So, I've had a tendency to roll my eyes at the "agile" movement, and how much faith people put in it that it must be the best thing since slice bread. I've frequently said that a lot of it is just common sense, some of it is just needless process that just happens to be different, not worse or better, than traditional processes, and there are a couple of good points that agile methodologies have brought about.
But then I look at something like this, and I realize that common sense isn't necessarily so common in a big bureaucracy like this. They have tried three times to automate this, and failed each time? They've spent 7 years on one attempt, with absolutely nothing to show for it? I can't imaging going seven years without some kind of incremental release that you can use to gauge whether it's working or not.
I mean, why do they seem to think that it's necessary to solve the whole problem in one big, opaque project? You would think they could do it incrementally, automating the easiest pieces or automating one step of the process while leaving a fallback for manual processing for the more complex cases or the other steps.
The counterintuitive thing that many people don't understand about government software projects is that failure is the norm. There are relatively few "big" consultancies that win these projects, and their numbers include Oracle, IBM, Accenture, Computer Sciences Corp, SAIC, Booz-Allen Hamilton, and a few others.
Their goal in government procurements is not to deliver working software, it's just to win the contract, and then staff it with as many people as they can. We've known since The Mythical Man-Month and other books such as Peopleware that high staffing levels does not help most software projects, in fact it hurts them. But if the project fails, which most do, they all still get paid. And they will still win other contracts. Delivery of working software is just not high on the priority list--winning contracts and staffing them is.
There are all the usual problems with big software projects, and add to it that most government procurement is still based on the experience of building physical things (jet fighters, ships, highways, etc.) which really doesn't work for software.
But the big reason governments spend millions of dollars and waste years of time on software projects is because the contractors are not incented to deliver working software. In fact their incentives are to deliver failed projects, so that they can bid on them again.
Even crazier is that the government Often creates these situations by requiring that bidders bid the same number of positions, and then picks the one with the lowest rates. Lowering cost with fewer, better people just does not compute to them at all, because they have tried that strategy and been stuck with fewer just as bad people in the past. They then just use these people as staffing resources and attempt to manage the delivery themselves.
The major companies in the space (including many you listed) are trying to push incremental, iterative delivery, but the customers won't get out of their own way long enough to let it work.
I work in a state government and personally failed at least twice to consolidate agency email systems -- a slam dunk if ever there was one. We made progress... Eventually getting about 55% of the target. It went nowhere until someone had the balls to consolidate the staff along with the email.
The issue here is the same. The people examining these records need to figure out where everyone worked, who was a veteran, who is entitled to X or disqualified for Y. All of this data requires records kept by another agency, and the folks between the retirement people and the records don't really give a shit.
The difference between the private sector and government is that the people "in charge" can't change the rules. Think:
Does it make sense for people processing retirement paperwork to work in a CAVE a few hundred miles from that Capitol? No. But congress said put it there. Does it make sense to have a system so complex that you need thousands of people to figure out what to pay a retiree? No.
But at the end of the day, the political appointees come in and say "Wow, the old crew was a real bunch of muppets!" Then they try to fix a people and organizational problem with computers and consultants. Then it happens again.
> I can't imaging going seven years without some kind of incremental release that you can use to gauge whether it's working or not.
That's because you don't work for the government. For the government spending millions and years on something and have nothing or worse to show for it is completely normal. It can not go bankrupt, it can not (excluding very rare cases) be destroyed by the competition, it can only drag along. Surely, people at the top can change, but the bureaucracy only expands, and changes very little whoever is at the top. It requires very determined idealist, powerful skills and huge amount of luck to change anything there. Failure to change is routine and normal.
No, look: this is not about lack of common sense. The people in this cave -- the Office of Personnel Management or whatever -- know very well that the system they are a part of is inefficient as hell. But they also know that, if the system was automated with modern day standards, their office of 600+ people would be cut down to less than 50. Can you imagine having to lay off all those people?
In other words, resisting and rejecting modernization in this particular case demonstrates the existence of common sense, not lack of it.
It's easy to blame government bureaucracy but the real problem is the combination of lack of empowerment and policy which discourages hiring qualified staff in-house. Most agencies have been facing budget cuts for awhile and there has been congressional pressure against hiring staff because hiring contractors allows you to campaign on having shrunk the government, you can receive donations from contractors, and few reporters will actually do the homework needed to note that "saved" positions tend to cost more once you factor in overhead.
