My Dream App

[da]

[edit: unlocked b/c round one is over and I didn't make it; maybe someone will take the ideas and make this anyhow?]

Yesterday I entered My Dream App, a contest to propose a new Macintosh application, judged according to its novelty, use of Mac OS features, feasibility, and marketability. Three winners will get to see their applications developed commercially, plus they get royalties. In a week, the contest closes to new submissions, and they weed the bids down to 24 semi-finalists. (Go check it out; I'd love to see what my friends come up with as their ideal applications! Plus, I'd love to bounce ideas off everyone, and help come up with something else as your bid!)

The initial bid is limited to 800 characters (eek!) and they're up to over 1,500 submissions in the first 48 hours (eek!!) So
I don't suppose I'll make it to the second round. But who knows.

People who've read my fuming about Quicken on Mac may guess where my thoughts were this week. So sue me, my dream app is... a Quicken-killer. Yah, boring. How many people would use something like this?:

Title: Tweek or Ka-ching. (Maybe something else. I've got a week to decide.) [Edit: how about 'Reggie' short for Register?.. ]

Description:


A modern money-tracking program. We've got email receipts, paypal, bank and card transaction downloads. Checkbooks are 20th century. Automate!

Default interface presents eye-candy for your chosen important items (budgets, recent transactions, balances).

Use spotlight to find emailed receipts. Attach web receipts and web proofs-of-purchase. Download .qif and OFX data. All automatically, & via task scheduler.

One goal: minimize manual entry. OCR paper receipts via scanner/iSight. Automatically reconcile where possible. Learns your behaviour well enough to make money-tracking effortless.

Another goal: use the network. Open scripting API for plugins (IO/storage/control). What if it worked with billmonk.com? What if joint expenses carried to other person's view on their mac?

---

(Please don't share this beyond my friends-list; I locked it b/c I don't want someone else to submit the same idea.)

Critiques welcome. After all, I have 12 characters to spare. ;)

Comments

[dr-tectonic.livejournal.com]

'Tweek' is a bad name. It doesn't connote finance, and is likely to be confused with either setup-tweaking or crystal meth.

I've never really understood the point of programs like Quicken. I don't have a problem keeping track of how much money I have, or where I spent it; I buy basically everything with a credit card, and I can download my transaction history at will -- ditto with checks. The credit card also buffers money, so my exact account balance at any particular moment isn't ever really important.

The thing that would be really useful would be to have all the purchases categorized, so I can see how much total I've spent on, say, groceries. And from what I can tell, that's something you have to do by hand, on top of entering all the data, and it's a gigantic amount of work, which is why I never do it.

Could this program help with that kind of thing?

[dcseain.livejournal.com]

I agree with dr_tectonic that Tweek is not a good name. Especially since what you're describing is more of a money genie.

[melted-snowball.livejournal.com]

I don't really understand much of money management, either.

Quicken is vaguely helpful for us because we keep quasi-separate finances (we have his, his and ours accounts), and I often pay for things that are household expenses in cash (mostly food: farmers' market visits, ethnic grocery store visits, ...) or via my own credit card.

My biggest issue with all things Quicken-like is that at the end of the day, I more or less know what I already did know, which is what my online bank balance was so kind as to tell me: I have about $X. It always feels like a huge payout for little payoff. (On the other hand, it's vaguely cool when da_lj tells me how much money we spend on things I didn't know we spent money on. And then I forget...)

[ng-nighthawk.livejournal.com]

I would use this because we manage several different accounts which must be merged. One thing to mention (it might make the folks less likely to pick your choice, not sure on the politics of such things) is that we are a Mac/PC household. Interoperability with other programs, or publishing to an editable generic format (.pdf and .html are not terribly editable, if you see what I mean) would be nice. But as I say, I sense there might be some annoyance with that kind of thing.

[da-lj.livejournal.com]

Could this program help with that kind of thing?

"Why yes," he said, with a twirl of his cape and a clap of his hands.

The (outmoded) state of the art is to automatically pair up new transactions from the same vendor, to the same category (or categories plural, by percentage, if you've gotten fancy).

Some credit-cards will do this automatically, on their website, for a small number of categories. That's OK.

Or Quicken will do it, using your own set of categories, when you download the transactions; then you correct any that it mis-categorized.

To my mind, Quicken is buggy in that it doesn't learn the vendors very well; it can get completely confused by any novel characters in the downloaded vendor name. I think software should do much better (just from what I know about Bayesian trained spam-filtering, which does a damn sight better than Quicken).

The reports you can generate, in Quicken, are really quite pretty. You can compare actual expenses to budgets, by month, if you're inclined. Or just notice how much you're spending on dog kibble versus movie rentals...

There's another layer, the receipts. I save them from credit-card transactions, because restaurants occasionally like to enter the totals wrong. And often the category is purchase-dependent; such as personal vs. business vs. joint expenses. Currently, this step sucks. I spend at least 30 seconds on each damn receipt.

I think/hope a modern system could OCR your receipts and grab date, vendor, total amount, maybe keywords. But at least reduce from 30 seconds down to 5.

Maybe OCR isn't that great. But if the postal service and British traffic systems can do sub-second captures on noisy data, I should be able to expect my computer to do 5 second captures on, say, the first and last line of a receipt, and maybe more. Shouldn't I? (He said hopefully).

