[apohn]

I work in the field of Data Science and one upsetting reality has started to sink into my mind over the last year.

In a business there is top line and bottom line. There are a lot Statistics/ML/Data Science jobs that are about moving that bottom line. You build something to optimize something to reduce costs.

The value provided by the bottom line people is less visible than the value of top line people. The easiest way to move the move the bottom line is by just getting rid of people. So when the axe falls the bottom line people get cut and it's hard to understand why.

It's the same thing as people say about fires. When you put out a fire you are a hero. When you prevent the fire in the first place, everybody thinks it's business as usual and nobody understands why you are needed.

[a4isms]

> It's the same thing as people say about fires. When you put out a fire you are a hero. When you prevent the fire in the first place, everybody thinks it's business as usual and nobody understands why you are needed.

I got a dose of very cold water about this thirty years ago when I was building payware that improved developer productivity. I gave a presentation about its ROI, and afterwards, a developer walked up to me and gave me some feedback that none of the business-types had articulated:

Products are either vitamins or painkillers. People buy painkillers, because they're in pain. People postpone vitamins, because nothing is wrong and the benefits are always "later."

I didn't 100% change what I chose to build over the years, but from that time to today, I have worked on always spinning what I sell as an antidote to a customer's pain point, rather than as an investment they make to pay off eventually.

p.s. I don't know where that dev got the "vitamin/painkiller" metaphor, but it's sticky!

[dathinab]

Ironically this quote also show how broken the US is: It's normal to take pain killers.

It should not be.

It should be a last resort.

You should take what fixes the problem and give your body time to heal not take pain killers and pretend nothing is wrong.

Pain killers are addicting, can have an increasingly reduced effect, can have a bunch of side effects and can make the end result much worse by not healing wounds (metaphorically) when they are still easy to heal(1).

(1): Through sometimes they can also help you healing by preventing you from doing pain-caused bad actions, like setting down your food in a bad angle.

EDIT: Just to be clear I mean pain killers for a "normal live" situation, not in context of you lying in a hospital bed or having extrema healthy issue which can't be fixed/heal anytime shortly.

[fshbbdssbbgdd]

It’s important to draw a distinction between narcotic painkillers like opioids and safe painkillers like anti-inflammatories. In general, it’s safe for people to occasionally take certain painkillers for minor pain. In some populations, drugs like aspirin can even extend life when taken daily.

[nemothekid]

>Ironically this quote also show how broken the US is: It's normal to take pain killers.

I've heard this metaphor before, by a VC, and it was medicine vs vitamins.

[StillBored]

Chuckle. I wonder about the "leopards ate my face" moment which seems particularly apt here if you consider Si valley VC money to largely be expensive steroids that appear to provide big growth fast, but will likely land you in worse shape long run either when you withdraw them, or by leaving you addicted trying to avoid the resulting crash.

AKA your removing the "vitamins" path in exchange for some future "medicine"

[treeman79]

A real fun one is rebound headaches. Spent a few months with horrifically painful headaches. Turned out it mostly from painkillers. More I took. Worse headaches Got.

My other less painful headaches that started the cycle were an actual brain issue. Just took a few years to get correct diagnosis.

Eventually had a cycle of one round of pain killers every other day. Cycling through To a different kind each time. This mostly worked until I got excess brain fluid drained off. Which actually solved issue.

[Obscurity4340]

What was the ultimate underlying pathology?

[treeman79]

High cranial pressure (IIH) and lots of minor strokes.

Apparently I’m genetically prone to clots. (Factor 5)

[fanick]

Damn, I think I might have similar issues. Also diagnosed with Leiden. Should I be worried?

[KptMarchewa]

My head hurts because atmospheric pressure changes fast. If you solve this problem I'll stop taking painkillers, and in the meantime you could stop using appeal to nature fallacies.

[fleddr]

It's accounting bias/culture.

Say you have a 100 developers and you reason each should get a second monitor worth 300$, because this increases productivity by 20%.

According to an accountant, you just added 30K in costs to the books, with nothing to show for it. You can't eat productivity nor is it a line item in the books.

Who is to say that this 20% of freed up time is used productively? Or used on things that increase revenue? If so, how much revenue? And when?

[midoridensha]

>Say you have a 100 developers and you reason each should get a second monitor worth 300$, because this increases productivity by 20%.

>According to an accountant, you just added 30K in costs to the books, with nothing to show for it. You can't eat productivity nor is it a line item in the books.