Even if you do get permission to hire techies, the GS pay scale limits who you're going to be able to hire. There aren't all that many jobs at the top end of the scale and even fewer will be non-managerial spots which means that your sales-pitch for putting up with bureaucracy and political meddling starts with a significant pay cut for high demand professions. (This is the same reason why regulatory capture is such a problem – we'd be much better off giving the SEC staff 400% raises than having them make decisions about likely future employers when they get tired of below-market pay)
If you do manage to make it that far, you can do some interesting things. I know the CFPB runs a far more modern IT shop because they were able to start without years of legacy-blinkered culture and hire skilled staff directly rather than relying on contractors through a procurement process which basically legally binds you to the least functional waterfall process. It'd be really good for the entire government if that became the model in the future – hire good people, give them autonomy to get projects done and responsibility to deliver.
"“I used to chase people for months — literally — for one signature on one piece of paper. You want to talk about an egregious waste of taxpayer money?” recalled one worker who left the mine recently and declined to be named because of fears of retribution."
And of course if they didn't chase down every signature, we'd be reading an otherwise identical article from WaPo whining about "wastefraudandabuse".
Exactly. I work in a government town (although not for the government), and the fear of "gotcha" is the single biggest inhibiting factor for getting things done. There are hordes of politically-motivated eyes waiting for the smallest chink in your armour, so they can blow it into a huge story. It literally doesn't matter what people try to do - if there are checks and balances it's "waste"; if they try to do something innovative and interesting it's "recklessness" and so on.
The people have the government departments they deserve.
Yeah, I second this, I've learned to phrase questions in such a way when dealing with government employees, where they can give approval, without having to either say yes or no - in other words phrase it in such a way that they dont have to say no - No with the government is the default answer.
> The root of the problem, he said, was that the system had trouble synthesizing information from so many sources and calculations based on so many laws.
So the gist of the article is that we pay these folks to do things with paper because the computer based replacements aren't flexible enough to handle the problem.
I can certainly imagine how that went. A monstrous Java Enterprise deployment, with thousands of classes and many more interfaces, and illustrous names like IRetireeAnuniatingSpecialCaseBenefitCalculatorSourceSinkBean.
Maybe the whole problem is too complex to be able to solve it in one go, but thats why we abstract complexity away and return to solving the simple things, such as reliable digital document storage. Just digitize papers as they come in or are retrieved from "cold storage", and over time you have enough in the working set to get most of these poor people out of the mine and into some sunlight, jesus...
This is exactly why many areas of government should be privatized. With zero competition and virtually no accountability, these bureaucracies have no incentive to become more efficient.
If you read the article again, notice how the fast majority of those millions of dollars have gone to private companies who failed to deliver working systems. Each of those companies passed multiple rounds of competitive bidding, too. The government staff processing things by hand can at least point to a ton of successfully processed requests rather than a complete write-off.
Outside of shallow libertarian magical thinking, privatization isn't a panacea. Anyone who has worked for or with a large corporation has stories which are at least this bad — the main difference being that you rarely about about the failures unless it leads to a major lawsuit. The usual argument is that you aren't required to do business with those companies but … just try to live without insurance, telecommunications, cars, etc. Markets allow you to select the least-bad option but it often takes regulation or changing the problem to make significant improvements. Before someone mentions Google: consider how happy you'd be if their QA-in-production / no live support mentality was applied to your retirement checks rather than your free email…
There are two real problems here: the first is the well-known fact that we're still figuring out how to deal with large-scale IT projects, which affects everyone, and the fact that government agencies are significantly blocked from doing things the right way. For the last couple decades, politicians have “shrunk” government by giving jobs to the private sector rather than hiring lazy government workers. What this actually means is that instead of hiring the same person directly they're hired at a significant markup – equal or greater pay plus the overhead costs of all of the contracting company's staff and profits – and all of the work is pushed into a procurement process which forces you into the worst possible form of waterfall.
