Comment by bmir-alum-007 | Hacker News Reader

Comment by bmir-alum-007 | Hacker News Reader

Parent Full thread

Stanford was taken to the cleaners to the tune of $1.5 x 10^8 USD in the deployment of Oracle Financials and related products via endless "consultant implementation" charges that didn't really deliver much value, were rarely on schedule or on budget. Oracle's enterprise calendaring program was totally inadequate and had UX that made most point-of-sale systems look effortless by contrast. Also, the assets managing app, Sunflower, was another dud. The only thing Oracle Database had going for it was no crippling license activation (license scofflaws are/were sued or fined into oblivion worse than M$FT), which one could say was equivalent to MS SQL. Unlike MS MSQL, Oracle DBMS has/had bazillions of support patches to apply to run a real production box, analogous to the previously separate Sun's Solaris patchsets.

Btw: For smaller enterprisey shops, either MS SQL or Postgres are the way to go. Often multiple similar components are needed because different apps have firm requirements that only support one or another; but generally try to avoid this because supporting too many heterogenous components is expensive (laborious)... hence the prevalence of local "standards." Deploying everything with cfengine3 or puppet can help reduce the manualness and nudge vendors into repeatable, idempotent deployments rather than clicking on inane GUI installers like an animal.

ProTip: Don't let consultants "provide" oversight / free-reign for their own projects, budgets, etc., that's like the wolf guarding the henhouse. The client must hire their own project managers, have clear accountability/authority paths to their management and know exactly what they need (avoid endless upselling). Or lots of money will be transferred from idiots to crooks (enterprise caveat emptor).

Edit: fixed grammar

I worked for a large company which will remain unnamed, who a few years after purchasing a 12-year unlimited support contract, switched to using Postgres. Read: they were already locked in to paying for Oracle's most expensive contract for 12 years, regardless of what services they used. In short, they concluded that due to the cost of developing for Oracle, it was cheaper to migrate to Postgres than to continue using Oracle even when Oracle was free.

It is somewhat rare that companies will recognize the fallacy of sunk costs ("We've already spent so much, we need to do this"). I think the world would be a much different place if we could somehow overcome this cognitive error.

makes for good examples that everyone can learn from

easier said than done.

I see this repeated all the time and tend to label it "the investment-bias fallacy" but it's also called commitment bias. [0] Go into any casino where people are trying to win back their lost money is one classic cliché.

0. https://en.wikipedia.org/wiki/Escalation_of_commitment

"Sunk cost fallacy" is what my college "Engineering methods" course called it, but we are talking about the same concept.

In case anyone were wondering: Postgres deliberately resembles Oracle. EnterpriseDB is one of the official Postgres commercial flavors which is an even closer "clone" of Oracle DBMS with support options to make migration from ora easier.

> Postgres deliberately resembles Oracle.

I don't think that's really the case. We don't try to do anything different for the sake of it, but that's pretty much that.

> EnterpriseDB is one of the official Postgres commercial flavors

There are no 'official Postgres commercial flavors'. EDB is well known and contributes a fair amount to PG development, but that doesn't make them official.

[citation needed]

I guess one could say they deliberately resemble Oracle in that both DBs roughly implement the SQL standard, but this is by no means a clone, nor was it a smooth transition.

> to the tune of $1.5 x 10^8 USD

Was it really necessary to type it like that?

For general sanity: $150,000,000

To be pedantic, 1.5e8 carries additional information about accuracy: $150,000,000 ± $5,000,000. By contrast, 1.50e8 would be ten times as accurate, and so on.

Of course, accuracy is hardly relevant here.

s/accuracy/precision

Sorry, I don't usually try to be pedantic, but when the conversation is already about nitpicking, I think it's necessary.

Significant figures are precision, not accuracy. Accuracy is "being in the ballpark". Precision is "tight groups".

I'm sorry for the confusion. English is not my first language, and I resorted to https://en.wikipedia.org/wiki/Accuracy_and_precision to pick between "accuracy" and "precision".

To put it another way: "pi is exactly 3" is extremely precise, but not very accurate.

On the other hand, if you had said "Pi is 3", you would have been accurate, but not very precise. If you said "Pi is "Pi is 3.14", you would have been so precise that you could have gotten to the moon and back.

Damn, and here I thought I needed to build a gigantic rocket.

My statement was accurate, just lacking in precision on the details :P

The very next task of mine

is a new value of pi to define.

I think I'll use 3

it's much simpler you see;

than 3.14159

You've simpified pi, that's fo' shore.

But rivalry says "level-up score".

With competitive drive,

It's yet higher I strive.

The value I'll use shall be four.

Well played

Or, the usual and shorter $150M.

At the very least keep it to multiples of three, as in 150e6 or 150 * 10^6! Though a very simple 150 million is easy to think of in terms of erasing past donations that sounded impressive.