Right, this is why developers should NOT get 2nd monitors.

Even better, the business can save another $30k by not getting the developers any monitors at all.

[jlokier]

You joke, but my current job sent me a Dell laptop with a tiny screen and a fancy 4-video port USB-C dock. Seem I'm expected to have my own monitors.

(Turns out I do have my own monitors, but they are for my own laptop, and are HDMI and Thunderbolt which the dock isn't, except one HDMI.)

[Eisenstein]

If you can't quantify it, it doesn't exist, and when you start to quantify it, the numbers turn into targets that must continually improve.

[fleddr]

Yep, and I think we can apply the accounting logic to the original article.

A team that regularly saves costs for other business units is a promise of cost savings. Dropping the entire time is an immediate and factual cost saving.

Short-sighted? Yes.

[zemvpferreira]

It's a trope with some truth to it, but it runs out of steam fairly quickly. Was original facebook a painkiller? Instagram? $1000 iPhone? Liver King?

I find it's a useful framework for selling b2b. Even then, desire can win over pain many times.

Fear and greed are the real big sellers in b2b anyway.

[matkoniecz]

> Was original facebook a painkiller? Instagram?

I would say that it was closer to a tasty pizza.

It was definitely not fitting either "vitamin" (worth investing for future payoff) or painkiller (solving immediate and urgent need[1])

[1] I guess that hiding/temporary fix is not intended to be part of this allegory

[fxtentacle]

At my university, Facebook was the painkiller for involuntary celibacy ;)

[astrange]

Lots of people run businesses out of their Instagram accounts. Might not be what it was for, but those followers can be valuable.

[a4isms]

It doesn't run out of steam so much as it fits a particular scope better than others. We can bend over backwards to make it fit other scopes, like another commenter points out: FOMO is a headache, and social media could be called the aspirin.

I personally find that metaphors work best as mnemonics. Unless you're in health sciences, products are neither vitamins, nor painkillers. If this metaphor helps us remember to distinguish products that solve urgent and existing problems from products that are investments providing an ROI over time... Good enough for me.

Now about fear and greed. I agree with you. In fact, I do so because when I was in sales and marketing, I read a book that said that the four motivations that mattered were fear, greed, exclusivity, and belonging. The first two are literally what you just mentioned.

That particular simplification doesn't fit every situation any more than Maslow's Hierarchy does, but it's another surprisingly useful lens to use when looking at value propositions.

At the end of the day, all these rules of thumbs and metaphors are tools, if you find one that's useful, add it to your toolbox and figure out when it works and when it doesn't. The more of these you have, the more ways you have of analyzing a situation and coming up with a rough model for how it works.

So I agree with you: Fear and greed are the big sellers in B2B.

[OkayPhysicist]

The original facebook was a painkiller the same way the Oxycodone you crush up on a table and insuffulate is a painkiller. The metaphor works amusingly well, actually.

[beambot]

Youthful hormones and social belonging are pains too... Consumer pains are often more abstract.

[mjevans]

iPhone was / is a 'status symbol' and 'fashion accessory', which happened to be way better than the clunky, expensive, and poor UI mobile phones which came before, (aside from Blackberry, which was a corp status symbol, work / gov focused, not average consumer.)

[astrange]

What actually happens with vitamins is people love taking them (because they’re colorful and some of them are food preservatives) but there’s like no evidence they have health benefits.

[rurp]

I don't know, I was surprised to learn that my vitamin D was quite low, and I get a fair amount of sunlight each week, probably more than a lot of office workers.

Now that I take supplements though my levels have been fine. From what I have read, quite a few people fall into a similar bucket.

[mirker]

Do you feel different? There is a theory that vitamin D is merely correlated with a healthy body. Sunlight may be required.

[rurp]

Honestly no, but I believe there is some research showing potential long term health benefits of avoiding low vitamin D. It's totally possible I'm wasting my money, but I'm happy with the potential benefit/cost ratio. I would gladly pay the price of some vitamins for even a small chance of avoiding significant health problems.

[mannykannot]

...unless you actually have a deficiency.

[TomSwirly]

The people in the United States who can afford to buy and consume vitamins are almost certainly not people with a deficiency.

[TremendousJudge]

Most people in the US are vitamin D deficient, it's very cheap, yet it's rare for people to take supplements.

[astrange]

Yes, D+K is the best one to take. D only can lead to heart issues (atherosclerosis), and multivitamins don't really have enough to help here.