You're not the first person to notice that things aren't efficient – and I'm certain that staff at OPM would agree, given how many of them were quoted saying just that. The solution is easy – allow them to hire good people in house at market rates and give them the needed support / autonomy – but that's politically dicey. This and your proposed solution are both likely to fail as long as Congress has a sizable contingent of people who benefit from problems, even those created by their policies, and everyone receives donations from companies which do business with the government. This also won't magically go away with privatization, as anyone who has seen large deals go to the company which bought the best lunch can tell you, but at least in the case of the government the details are public knowledge.
> If you read the article again, notice how the fast majority of those millions of dollars have gone to private companies who failed to deliver working systems. Each of those companies passed multiple rounds of competitive bidding, too.
Those companies didn't fail of course, they were wildly successful at raking in millions of dollars. It's the government that's responsible for spending that money well, and it's the government that's failing at it.
> The solution is easy... but that's politically dicey.
Which is another way of saying the solution is not easy.
> Those companies didn't fail of course, they were wildly successful at raking in millions of dollars. It's the government that's responsible for spending that money well, and it's the government that's failing at it.
This is true but misses the point that the government really needs to develop in house talent. If you looked at every one of those contracts, I'm sure you'd find that part of the work to be performed was project management, design, etc. but the problem is frequently that the government isn't allowed to hire people who can evaluate the options. It's quite easy to make a project fail by giving the client exactly what they asked for…
> > The solution is easy... but that's politically dicey.
> Which is another way of saying the solution is not easy.
True – we're basically debating semantics but I would argue that the solution to this problem is easy but that's because the real problem is something else. As a country we tend to waste a lot of time talking about problems which are the product of other problems we've chosen not to deal with.
> Those companies didn't fail of course, they were wildly successful at raking in millions of dollars.
Yes. These contracts essentially amount to paying a company to work on a problem. There's no incentive for the contractor to finish. When they finish, the money stops. The most profitable activity is to drag the project out as long as possible until the government kills it.
The client needs to actively manage the projects for risk so that the do succeed and don't pretend to "succeed." In other words, doing a good job making sure a good job is being actually done.
Correct – my argument is that the government is actively hindered in having the ability to perform that task. You simply can't contract out the skills needed to manage a non-trivial project because too much of it requires local knowledge and experience.
I disagree based on my experience with equally bureaucratic and wasteful private organizations. My take is that we simply don't have the right organizational frameworks to undertake large, ongoing tasks involving lots of people.
I don't know what the answer is, but watching the results of privatization binges in a number of countries in the past, I don't think it involves moving things along the public/private axis. There are more ways to motivate people than just competition.
Just because you shift the organization from public to private does not make the problem goes away. The true problem is incentive.
For example, private contractors build the rockets for NASA. But there is no incentive to do it cheaply and right. Any cost overrun is covered by the government.
This is in contrast to the COTS program initated by NASA which is oriented toward rewarding space contractors on efficiency and milestones reached. They don't cover cost overrun and responsibility lies with the suppliers. For example, if SpaceX or Orbital Science's cargo ship blows up, they're responsible.
The US Government has a conflict of interest problem with private contractors in that they tend to be the next stop of government workers. If you actually want competition[1] at the Federal level than this really needs to stop.
I still say it should be illegal for government employees or elected officials to take jobs with a company holding or bidding on Federal contracts for a period of 5 years after leaving government service. The only exception I can see is military personnel that have not served on a procurement job.
1) competition for contracts at this time is a joke
I would go further: at least double the salaries of the government staffers who approve or oversee non-trivial contracts on the condition that they have a long ban like what you described on receiving any form of compensation from that company, its officers, or the employees which they dealt with.
When you consider the cost overruns on, say, a single massive project it's much cheaper to pay people well than having a routine cycle to make up for low civil wages.
Yes. They're well compensated compared to the average office worker but not compared to private sector managers overseeing equivalent amounts of money. Someone making low six figures in the DC area is doing well but hardly living lavishly enough not to be tempted away by an offer of significant raise in return for selecting someone's bid.
I'd pair a future employment restriction with enough money to make the job interesting to good people who otherwise aren't going to be willing to put up with the higher scrutiny of working in the public sector.
If you're the GS-14 overseeing a contract with hundreds of contractors on it, it's unlikely you're using enough benefits to make up for the fact that your compensation maxes out at under $140K year in DC, particularly since your peers tend to be working for major corporations which have decent benefits. It's not like you're comparing the federal benefits with a bare-bones Wal*Mart plan.
