Something is very wrong with those spreadsheets

As we have continued to develop our analytics surrounding summary financial projection information, we have continued to find all sorts of things that are not as they should be. The more we see them, the more we automate, and the more we find.

Why is this?

When I first encountered this, and over the last many years, I put it down to one of two things; (1) Incompetence or (2) Intentional subversion.

• Incompetence (a.k.a. naivete): Most folks in startups doing projections don't seem to understand what it takes to run a business. They forget things, make ridiculous guesses, don't think through an actual business actually scaling, and so forth. I think it's because most of them have never run a company before. So many of them hire an accountant or someone who is supposed to know what they are doing. Most of these accountants apparently don't know how rapid growth works, or don't know about the specifics of the kinds of businesses they are helping with, and just ask questions of the CEO and fill in accounting magic to generate answers.
• Intentional subversion: These folks seem to want to tell me what they think I want to hear, that they know what they are doing, that they will do amazing things, and that I should send them my money for their experiment. Some, of course, are frauds, but I'm not talking about those folks here. I am talking about the people who insist that they are completely rock solid and growing sales at better than the required pace, or fully able to scale execution having never delivered anything, or having made one sale. They seem to not get it that I want the truth, not a best case possibility, and not their overly enthusiastic vision of a possible future. They seem to think they are selling me when they are actually putting me off.

But I think I might have been wrong on this one...

I think that there may be another huge reason we get ridiculous nonsense out of projections that don't pass the smell test when you start to look for strange stuff or inconsistencies. I think it might just be...

The spreadsheets

Yes, the spreadsheet is a fantastic technology for all sorts of stuff. The only problem... well not the only one, but... is that it's nearly impossible to get a spreadsheet right for anything complicated. And even if you should happen to get it right, there would likely be no good way to prove it, at least today. And as soon as you change something, you will likely never fix all the indirect implications it leads to.

The complex interplay between different elements of running and growing a company is not really well reflected in a typical spreadsheet. There are lots of places where you hire or don't hire someone, and while this makes little financial difference at large scale, at small scale, the 3rd, 4th, 5th, etc. employees usually drop cash flow further negative just as it's about to turn positive, and by delaying it by a month, the cash flow is better but the long term performance is worse. This kind of variation of parameters analysis is beyond what almost anyone can do well with a spreadsheet, and calls for a more advanced simulation capability.

Another big problem with spreadsheets for these functions is that they tend to operate on a monthly or quarterly basis, whereas bills and payments tend to come in on a day-by-day basis. The spreadsheet may say that at the end of the month or quarter you are OK, but the actual cash demands in between may be for greater than the spreadsheet shows, which is one of the many reasons companies hit the ground even though they would be OK if it weren't for the lack of fidelity in their analysis. Add to this early outlays and late income, and you have a crash waiting to happen.

Yet another spreadsheet problem is the variations in probabilities of moving from step to step in a sales sieve. Very few spreadsheets take into account different results from each step in a sales sieve, and this gets really complicated because:

• More leads moving forward at an earlier step may end up being worse than fewer of them because the earlier you differentiate the ones that will end up as customers, the less expensive it is to sell to the rest of them;
• As you execute a sales sieve, you get partial results for each step, which the spreadsheets are rarely updated to reflect. By updating the analysis with the feedback you get along the way, you can adapt your tactics and projections BEFORE the end of the sales cycle, which means you can try to fix the problems before they cause bigger problems.

Of course all of this can be worked out in a complex enough spreadsheet, and by constantly updating it to reflect changes, but that gets really complicated and error prone.

Alternatives

The path we have taken in most cases is to implement applications customized to specific business functions.

• Cash flow analysis seems to be the biggest issue for companies close to the edge, as most startups are. We have implemented various cash flow simulation systems over time, and typically use them at most quarterly to give us a sense of what to expect and how much cash reserves we will need for different plans. But once we have a plan we rarely change it significantly more than once a quarter.
• Sales sieve analysis is closely related and also critical for keeping sales on track, adapting tactics rapidly, and recognizing the real resource requirements to hit the numbers. We automated this many years ago and have made minor adaptations since. This is something we update at just about every step in a sales process by putting in the new numbers and seeing how bad(or good) it is. This level of adaptation is critical to debugging sales processes, but once they are working, we tend to update them only as part of our quarterly cash flow process.
• Projections seem like one of the hardest things to make sense of because of all the results that come out non-sensical. Analysis of projections for inconsistencies and other concerns is something we have automated and continue to rework, but it saves a lot of time and effort in trying to check results of other analysis. We rarely re-look at our projections more than once a year, largely because we already know we are or are not meeting them, and annual reports should accurately reflect what we actually know. But projections before funding startups is another world. Here, we adjust until they make sense and then reconcile with the spreadsheets or simulations to get consistency and clarity of understanding. We also do a deep dive into an accounting perspective on it to make sure we aren't missing anything obvious (to an experienced person running the same kind of business for the last several years).

Integration and Updates

Ultimately, all of the information about a company and its paths forward should be fully integrated and change controlled so that when you propose to change one thing, everything else is updated and all the changes reflected in the proposed change with all affected parties notified and a management control system supporting it all. But that's not going to happen, or particularly help, small and early stage companies much.

So we remain largely not integrated, but rather, connected. We take the approach of making it easy to check things out by allowing rapid viewing of related material. And we don't so much control change as record it all so it can be undone if desired, or at least, previous versions are available for inspection and use if desired. It makes it more efficient and easier to find what we are looking for, without all the harsh discipline of rigorous controls over inter-related applications and content.

I also thought I might mention that integration between and across spreadsheets is a function they have, but it's really complicated to get right, and very often fails because spreadsheets are sent from place to place without the relevant related content. They become something like a database with complex calculations, but without the central control of a database to support availability of relevant information or the ability to remotely go get it when you want/need it.

We work closely with Chris Blask and the folks at Quietwire.AI for a range of things. Recently, they have experimented with substantial success in using AI to search and combine content from across companies of all sizes through narrative, and it seems to be a direction that can succeed in largely solving the top level integration across an enterprise. This is closely related to the work on standards of practice that we use in decision-making and structuring understanding for cybersecurity and business development (at Management Analytics). There is a lot of potential in this arena, and over the coming years we anticipate that attribution infrastructure for narrative integrity combined with improved linguistic processing and modeling will make a fused understanding and capacity to query and analyze businesses will greatly improve. To a lesser extent this is already done in our GWiz™ and Diligence systems through their internal and external automation (some would call it AI) functions.

Conclusions

Spreadsheets are great tools, but we long ago exceeded their limits in understanding companies. A new integrated approach to analysis is needed and being developed even for small companies and startups.

More information?

Join our monthly free advisory call, usually at 0900 Pacific time on the 1st and 3rd Thursday of the month, tell us about your company and situation, and learn from others as they learn from you.

In summary

Spreadsheets are great but almost impossible to get right for anything complicated.