It doesn't replace getting real sunlight though. Or if you're an Inuit, eating polar bear livers.

[layer8]

Technically though, vitamin D is not a vitamin. ;)

[mannykannot]

I happen to be one who does, and no, it's not a consequence of a bad diet or unhealthy lifestyle.

[abakker]

Cutting costs is always a marginal thing, because businesses tend to value growth. Oversimplification: If you have a 50% margin business, the value of one more dollar of revenue is $.50. If you cut costs and change the margin to 55%, then you've added only $.05 of revenue to that additional dollar.

Now, a sane person will look at the improvements to margin across the whole business and still want to make those improvements because in aggregate, they add up, BUT, you cannot improve margin forever as a strategy. Eventually, hard limits come up and the incremental gains shrink and shrink. At that point, growth dominates.

Most mature businesses need revenue growth much more than they need marginal internal gains, especially because as businesses get bigger, marginal gains tend to apply to more limited segments of the business. E.g. improving one product is marginal and applies to only the sales associated with that product.

I think the claim that data science is about moving the bottom line is right, but I think the other way of thinking about this is that Project/Consulting is probably a more relevant way for companies to buy these skills than Salary. Many companies can see the value in an incremental move in the bottom line, but most companies don't have a sufficiently large problem space to worry about paying a continuous cost to focus on this.

I've seen a lot of big companies say that they need these skills, but also believe they can't attract talent because they wouldn't be able to keep a data scientist busy.

[a4isms]

I've been a part of this argument before. I have another, additional perspective on why growth is more important than cost-cutting in many cases. If there are costs to be cut, you can cut them today, you can cut them tomorrow, they're right there and eventually, you can hire someone/buy something to cut those costs.

But growth is a tricky thing. If you're in a land grab market and you cut your costs at the expense of growth, you may find that you lost your chance to grow, because the market is now dominated by other people.

For people with this mentality, they expect in the long term to cut costs, but only after growth has slowed for reasons out of their control, e.g. the makret is stabilizing and has already chosen the #1 big gorilla, the #2 little gorilla, and numbers #3 though #100 small monkeys picking up scraps.

[i_am_proteus]

And if you cut costs in a (prospective or current) operating area from 120% of revenue to 90% of revenue, you've opened up an entire new operating area to profitably grow in.

Developing the technology to do a thing profitably that previously could not be done profitably is the stuff unicorns are made of.

[abakker]

Absolutely! I hope my reply didn't imply that I thought there was no value in doing things more efficiently. There clearly is, and as consumers we love marginal gains in product quality, efficiency, and price.

I'll nitpick a bit to ask, though, how many times has a new entrant to a market gotten a process/business/tool/etc from 120% operating to 90% through marginal gains? I'd wager almost never. Process improvement can be marginal or stepwise/punctuated. I think most unicorns create punctuated change in ossified industries, but, I don't think any big companies are likely to hire a data scientist and through years of grinding through the margins achieve that 30% improvement.

put differently, the decision to focus on revenue vs profit is a decision that typically does not include the NPV of R&D investments. those are uncertain and have some probabilistic value, but not so much in accounting terms.

[marcosdumay]

Competition in mature industries is all about processes costs becoming 110% or 90% of revenue, usually through a series of marginal changes.

That's not how unicorns are made, but it's how most of the economy operates.

[Karrot_Kream]

> Eventually, hard limits come up and the incremental gains shrink and shrink. At that point, growth dominates.

The trick is understanding where the hard limits are. I've noticed that upper leadership tends to be pessimistic about these hard limits (they come quickly) and engineers on these teams tend to be optimistic (there's a lot of fat/cost to cut so the hard limits are quite far down.) Now naturally, the engineers on these teams have a vested interest in being optimistic, as their team charter is based around their work. But I've seen this conflict play out in many organizational situations and I'm not sure this interplay between upper leadership and engineering about these margins is illuminating for the business.

[dh2022]

Not to nitpick, but 5% improvement profit does not apply to the additional dollar. It applies to all the revenue.... So the improvement could be massive.

[abakker]

That what I meant when I said the decision would be weight to do that because in aggregate it pays off, but, the payoff for those one time things is not fully retroactive. E.g. for sold products it does nothing. For services, it can be much better!