"maxes out at under $140K year in DC" - you're missing the executive pay schedule (same site), critical position pay, special rates, performance awards, days of leave, and their health plan is much better. Also, the security of a government job with all of the stops to keep you from being fired.
> you're missing the executive pay schedule (same site)
I actually linked to that.
The benefits are a little better but not as much as you're making out. Nobody gets European-style leave allotments; health care is nice but again not noteworthy compared to how executives are compensated and not a benefit most people fully utilize. An extra week of vacation per year doesn't close the pay scale for an IT person, much less a managerial position in charge of similar numbers of people.
> Also, the security of a government job with all of the stops to keep you from being fired.
That is a perk but most of which does not apply to executive positions and, again, I have never noted corporations to be much faster to let people go.
The major mistake you're making, however, is assuming that the overseeing major contracts are all in SL positions. Yes, there's probably general in charge of developing a new fighter plane but an awful lot of projects are significantly run by people at the upper end of the GS pay scale who are not receiving anything noteworthy.
If anything this shows there is a ton of low-lying fruit for the Feds to pick to improve operational efficiency. This along with the Pentagon's inability to do accounting should allow for big improvements. And due to budget constraints and more government services coming online (Healthcare) there might actual be a reason to push through some of these changes. Or at least that is my hope.
My favorite example of just how dysfunctional it all is was an agency which made a small fuss about how they were modernizing by allowing online job applications 10+ years ago.
Behind the scenes the applications were simply emailed to a shared mailbox then printed out and re-entered by hand into the actual position management system.
The position management system was so dated and overloaded that the folks entering data had learned how long they could expect to wait (measured in minutes to tens of minutes) when moving between a given set of screens. They would submit one screen then work on something else.
Eventually, the agency spent an enormous amount of money implementing some combination of systems from SAP to consolidate / replace everything.
Now the same folks who were retyping resumes are cut and pasting them into a series of webforms and emailing around spreadsheets of all the information that the system lacks forms for.
Hiring managers who had little or no input on requirements for new system don't even log in, they request printouts from HR.
---
Having seen things like this play out a few times now I have a few ideas about what could help one of these projects actually succeed:
* Generate or at least inform requirements with honest information from people at the ground level. Those people toiling away filling up manilla folders and doing data entry know the actual process. The farther up the chain you go the story will move away from reality toward what's effectively nothing more than the way things are supposed to work.
Problem is, the layers of management don't want to allow that kind of access and will retaliate against their subordinates for being honest. Also, those low-level people are likely aware enough to know that they're talking themselves out of a job.
I think the former requires a powerful and attentive enough personality leading the effort to simply bulldoze the managers. I wonder if the latter can't be addressed by buying everyone out with generous severance or early retirement. It would puts a cap on the cost of these broken jobs and processes rather than continuing to cycle people ad money through them.
* As a rule, the technical and procurement staff of these agencies aren't competent to generate requirements or evaluate what's produced when dealing with outside contractors and implementers. At the same time, the agencies don't pay enough or move fast enough to maintain a competent internal staff.
This is really hard. Hiring outside experts seems to actually work against the goal as they will invariably work more for some particular vendor than the agency.
Big companies absolutely collude and actively work to hold down smaller competitors so "competitive" anything doesn't really work.
One idea I've been kicking around is a sort of a open source template for government system/project requirements. It would be a easy to follow "test suite" which proposals could be compared against. It would raise flags or fail plans and terms which don't meet the standard of contributors with domain knowledge.
I'd like to write more on this, but I've got a few things to handle this evening.
The 2010 US Census was a good example of this sort of dysfunctionality (I worked in one of the offices for it).
Hiring anyone for field jobs was a process something like:
* field operations clerks arrange application/testing sessions and get all applications in on paper
* administrative clerks (other side of the same office) enter all applications into the computer
* when the people running the field operations want to hire field employees, they fill out paper forms with manually selected criteria (local area, languages spoken, etc) and give them to the administrative fiefdom
* administrative clerks enter the form criteria into the computers, print out a paper list of matching applicants
* administrative clerks call down the list until they have enough who have said 'yes' to the job to match the form
* administrative clerks hand the paper lists with selected applicants on them over to the field operations fiefdom
* field operations clerks try to organize the data in an actually usable way, usually by manually entering the lists into Excel or Access
The most bizarre part was the "manual hiring forms" thing. Beyond the cargo-cult hiring practice of the forms literally only existing so that they could be filed away somewhere national in case of cover-your-ass legal issues, the Census didn't actually give us access to any resources for demographics... and so the only reason my area had anything like representative employees (important for getting into gated communities, dealing with different languages, etc) was because we had a set of people spend a week just comparing paper county tract maps supplied by the feds with locally-supplied city and county demographic info and writing up dozens of forms with criteria tuned for specific areas.
