clock menu more-arrow no yes
Theranos Founder Elizabeth Holmes Attends Criminal Trial David Odisho / Getty

Filed under:

The Holmes trial gets into the Excel nitty-gritty

But will the technical jargon overwhelm the jurors?

It was around the time that the defense counsel referred to “deferred revenue” that I began to feel genuinely bad for the members of the jury in the wire fraud trial of Elizabeth Holmes.

Tuesday’s testimony involved the nuances of accounting, how quality control works in labs, and way too many Excel spreadsheets. I get it — you have to show the financials to show that Theranos was in a tricky financial position. And you have to show how the labs were failing in order to establish that they were failing and that Holmes knew it. But go through this too fast, and you risk confusing people.

The case’s first witness, Theranos’s corporate controller, Danise Yam (she also goes by So Han Spivey), gave testimony that was crucial for making the government’s case. Every scam is, at bottom, about the same thing: getting money. If Elizabeth Holmes committed wire fraud, as the government alleges, it would have been with the goal of getting more cash.

To that end, prosecutor Robert Leach had Yam go through a series of Excel spreadsheets, often attached to emails. There were two major cash crunches, Yam testified: one in 2009, and another in 2013, when the cash position “started to get a bit tight, not to the extent of 2009,” but enough that Yam was sending updates to Holmes and Balwani on a weekly basis.

In 2013, Theranos was burning about $2 million a week.

In the financial statements for 2012, there is no line for revenue — because there was none. There’s none in 2013 either. This is not, in and of itself, especially damning, but it does paint a picture of a money-losing company on the ropes; by 2012, Theranos had lost $161 million over its lifetime, and it lost another $92 million in 2013.

In the course of working at Theranos, Yam consulted with Aranca, an analytics firm, to price stock options for Holmes and other employees. To do this, she consulted with Holmes, because Holmes had “the best information” about what financial projections to give to Aranca. In one email from Yam to Holmes, Yam asks to use numbers she’s previously used. Holmes writes back, estimating that 2015 revenue will be about $100 million.

This was all fine and sort of boring, until another document was introduced, one Yam hadn’t seen before.

This document, Leach said, had been given to investors. It showed revenue projections of $140 million in revenue in 2015, and $990 million in 2016. Not only did Yam not provide numbers for this, she didn’t know where these other numbers came from.

Look, that’s weird. Let’s say you’re giving lower numbers to the people who price your stock options. That means that your options will be worth more if the higher numbers — the ones you gave to your investors — are right! On the other hand, let’s say those higher numbers you gave to your investors are wrong, and the lower numbers you used to price your shares are right. That, um, seems like maybe you lied to the investors in order to get their money.

The discrepancy here doesn’t establish intent — the government has to show that Holmes knowingly lied — but it doesn’t look good. Which is probably why the defense counsel got up to talk about deferred revenue.

Lance Wade, Holmes’ lawyer, didn’t exactly define deferred revenue, which is the accounting world term for advance payments for goods or services that aren’t yet delivered. (If the goods or services aren’t delivered, a company may have to give that money back; deferred revenue shows up as a liability on the balance sheet until the work gets done.) In response to Wade’s questioning, Yam says she provided Sunny Balwani — Holmes’ co-defendant, being tried separately — with Theranos’ annual deferred revenue estimate: $169 million.

So Theranos’ situation wasn’t as bad as it looked, Wade argued: money was coming in from deals with Safeway and Walgreens and some other companies — Theranos just couldn’t mark it as revenue on its books.

Okay, but what about the mismatched projections? Like, I’m not a math whiz, but $100 million, the 2015 projection given for stock option pricing, and $140 million, the 2015 projection given to investors, are different numbers.

I didn’t hear anything satisfying to explain that. Wade instead made the argument that different accounting practices can lead to different outcomes — pointing to another part of the document the Aranca consultants had prepared for Theranos as proof. See, one way of valuing Theranos meant it was worth $1.9 billion; another way meant it was worth $9.5 billion. This is not quite the same thing as using different numbers when you’re speaking to different audiences, though!

After Yam stepped down, former Theranos lab associate Erika Cheung took the stand. She’d joined Theranos directly after UC Berkeley, and at first had been “starstruck” by Holmes.

The secrecy started immediately, in the job interviews where neither Holmes nor Balwani would say exactly what Theranos did. Cheung was presented with a non-disclosure agreement on her very first day. Then, Holmes’ brother, Christian, made it clear that Theranos employees couldn’t list the company on their LinkedIn, or describe their responsibilities.

Cheung was only at Theranos six months before she quit.

Before that, though, Cheung did work on validation, a series of experiments that were run to make sure the results were accurate and precise. Samples of Cheung’s own blood were run as part of validation testing. Though traditional methods showed she was in the normal range for vitamin D, Theranos’ machines consistently showed that she was deficient.

Cheung also discussed a similar kind of testing that is used to calibrate equipment on a daily basis, quality control. As part of the process, you take a sample with results you already know, and run it through the machine. If the machine’s results don’t match the sample, you troubleshoot to see where things went wrong.

But Cheung’s testimony called into question how good the quality control was. She’d emailed an internal helpline because, despite troubleshooting her machine, she couldn’t get it to work properly for the vitamin D test. This chain is escalated to Balwani and Holmes.

The problem was resolved, but not in a way Cheung was comfortable with. See, the vitamin D reading came from six data points, and to get the machine to pass quality control, two “outlier values” had to be discarded. That was when we broke for the day.

Cheung did a better job than either the defense or prosecution lawyers at clearly explaining technical concepts, but even so, it was a lot to throw at people all in one day, especially after the bickering over accounting.

Given the presumption of innocence and the standard of reasonable doubt, confusion works in Holmes’ favor. If the jurors get lost, it’s easier for Holmes to win — because if they can’t keep this all straight, then it might seem reasonable she wouldn’t have been able to, either.


Hospitals say cyberattacks increase death rates and delay patient care


How Elizabeth Holmes sidelined the real scientists at Theranos


Making Elizabeth Holmes’ cringey texts public is the ultimate crime deterrent

View all stories in Science