’s the CEO of TRIPOINT. We’re in the implementation phase with him a little on the first end of it I’d say but he has a great story to tell and I think we are excited that our expectations will be fully met and he has some great expectations as well. So we’ll be covering that in today’s session. So just to give you a glance at the agenda, we will be going through what TRIPOINT is and what they do. And then we’ll be talking about their current cash forecasting practices right now. And then Jeff will be talking about the challenges that they face right now, and why he came to us. And that’s where I’ll step in and show you a live demo of the system. And then we’ll bring it back to Jeff. And he can explain what his expectations are. So I’ll give it to you.
[1:35] Jeff Martini:
All right. Thank you, Bernice. Appreciate that. Thank you very much. Right, so a little bit about me, what you’ll hear from me is sort of the comments and in the construct that the way we’ve come to HighRadius and the way we do our cash forecasting, or liquidity forecasting a little more wholesomely, is colored by my experience in oilfield service. So for the last 20 years or so, I’ve been in a variety of different oilfield service companies, larger publicly held companies and in the smaller middle market, private equity-backed companies as well. But in the last 15 of those years, I’ve been in a variety of standalone CFO roles and took the CEO role at TRIPOINT just within the last several weeks here. So TRIPOINT as a company where we’re a manufacturer, so we’re probably again categorized within oilfield service. But what we’re doing is manufacturing production equipment is what it’s called. So, to oversimplify a little bit, you see one of our welders on the screen there. We turn flat steel into circles.
[2:54] Jeff Martini:
We put caps on the end and calm vessels or tanks depending on what size they are. All the equipment is to help producers, EMP companies, exploitation production companies, such as Chevron or Oxy. When they produce a well, three phases make up the word TRIPOINT. There are three phases of substance that come up out of the ground, there’s oil, there’s water, and there’s gas. So in every single well, you’re going to get all three. So what our equipment does is it separates those three phases. So the oil and the water generally are our house on the production pan. So they’re big tanks. If you ever drive West Texas or East Texas or South Texas or anywhere, you’ll find these production pads with big tanks on them. So those are there to hold the oil in the water and then separately, the gas gets compressed generally sent down a pipeline and used in the industrial or residential applications.
[3:58] Jeff Martini:
As far as our footprint is concerned you’ll find us in the usual spots across the world patch of the US. So we got locations and Midland, up near Oklahoma City, East Texas. And we’ve got a footprint up in the northeast from Pennsylvania, Ohio, as well as up in the Rockies north of Denver up that way. So as far as employee count and sort of the rhythm of the business, we got about 500 employees. So our average ticket size is somewhere call it half a million to a million dollars of invoicing. But there is another business in there that is more of what we call the supply store business. It’s much more transactional, sort of a parts counter kind of a business that sits on top of that as well.
[4:47] Jeff Martini:
The cash forecasting is sort of that we’re transitioning away from using the HighRadius software, which was/is a very manual process. So it involved being set up in regions again, sort of the obvious regions, the Rockies of the Permian, the Northeast, etc. Each one of those has a finance manager associated with it. And each one of them independently would run every week run the aging, the A/R aging, using their understanding of what’s going on with the business with their customer relationships are like and what our history is like, would be inspired by but not bound by the due dates. Again, fees are sort of notorious for the slow pace. So using their best human heuristics, they would guess at what the expected payday was. So we would gather all that up every week. Everybody’s got supposedly standard spreadsheets, but inevitably they’re not. So we would gather those up and corporate, add it all together and come up with the cash forecast. That cash forecast when would roll up into some more modeling of a 13-week cash forecast that also has some debt associated with it as well. We’re also bound by an ADL and asset-based lending facility. So all that rolls in, together and makes a sort of a stew of goodness, let’s say that at the moment.
[6:26] Jeff Martini:
It’s just a sort of baseline, where we’re moving away from? See some statistics here you see a pretty poor forecasting number. In our defense, it’s not quite that bad. That’s 50% every week. So every month, you zoom out a little bit, things look a little bit better. But it’s still not great, not world-class by any stretch. The forecasting was either done as a setter sort of every week or on a sort of more intense kinds of liquidity narrow channels as I call them to navigate. We’re refreshing that forecast much more regularly in the more abbreviated way. And the forecast goes out for about 16 weeks. So after the next few weeks go by it comes down to a 13 week and we had some more columns in the Excel spreadsheet. So there’s isn’t a whole lot of sophistication to it. That’s why it’s 13 to 16 weeks just because it’s kind of a pain to add additional columns and oodles and gobs of sheets.