As a minimum, I think they could scan papers into a document system. We had this when I was in insurance. It still required review by a person but most files were digital.
This basic thing already exists. I don't understand why it is not being used for the federal retirement system. Insurance is probably just as bad as the retirement system: Policies have to comply with state laws in fifty different states and there are all kinds of SNAFU things that go on because of it.
These organizations seem almost organic, an emergent phenomenon that protects it's own existence. How does one kill a bureaucracy? It's times like these I'm tempted to throw my support behind the "Eric Schmidt for CEO of America" stunt.
The article actually hints at how the problem could be solved: Simplify the rules (read: laws) that the system has to implement sufficiently that switching over to a computerized system becomes simple enough. However, such efforts are boring and don't score political points when they're meant as genuine help.
The other point to note is that they are already partly computerized, as the article admits. Part of the problem may be that retirement systems have strong backward compatibility requirements by their nature. So this is actually a problem that might fix itself given another decade or two, as more people's files become fully electronic.
It seems to me that (as often) announcing a big project to try to fix it all in one go is just bound to fail. Instead, the best way forward would probably be a small number of competent people who are given the freedom to refactor the process from within. A good starting point seems to be to reduce the number of computer -> paper -> computer turnarounds that seem to exist with the current system.
It's a broken feedback loop. The people who feel the pain are hardly any more connected to Congress than I am; how can information about the pain get back to the body that can actually do something about it? In a nutshell, it can't. Further, I as a taxpayer am paying for this, so technically I'm in pain too, but how can I close this feedback loop? I can't.
To those who are saying here "Privatizing isn't a miracle solution", since, ahem, you are trying to preempt the libertarian argument since you sense that it's a rather strong one here, well, of course it's not a miracle cure, but in general in the private world you can't get this level of brokenness in the feedback loop. Somebody would have had to do something decades ago because this would have gone out of business, or some equivalent. Private industry's advantage here isn't that it could somehow fix this problem, it is that it has the "advantage" of being unable to sustain this level of pathology in the first place. Before you post your example of equal pathology in the private field, my question to you would be, has that pathology been present for 40 years? Some private entities do indeed become healthy enough and bank enough funds to be able to descend into this level of pathology, but they can't do it indefinitely; eventually they must either clean up their act (IBM) or go out of business.
And when we talk about "big government", it isn't always literal size... sometimes it's things like an enormous maze of interlocking regulations that nobody ever asked "Does the costs of this regulation exceed the benefit when combined with our other enormous pile of regulations?" Obviously, Federal government workers are going to get their retirement benefits, and growing the employment of the mine isn't going to particularly cause sudden new regulatory overreach or anything. The literal physical human footprint is not intrinsically a problem. But this is "big government" at its worst, a regulatory footprint so large it half paralyzes itself... is there really anybody who would argue in favor of this? (We can't even particularly argue that we are better off with this particular department paralyzed on some weird libertarian "better a paralyzed bureaucracy than an efficient one" freedom argument... who benefits from this accounting being this slow?)
Bureaucracy is a wonderful thing, an amazing achievement for humanity. Compare a nation with a functioning bureaucracy vs. a nation that doesn't. The paperwork chain breaks corruption, breaks broken incentives, localizes the impact of cheating and makes it accountable. If it's implemented well.
Bureaucracy doesn't have to be a dirty word, although we're all disgusted by the waste it can cause. Especially I see in the U.S. where government systems tend to be waayy worse than I'd experienced working within the government in Canada.
Even banks have a different internal character between the US and Canada; at least the one Canadian bank I've worked in simply did a much better job of having a functioning, appropriate systemized bureaucratic means of operation than the inane politicking-slowpoke-dysfunctional bureaucracies of some of their competitors.