I don't think that they mean , one hundred and fifty times ten to the sixth factorial

     _|_    __  ___   ___      ___    ___    ___      ___    ___    ___      ___    ___  
    / |_)  /  )| __) / _ \    / _ \  / _ \  / _ \    / _ \  / _ \  / _ \    / _ \  / _ \ 
    \_| \   )( |__ \( (_) )()( (_) )( (_) )( (_) )()( (_) )( (_) )( (_) )  ( (_) )( (_) )
    (_|_/  (__)(___/ \___/ /  \___/  \___/  \___/ /  \___/  \___/  \___/ () \___/  \___/ 
      |

The place where I work, regret a decision to use Oracle, our application build with Oracle Form Builder, which is awfully hard to use, broken easily. But I must admit, their pre-sales really good at describing their product, my boss really hooked up by them. "Oh, for that problem we have this, it will cost this much, but for now, you can just use it for free" then some wild invoice came to our office.

The dept I used to work at was mostly ex-B4 people and I later did enterprise consulting later: What happens is that sales people talk to people high enough up on the food chain that they get the run of the place, and so it's nearly impossible to kick them out or refuse their requests without a substantial political cost/justification. It's called "building a beachhead" and involves the engagement team worming their way up to a client's CTO or even CEO, if possible. I've been deployed as a client-facing "technical"# consultant, and I got really good at it (if I do say so myself): creating additional staff roles, developing opportunities and frenimizing other consulting shops... without inventing things that didn't need doing or creating bullsh!t jobs.

# My mind was more on how to grow the business while simultaneously helping the client fix their deployments. Seriously, the sales guy was dead weight and dumb as a bag of broken rocks but looked good in suit and could talk in nonstop sports metaphors and bizease clichés.

The measure of how a software company values its customers is in the post-sales support... just asking some users how it's going is better than hiding out or DKing them. (About '03/'04, Dell took us out to see the server factory in Round Rock, but went radio silent after a significant purchase.)

I was in a situation where they were selling something to a business leader that my side of the org wasn't in favor of buying (5x cost difference). I drew the short straw and got to meet with them and the business leadership people.

In the meeting, I literally had this dude and his presales guys in $5000 suits sitting across the table from me, berating our position and demanding (quote was "You have no fucking business saying no until you provide us with <our numbers> so we can do your ROI") all sort of crap from us. Ended up prodding the salesbot and let him run his mouth for awhile until he said something really stupid that ostracized the business folks on our side.

Wow that's wildly aggressive. I know people say 'push for the sale' but that seems counterproductive. It seems like they would have got further if they listened more.

Dell Hell.

iSCSI purchase that was promised to be supported on CentOS (it wasn't: RHEL only, despite no deltas), and which Dell itself didn't understand. Ended up getting RHEL just to get a comparison baseline install.

At one point, got cussed out by Dell's support manager in the process (the front-line support team was good). The quote was "I'm not here to support you." Ultimately cost me my job (though we did get the product running).

I'm usually pretty free with sharing my documentation, but in this case made an exception: Dell's support was so fucking crap I refused to provide any assistance for them at all.

Dell's academic-facing sales force (sales and support are their core business value) seemed to start sucking around 2007+. For enterprise IT critical boxes, I would today probably shop around IBM, HP, Cisco, Dell but for web I would go the ODM/OEM route if it were a big enough order. For test lab stuff, Unixsurplus FTW.

(Dell tried to hire me thrice because I could actually use Visio to communicate visually and detail drilling quotes (BOMs) to ferret out upselling and ensure we had supportable configurations (SANs and such). I even had the pleasure to due-diligence several of their solution presentations across the table from 6 of their SE's in Round Rock with the worst hangover of my life, and we still managed to extract maximum intel from that meeting. Dell's client reception secretaries helpfully stocked boxes-upon-boxes of Advil and Tylenol samples and a righteous Keurig binge didn't hurt either. Interestingly, Dell+Redhat had a massive champagne orgy in an adjoining room and we wondered what sort of treatment the several megabucks USD customer level receives, hopefully not just a backpack or stay at the Dell cobranded hotel.)

> The client must hire their own project managers, have clear accountability/authority paths to their management and know exactly what they need (avoid endless upselling).

This so much. Especially the "know exactly what they need" side of things. I speak from the consultant side of the fence, and we've had a number of projects where project management was lacking on the client side; it ends poorly.

We always have our own project manager on projects, and we strongly encourage the client to have their own as well. It really helps to maintain a clear escalation path, and makes everything run much more smoothly.

Large enterprise account engagement is trickier, than say a solo consultant contract with a startup, because there are way more expectations to manage and moving parts to coordinate. As a client-facing consultant, you're often having to balance multiple masters of client and internal politics without having bleed-through and maintain a unified position regardless of other people's screw-ups, accidents or outright lies.

Reply on news.ycombinator.com