[7:35] Jeff Martini:
So what that has a knock-on effect of because it’s unpredictable and unpredictable as far as our AR is concerned or poor forecasting is really to be blunt about it and some of the things that sort of make it even harder, said this is a capital equipment business. So the timing of those receivables trying to spot what week those are going to happen, 13 weeks out is pretty tough. So there’s a lot of luck involved in trying to get that done. Again, given the way that we’re doing is bottoms up human heuristic, intense kind of a structure inside the week, so again, I talked about the burden with OFS. Right, it’s all about that it’s not a great space to be in, characterized by thin liquidity these days.
[8:23] Jeff Martini:
So, predicting what the receipts are going to be daily, also pretty important. So we have an intro week shaping of what we expect the receivables to look like. That’s based on history or inspired by prior history. There is some seasonality here as well. So inside each quarter with EMP’s tend to do is they dress up their balance sheets as at the end of the quarter. So that leaves the slow pace and that net third month, especially in the last two weeks of the third month, and then it catches back up in the first few weeks of the following quarter. So all these different rhythms going on. All these things to try and keep in mind as we maintain this robust 13-week cash forecast all that leads to a pretty inaccurate forecasting construct. So, what’s behind this right so let’s play the five why questions and figure out what is the root cause of the problems of the inaccuracy. It requires a lot of very knowledgeable people, it demands your perfect knowledge for our finance organization.
[9:38] Jeff Martini:
As far as customer behavior is concerned, as far as the invoicing patterns are concerned, the collection’s patterns are concerned. So there just aren’t enough smart people affordably to fill up the business that way, lack of access to the right data, right. So the data is there. We’re proving that now. That the data is available, it’s just inscrutable. It’s just, in the ERP system, it’s in a dozen different tables. And it’s impossible for any person or even a group of people to push it all together on a real-time basis and understand the dynamics. And then as I mentioned, the limitations of spreadsheets, they’re great for a lot of things but not everything. And they’re just not up to the task of a real-time moving forecast.
[10:32] Jeff Martini:
The knock-on effect to the business right, so we’ve sort of marched down the consequences here, that there needs to be additional liquidity on hand to be able to give enough cushion to account for this uncertainty. So because liquidity and cash are unpredictable or poorly predicted numbers, we need to make sure we have more of it. So that means additional depth and it’s more expensive than means more horsepower, more people to try and keep up with it all. So ultimately as a fleet some more costs and more burden to the business.
[11:13] Jeff Martini:
So we’ve made multiple runs at trying to fix this and the first one was by swarming it with human resources. So we gathered up the organization and said this is job number one. We got to figure out our cash, they’re all hands on deck and let’s figure it out. And all the dozen ERP tables or whatever the case may be. This is what you need to do. You need to go figure it out. So that was good that helped. That helped as long as there wasn’t anything else going on. So as soon as the organization’s focus shifted somewhere else. As soon as we needed to close the books or since we needed to go through a year-end or Quarter end or there was some project or budget focus shifted away, and our forecast accuracy went back down. So it was an unsustainable way to try and manage this. Then we said, all right, well, it must be something wrong with our Excel abilities. So let’s get some consultants in here. Maybe we can come up with a better way to cobble these spreadsheets together, a better way to get these bottoms-up data in one tool. And then maybe we’ve got to get a better way to accurately forecast. That didn’t work either for a couple of reasons. One is the consultants they did get exactly what they were asked to do.
[12:49] Jeff Martini:
They came up with a more robust Excel spreadsheet with some sophisticated assumptions. It was model-based and very complex. And as long as they were there, it was okay. But here again, just like the first run of this, as soon as the consultants left the building, we were left with a tool that we couldn’t keep up with. The tool was too complex, too demanding. And, just a few weeks after the consultants left, we found ourselves in a liquidity crisis, because there was an assumption very deep in the modeling that said that, assumed DSo was going to be constant. So it wasn’t in our world, it was coming down we’re doing a great job of collecting but the DSO was assumed to be flat. So it showed a projection of a nice fat borrowing base that wasn’t there. So that’s a nearly one mistake job. And that was almost the one mistake that caused some substantial stress to the business. So we just scrapped everything the consultants did and reinvented ourselves and came up with something simpler and more manual. So the tool that we’re using while we’re going through the implementation of the HighRadius product is one that’s got sort of some built-in checks and balances. So effectively, we’ve got two people, I’m one of them that is doing dueling 13-week forecasts. So we’re using similar inputs to both models. One is a daily rhythm for 13 weeks. Another is a weekly rhythm for 13 weeks. Each of them has slightly different inputs.