A few of my friends at Google have complained about bureaucracy there, so shrug I'm not sure Eric Schmidt would exactly be our saviour here.
My feeling, from limited experience, is that it is easier to bring/hire people into an existing/new process than it is to change an existing process people are familiar with.
"Risk management: OPM did not have a process for identifying and tracking project risks and mitigation strategies on a regular basis. Thus, it lacked a mechanism to address potential problems that could adversely impact the cost, schedule, and quality of the modernization effort.
Organizational change management: OPM had not adequately prepared its staff for changes to job responsibilities resulting from the modernization by developing a detailed transition plan. This could lead to confusion about roles and responsibilities and hinder effective system implementation.
Testing: The results of tests 1 month prior to the deployment of a major system component revealed that it had not performed as intended. These defects, along with a compressed testing schedule, increased the risk that the system would not work as intended upon deployment."
'Work in Progress' is a key form of muda or waste in lean philosophy. The key is to eliminate waste coming into the system. Employees need to be empowered to, send straight back incomplete paperwork and to update the forms & instructions coming into the system that are contributing to the waste.
Considering the gravitational mass of all these processes I would work on bringing lean manufacturing approaches in first before deriving a computer based solution.
As per all these things its going to take a lot of leadership to drive any type of real change.
Oh, well, let's not be too hasty in becoming efficient and up-to-date here. Obviously, having a sneaker-net is fast enough for you simply because the name has the term "net" in it. One of my old employers has a few vintage DEC VAXes if you want to start out on saving on your pencil budget.
I can confirm, though that the "chase the physical signature" game is still alive and well. /shiver
I would, however, like to be the pizza guy with a security clearance. Employment while it lasts :D
I wrote an article about this having lived through it for a shilled called Why a hammer costs $5000? And it really all comes down to motivations being very badly lined up trying versus just doing the same. The link is: http://www.followsteph.com/2011/11/22/why-does-a-hammer-cost...
A 61-day processing period is not cheap for the person waiting for money.
You want to digitize old forms for retired people? This is entirely the wrong approach.
New data should be in digital form. Not even all new data, just one vertical at a time - new payroll data, new business expense data, new insurance claims data, etc. This shards the work into tractable, testable, measurable units. There will be a (long) period when processing retirement forms will need to pull data from both analog and digital sources. That time will eventually pass, and the analog gradually phased out.
Are you aware of time periods involved, though? When you say "new data" do you understand that this new data will only become pertinent in about 50 years, when the employee that just came in to system, retires. 50 years later. And meanwhile, you are processing paperwork for people that entered the system 50 years ago: on today, 1964.
(Ok, maybe closer 40 years, but the point is the very long storage time.) Even if all new data is digital today, what about 50 years later, how easy will it be to use digital files that were generated 50 years earlier. What if file formats change every 10 years. Then 50 years on, you got 5 different file formats you must be able to work with.
Investing 100's of millions into a parallel system that won't yield much until a good portion of your future retirees are in it (and all 50 diferrent federal agencies records are integrated in) is a hard sell in any universe.
In the meantime, that system will obsolesce in 5 years and you'll have to redo it all again.
What was their accuracy rate? What's your process for not being sued when someone's retirement is calculated assuming they made $10,000 when they retired instead of $70,000 because your OCR glitched?
Google Books is a wonderful service but few services meet the level of importance and liability which a retirement system will have.
Wait until those glitches show up on 60 Minutes and Congressmen are ranting about another digital disaster...
There's no win here when you're dealing with ancient records and dozens of agencies with different standards.
It makes me laugh when people see this as an easy project and don't grok the politics and all. Decades back, I saw a "by hand" project that generated >$100M in tax revenues from corporate non-compliance in one test region, and the next year, Congress cut/delayed an automated expansion because the biz folks whined about getting caught.
But then I look at something like this, and I realize that common sense isn't necessarily so common in a big bureaucracy like this. They have tried three times to automate this, and failed each time? They've spent 7 years on one attempt, with absolutely nothing to show for it? I can't imaging going seven years without some kind of incremental release that you can use to gauge whether it's working or not.
I mean, why do they seem to think that it's necessary to solve the whole problem in one big, opaque project? You would think they could do it incrementally, automating the easiest pieces or automating one step of the process while leaving a fallback for manual processing for the more complex cases or the other steps.