[Karrot_Kream]

Yeah I've worked in infrastructure through most of my career wherever such a distinction is available (or when it opens up), and this is a common complaint. Product folks get the most visibility and get kudos and parties for product launches. Meanwhile, the deployment infrastructure staying up is just expected, even though the engineers responsible for it are working hard to keep it up. It affects team morale (infrastructure teams are unrecognized for their hard work) and also has material affects on promotions and compensation as it's harder to justify business impact on these teams. I know folks that left infrastructure teams because of this dynamic.

[treeman79]

Hired into a company. First day on job I find that the entire infrastructure team had quit. It was in a failing state.

told them flat out that they are most likely going out of business, but I’ll get it a try.

Couple of times owner tried to Ask me when feature X would be delivered. Just told them no. Managers were wise enough to understand they were one pissed tech guy from failure.

3 years of endless late nights to get company back to a good spot with a rebuilt time, new infrastructure. Proper documentation, the works.

Finally left after being passed over for promotion to a guy that did nothing, but promised the world. (He never delivered)

Took me a couple years to recover from that job.

I don’t work late nights anymore. If company doesn’t care to invest in infra, I look elsewhere.

[anonreeeeplor]

These lessons are just brutal.

If you are high in EQ and vaguely likable you can substitute for technical skill or hard work.

I had a colleague who was like a golden retriever and lacking in all talent. Everyone loved him. He never got anything important done and always worked on superficial shiny objects.

He was basically untouchable. Being optimistic and having no talent is a huge advantage.

The hard working workhorse industrious person always griping about how broken everything is: Can’t wait to get rid of you.

Don’t do great work for morons.

All of the collective results of these brutal lessons for me has been to become ultra cautious about where and who I work for and to try to do a much better job reading the room and analyzing people.

If you are an Aspergers person, this is super hard. I now do my best to get multiple second opinions about the situation because I learned my personal judgement and evaluation was nearly always wrong.

[andrewflnr]

It's tricky, because there's genuine uncertainty about whether you have prevented a fire, or just wasted some time and maybe added some overhead. Even people who understand a system deeply can have reasonable disagreements about whether a preventative measure is worthwhile. Executives whose only interaction with the system is feeding it money have almost no chance of figuring it out in the face of any amount of conflicting info. And of course a mixture of natural human optimism, aka blithe disregard of danger, and having their salary depend on believing there are easy things to cut, makes it quite difficult for them to believe in any particular instance of a fire prevented.

I hope it's clear that I don't mean to excuse them for giving up. It's hugely destructive both for decision makers and everyone around them. I just want to show that the problem is substantially harder than "just reward preventing fires already".

[brutus1213]

I understand the top-line bottom-line divide, but I am not fully convinced if the top-line projects are any safer. Wouldn't another reasonable business strategy be to get rid of all new projects, and only focus on operations-as-is during times of economic uncertainty?

[angry_octet]

That's exactly what weak management does. Family management is especially prone to this IME. Cut new investment, cut cost of inputs, labour, quality control.

That works as long as you have weak competitors (or a moat) and nothing terrible happens, like high defects. Essentially you're coasting on prior investment. But as soon as something changes in the market you're falling behind.

What I've often observed is that new low cost competitors introduce features which are often reserved for high end devices/products due to market segmentation. The dominant player refuses to adapt and hence they lose all their low end market share, the volume of which is necessary to make the whole thing work. Meanwhile new customers start with the lost cost ecosystem.

I've seen this happen with e.g. Agilent, or SaaS companies, who charge 10x for something that costs little, like SaML/AD auth.

Imagine if NVIDIA had charged for CUDA or considered it a distraction from selling graphics cards. They wouldn't own the HPC/ML space if they had done that.

[pixl97]

That depends if you're about to get ate by your competitions new product

[nemo44x]

That would be an extreme action. You do still need to be working with the future in mind. Anything that looks promising to revenue growth in the nearish future should probably continue to be invested in. You may ask those teams to become more scrappy and figure out how to achieve their goals with minimal new investment, especially if the new revenue streams are still a few quarters from coming online.

[jldugger]

> Wouldn't another reasonable business strategy be to get rid of all new projects

Only if you want to close the company in ten years

[kenjackson]

A very simple question that I've had to ask is "what likely happens if we cut this group?" then "what's the 'likely' worst case if we cut this group?"

That problem with Eric's group and most Data Science teams is that the company continues to move along. There is some long-term cost, but there are likely teams where there are severe short-term ramifications if they are cut. E.g., imagine if Windows cut their servicing team (snarkiness aside).