[14:37] Jeff Martini:
But they’re coming from the same source data. So to avoid the crisis that we had, where there was an unknown input deep in a deep in a schedule somewhere. I have a colleague of mine that keeps up with her version of the 13 weeks I keep up with mine, and then we manually sync up once a week. So we’re sure that we’re not going to make any mistakes that way or minimize the risk of making any of these silly spreadsheet mistakes. But again, that’s pretty costly. What we have done in the meantime, is we have upgraded our collection software, we’re logging our promises to pay. So we’ve got a more efficient sort of source data. We’re not regionalized so much anymore. We’re relying on the regional managers to come over what they think the promises to pay. We’ve centralized that within our collections group. So we got one source of data from the collection software that we’re gathering up the promises to pay or the due dates, and then that informs the receivables piece. So still suboptimal, right. Our accuracy is a little better, my confidence is better that it’s not mechanically wrong. My confidence is not all that much better but it’s more accurate. Just because the source data is heavily dependent on these human heuristics I keep talking about. So setting up the HighRadius commercial, here we go. So this is what we’re going through right now. Right? So my process and if you hear a little bit of pain in my voice, it’s because it’s real.
[16:16] Jeff Martini:
It’s just sort of sitting and just thinking through what makes sense. If we were just to throw all these spreadsheets out and start over from scratch, what would a functioning receivables or cash forecasting solution look like? And it sounds like I’m on payroll, but I promise I’m not. I got to thinking about how computers can do this better than I can. So I’ve got this rich data set and all this history, their patterns in this data, I just can’t tease them out by myself. So I wonder if there’s somebody out there that’s ever done that’s had the same thought. And I’m sure there is. So Google was HighRadius’s friend in this case. So within just a few moments of looking for keywords, that sort of this described what I just dreamed up, found my way to HighRadius. And we got to talking and the software makes a lot of sense. I mean, it just does. Right now, it’s a common-sense solution using technology to solve a real problem. the solution right there. So, Bernice, you want to take us through a bit of a demo and some thoughts on what you do there.
[17:29] Bernice van der Velden:
Thank you for laying that. As Jeff said, he’s been facing mainly with accuracy and efficiency. And, lastly, you were also talking about that it’s not user-friendly. So we want to be able to have a software where when I’m gone, you can use it without needing consultants to guide you through that. And so that’s where I think HighRadius does a great job is Jeff just log into this application and just check his forecast right off the bat, if he’s in Uber or his hotel. So the biggest thing, of course, that he’s mentioned was the inefficiency. He spends a lot of his time on forecasting. And as a senior, you don’t want to be focused on modeling, you want to be able to make the decisions and create a strategy around those forecasts. So that’s where we can play. First of all, we managed to aggregate all that data, instead of definitely to find that data, making sure that it’s the correct data. We come in and get that all into one place. Then that data is also a large volume of data. And we can put that into this cloud system and in a manner that looks visually appealing as well so that you can direct the see what your forecasts going to look like in the next six months.
[19:03] Bernice van der Velden:
And also, the accuracy part is I’d say, the highest point that Jeff made there is if you want to be able to depend on your forecast. So you don’t want to be having some kind of forecast today. And then tomorrow, it looks completely different. You want to be able to depend on it so that you can make those strategies. So, this is right here, is where Jeff will be logging in. So you can, as I said, when he’s in his car, you can just pull this up and right away, can you still hear me?
[19:45] Bernice van der Velden:
Okay. And you can see what his forecast is going to look like in the next three months. Jeff right now is looking at that 13 to 16-week window, but we’re adding some added value by being able to call that goes to six months as well. By the way, this is not his data. So we’re gonna look at this with any analysis or anything, this is just their demo. So he would be able to toggle to six months and look at a longer-term and make decisions based on a longer-term, not just the 13-week window. For Jeff, especially what we’ve noticed is he wants to be able to rely on his data, on his forecast. So when I showed Jeff this demo he enjoyed looking at these forecasts versus actuals. I think it captures that problem there that he’s facing right now is that he can directly look if his forecasts are making any sense. So with any new model, you want to know how much can you depend on it and this is what that shows. So, you can drill down into the numerical level and take a look at that. I don’t have it on my screen here. So that’s I’m looking behind me. But you can look right here. And so what this means if I mean it might look a little small for you.
[21:33] Bernice van der Velden:
But what this means on December 23rd, the forecast that was made on December 16th for December 23rd was about 67 million. If anyone can direct me there, I think you can probably not see it. But the actual for December 3rd ended up being 61 million. So there it’s showing you that we’re about 6 million off, which I’d say as a relatively good accuracy and look at that and in another fashion, you could go to this bridge and see that for December 23rd, the percentage variance was about 9%. So that’s how this would work and how you will be able to look at it. An added piece that you’ve probably already seen on other sessions as well is that say, Jeff gets a phone call, something big is coming in, or actually, something that knows wouldn’t be in any of the data that we’re getting, we can manually open it up, open this Excel web, so you’re not fully removed from your Excel hell you just put yet. This is what we call our high sheets. So you can enter in, manually enter in some values if you’d like. So, with AI, it can give you a lot of information and it can predict on a high scale, obviously with more data that comes in, it’ll predict with even more accuracy. But of course, there’s some human input that should be always there so that you can add that. And to show you even further, once you add that manual input, you can toggle what your model looks like, with or without that mainland. But you have a question?
To identify what’s the cause of the variance. So I want to drill down and see okay, here it was 9% off. But where was it in the model? Was it the model or was it that I did not provide the right data, which department maybe, there was even a specific department that’s consistently causing the model to be off, is there a way to drill down and see that?
[24:03] Bernice van der Velden:
Yeah, so what he’s asking, and I know everyone heard the start of it, but if you can look at it on an invoice level, drill down further into where that variance is coming from. And that is something that will be there. So you’ll be able to see whether from where that variance is coming from so is it from one specific customer? Is there something on your collection side that is not performing as well as they should be? So yet that functional?
So we’re like a moment of questions really, are you going to show it to me or I have to sort of deep dive and kind of figure it out like what you like to highlight and say this was the thing that’s off, it was this, you know, save me the time of having to do that investigation?
Under A/R is a one cash flow category that will have some sub cash flow category that can be my retail receivers or locked box. And then, this is 85% or 99% that I need to focus more on 85 then there are if I have categorized more sub and categorization can be different. So it’s an aggregate level 85% A/R for company one, whereas for my company two for the same subcategory is 99 so you can pinpoint for this company code for this cash flow category, I have some high variance. So I need to see what type of model is being used. And that shows our analysis report to call it a variance analysis. Invariance analysis, you can drill down to specific combination company code, region, line of business, that’s where specifically I’m getting a variance and cash flow subcategories. So that will help you and you can drill down the system, it will tell you, you need not run around for that support, indeed intuitive in that way.
[26:04] Bernice van der Velden:
To add on to that. I have a microphone. But we are also looking to add different models. So you could compare different models and see which one would fit best for you. So you can compare the variances between the different models. And that kind of goes together with that manual input. So you could compare with that manual input. Say you have a certain tax amount that you want to see how the model reacts to that you could compare that as well. Let me actually switch back to the PowerPoint and hand it back to Jeff, will we have more time for questions?
[26:47] Jeff Martini:
So just in case, it’s not completely obvious what the expectation is, for the future state there is better accuracy and better efficiency. And all in the interest of sort of the punch line here is confidence. Right, that what this is. That’s what it all boils down to. And in my role here as CEO, what is my competence? And how can I confidently manage the future of the business? What levers do I need to pull? What buttons do I need to push? What liquidity do I have available to me to invest back in the business in future capital product projects, or additional headcount? Or, do I need to be thinking about the other direction? Do I need to be thinking about trimming costs and tightening the borders and bringing the cash then we need, to raise prices, etc, all these are what I mean when I want to talk about the levers to pull and buttons to push. That’s the commercial outcome. Let’s say the punch line of all this is actionable, business data.
[28:03] Jeff Martini:
Some, so I know we’ve got some treasurers here in the room, and this is sort of one of the things on the agenda. So I thought it might make sense just to spend a moment here as far as from the seat of the CFO, what is the expectation out of the Treasury Department or more specifically for the treasurer, and it gets back to the decision support, right? So we need a steady hand on the stick and somebody that is, I can rely on for good data, good information better than data, to be able to run the business. So it’s somebody that’s going to bring solutions, not just problems, somebody that is, in tune with, technologies such that they’re prepared to bring new tools to bear and they have the wherewithal to do it, right. So, the idea of a new software platform, that’s all well and good, but I can’t implement something or go through a detailed evaluation, my expectation is that the treasurer would bring the solution along with a capacity to be able to do something with that solution to own it. And to own the only improvement and to own the improved information in the future. So I think that concludes our slides here. And I know we’re tickled to answering questions.