[angry_octet]

It's a failure of the data science team management that they didn't make themselves a front line capability. It is easy for OR (Operations Research) to explain their business value, any DS team that only stays at the tail end of building capability is liable to be cut (or under invested).

For DS it might mean being more on the market research / customer requirements / subscriber churn side, instead of being on the back end of services improvement / risk reduction. Be the thing that customers are asking about, that brings new customers.

[nightski]

I think this is an insightful assessment. Not everyone in a company can be top line. But I also think there's a lot more opportunity in using statistics/ml/data science in the top line than most companies practice.

[apohn]

>But I also think there's a lot more opportunity in using statistics/ml/data science in the top line than most companies practice.

I consider myself a fairly honest Data Scientist, in the sense that I like it when I can map what I'm doing to the value it delivers. I know some other great people I've worked with who are like this as well.

This is anecdotal, but all of us have hated working with many top line people because there's some really fuzzy mapping from goal to value (since value is realized in the long term), and some of the people are champion bullshitters. I don't need to explain sales people. But marketing, corporate strategy, and even upper product management - they drove us crazy because their standard of being data driven was absolutely not consistent with how we thought about things at all. All of it was because the mapping from project to revenue was over years, not quarters. And it was all projections.

Compare this to bottom line people, where the mapping from project to cost savings is on a shorter time frame. The types of personalities this attracts is different.

Maybe the growth hacking stuff at software companies is different and you can focus on revenue growth and still connect what you are doing to that. I've never worked in that role so I don't know.

[greesil]

But that would lead to accountability...

[kneebonian]

This is the real problem. Visibility is to be abhorred at the top levels because viability brings accountability. How many Dilbert comics are there out there with the punchline being "I don't care what the real numbers are these are what I want the numbers to be" from the PHB

There is a large swathe of middle and upper management that gets by due to continually making sure their actual impact is never measured, and they are only a "force multiplier." not that you should do away with middle management, but there are many in middle management who could be done away with, with very marginal loss.

[angry_octet]

OMG the 'Technology Foresight' group, the 'Process Improved Team'. Cross functional synergy!

We all know what the problems are and where investment is needed, but management pretends that they don't know so they can have An Initiative to discover it, but not really address it (because e.g. the problem is one they caused with previous poor management).

[bravetraveler]

This is a running joke for every systems administration/operations job I've had

A common theme for commiserating, the only investment we get are complaints

Make it work again with what you had or we have problems, must avoid OpEx at any cost

[higeorge13]

Yeah that’s especially unfortunately true for data science and data engineering teams in companies where ml or data are not the core business but nice-to-have. They are usually the first ones from engineering being axed in times of lay offs.

[WorldMaker]

Even for companies that have ML and/or Data in the core business. I think few would argue Meta in this specific layoff example doesn't have data as a core business.

(And those few are probably the ones drinking the "metaverse Kool-Aid" and thinking the pivot away from data siloes is already complete to some sort of VR scape where data somehow doesn't matter or doesn't exist, that Meta still hasn't actually convinced consumers to buy or figured out how to build. They finally figured out "legs", pivot complete I guess?).

[MattGaiser]

A friend had a job where a team there just let things fail rather than prevent fires. Lots of raises and praise for literally not doing their jobs.

[midoridensha]

It sounds like they did their jobs exactly the way management wanted them to do their jobs. The proof is that they received lots of praise and raises.

[dh2022]

Cutting costs but bringing no revenue shows as Cost Center on any financial report. Revenue though shows up as Revenue center. Thus this decisions which sometimes are illogical. Sad but true :)

[serverholic]

This is one of the reasons why I think making the workplace Democratic is a good idea. The workers have a better idea of what is important than the management.

[qbasic_forever]

Incorporate as a worker cooperative and not a corporation beholden to shareholders.

[substation13]

The potential gains from cost saving is always capped at total cost. The potential gains from increasing revenue are (typically) much larger.

[acdha]

Also cost savings has a hard, well known upper bound but revenue growth is speculative with many opportunities for pleasant fantasy. Business leaders love the idea of being the visionary who takes a big gamble and makes it work.

Facebook is an example of where that breaks down: there isn’t an easy way to grow that much larger so they would likely see greater return from cost savings than they are likely to make from VR, but after a couple decades of thinking of themselves as this incredibly innovative tech company it’s hard to accept that they’re stable as an ad company.

[TigeriusKirk]

There's also cost savings that are numbers shuffling on a spreadsheet and then there's cost savings that are actually less money leaving the corporate accounts.