[29:46] Audience 2:
Through the name of the models taken aback by the macro factors, like the price of oil or anything like that?
[29:51] Bernice van der Velden:
Yeah, so actually, in Jeff’s case, we are also looking at a model with that in mind. So what we do and then condition-based, consultants obviously, I know the whole implementation phase. And what we do is we want to understand what your business model is right now. How do you forecast right now? Which criteria do you use for forecast right now? Which criteria would you like to even use for your forecasts right now, but don’t have that capacity to do? So we opened up that conversation with you as well. So in the case for Jeff, I’ll be saying some oil, in the oil industry. So that would be factoring to improve the model as well. Yeah.
[30:37] Jeff Martini:
Right. So the macro impact to the business and more generally, is probably outside the view of a 13 week, because inside 13 weeks, at least again in speaking from my base of experience here, it’s not going to change the capital equipment projects aren’t gonna change for that quickly. It will change over maybe twice that over a 26 week kind of an outlook. But 13 weeks having the most confidence and the sort of this, which is called that high sheet. Yeah, the high sheet can help support sort of turning up or turning down revenue over the weeks and can help inform that.
[31:29] Audience 3:
Did it include all other categories, like A/P?
[31:34] Jeff Martini:
Right, so that’s a phase two that we have in mind. So this phase one is focused on the quick hit, as I call it, A/R is the big pain point for us. Because it inputs both the cash as well as to our bond dates. So it’s sort of a double important if you will, and then the AP is secondary, with the end goal of being sort of put it all in one model and be able to predict out what the cash is going to be and cash balances of one day over the planning horizon. And obviously in that is all there is that’s right. So it includes payroll includes taxes, etc. Right? Yeah.
[32:25] Bernice van der Velden:
With Jeffrey now we’re focusing first on A/R because it is the most difficult to forecast room. But then our phase two would be definitely to add those categories. And that also would be in that conversation. So if I’m talking to you, we will discover what is your hardest to forecast and what takes the most time for you to forecast as well?
[32:51] Jeff Martini:
Right so what’s encouraging about that, spreadsheets view those things that people know that software can’t possibly know or can’t know in any reasonable amount of time so, holidays, for instance, payroll is going to be lower on holiday weeks. So there’s not the sort of, I call it the safety valve is that spreadsheet. So if I know something that software is not going to know about, I can turn it down or if I get a call from a customer that owes a large amount, and then they tell me it’s going to push out a week or whatever, I can use the safety valve which can fix things topside.
[33:36] Bernice van der Velden:
That means that we put everything thoroughly. And if you do happen to have any questions that come to mind later, feel free to stop me or come to our demo.
[33:46] Jeff Martini:
Thank you very much. Thank you.
[0:00] Moderator: Thank you for joining today’s session on the cash forecasting case study of TRIPOINT and expected benefits with AI. We have with us Jeff Martini, CEO at Oil and Gas Production. Jeff loves helping organizations succeed and finds it essential to focus more on the stories and opportunities behind the numbers than just the financials. And also we have Bernie’s vendor Weldon, a functional consultant at HighRadius, Jeff and Bernice, the stage is all yours. [0:30] Bernice van der Velden: Hi, everyone. You’ve already seen my face a lot by this point. My name’s Bernice, and I’m talking today with Jeff. He’s the CEO of TRIPOINT. We’re in the implementation phase with him a little on the first end of it I’d say but he has a great story to tell and I think we are excited that our expectations will be fully met and he has some great expectations as well. So we’ll be covering that in today’s session. So just to give you a glance at the agenda, we will be going through what TRIPOINT is and what they do. And then we’ll be talking about their current cash forecasting practices right now. And then Jeff…
The HighRadius™ Treasury Management Applications consist of AI-powered Cash Forecasting Cloud and Cash Management Cloud designed to support treasury teams from companies of all sizes and industries. Delivered as SaaS, our solutions seamlessly integrate with multiple systems including ERPs, TMS, accounting systems, and banks using sFTP or API. They help treasuries around the world achieve end-to-end automation in their forecasting and cash management processes to deliver accurate and insightful results with lesser manual effort.