So, in addition to reconciling the downloaded statement to the downloaded total (which is stupid; I've never had more than a 1-cent difference; but that's the sort of reconciling Quicken does most easily) I want to reconcile my receipts against the downloaded statement. Quicken's way of doing that isn't very intelligent either: you have to enter the receipts before you download transactions, because it only makes one attempt at matching them up, at the download stage.

If you download them first, then for each receipt, you have to either sort the paper receipts by date or manually cursor around the checkbook register to see if they're missing. And fix the categories, which is an annoying amount of mousework (more so than entering the transaction in the first place).

If you don't do this religiously and you have a big stack of receipts, this can be a lot of sorting and cursoring.

This is my experience, at least.

Now, dan just sorts paper receipts by purchase category, and adds them up on a sheet of paper. And eyeballs his bank statements. I'm not totally convinced that's not a better solution in the end.

[da-lj.livejournal.com]

And the name 'Tweek' is now struck from the record. ;)

I'm not sure I like 'Ka-Ching'. At least it's onomatopoeic.

How 'bout 'Reggie' short for 'Register'? :)

[da-lj.livejournal.com]

And the name 'Tweek' is now struck from the record. ;)

I'm not sure I like 'Ka-Ching'. At least it's onomatopoeic.

How 'bout 'Reggie' short for 'Register'? :)

[dcseain.livejournal.com]

Cute, Mac-like, and i ilke it. But will it really mean anything to people 10 or more years younger than we are, that is if were marketed that Reggie is short for Register?

Maybe it can be in the release notes, complete with graphics of old-school, filled-out registers. I could probably dig some up for that from my dirt-poor college days.

[dr-tectonic.livejournal.com]

See, I just throw the reciepts in a box and glance at the credit card bill once a month. But we're also pretty much a single-account household.

I'd never do anything with reciepts, but smart categorization would be awesome. Is that an area where you could harness the power of user contributions? Like, suppose you can export a file (suitably anonymized and sanitized) that listed a bunch of vendors that fell into some useful category, and share it with other people. Sort of like Flickr tagging. Could I have most of my vendor categorization done automatically by relying on categories that other users have published?

[ancawonka.livejournal.com]

This is the kind of idea, where, if they don't accept it into the program, you can get together with three friends, built it over a couple of weeks, and either open source it or sell it to some company (apple, quicken, google, etc)...

I would totally use this kind of thing. Anyone who does budgeting and tracking of detailed expenses or is managing property or runs a small business could use this.

[earthling177.livejournal.com]

I think that of the names mentioned, "Ka-Ching" is definitely the most attractive. If you have to explain why it's named Tweek or Reggie, some part of the population won't even look at it twice on the shelf -- with Ka-Ching you at least get a second look, even people who don't ordinarily use apps like that will at least want to know what it is and will have basically no problem remembering the name to tell their friends who are looking for one.

The OCR is very interesting. Why enter stuff by hand when the computer can do it for you, even if you have to then correct a few characters? And your description of Quicken makes me happy I've never used it -- it is just asinine to make users enter all their transactions before you import the .qif, WTF?, it should prove much easier to download the file and let people point and click to transactions and maybe even drag and drop them in the right places when they find the paper receipts than entering stuff by hand.

I have not looked at Delicious Library in a while, but the guys there used to describe in general terms what they did to read the UPC thru iSight so they can enter the book in the library. I have the app, I like it even though I have not really used it a lot, but I ended up buying a dedicated, hand-held UPC laser scanner that uses Bluetooth because holding books to the iSight gets old fast. On the other hand, books are heavy, receipts are not.

Either way, I love your idea, and I figure once it's out, one mention on the Mac sites should get you plenty of customers, particularly if you use a portable file that can easily be converted to/from Quicken, like ng_nighthawk's example.

Good Luck!

[da-lj.livejournal.com]

"Why yes, it could," he said, with a flourish.

But I was thinking about that before, and people are more funny about money than they are about shared photos or bookmarks. All you have to say is "phones home" and you lose a chunk of interested users.

Even as little shared information as, "anonymous user X shops at store Y and calls that category Z." If the data were stockpiled by some random company, I wouldn't necessarily trust that they weren't collecting my shopping patterns, even without pricetags.

Though, there could be a pre-seeded list of stores, and there could be separate options for automatically doing a lookup, and automatiacally submitting new store/categories. (filter out categories marked private?)

Glad you mentioned Flickr tagging. This gets into something else I didn't have space to write in the short form. Tags: why limit to one category like Quicken does? Categories shouldn't be mutually exclusive; a shipping purchase can be personal, joint, business; it could be for a particular project I wanted to track; or all of the above.

Quicken's model is convoluted: there's a "category" (limited to one, though they can be hierarchical and therefore inherited). You can have a "class" which is sort of a second independent category, but not really. Limited to one.

And you can have a "transfer" which means one side is a credit and the other side is a debit. But a transfer can't be categorized, which means you need to mess around with double-entry accounting and set up the other account to be normalized to zero if you want to categorize those transactions. And then, they turn out to be negative. I've tried going the double-entry route for categorizing joint expenses, and damn but it's complicated and annoying.

Hm, I hadn't realized I would vent so much about Quicken categories. Guess you hit a nerve. ;) I don't know any program that does this right, either.

[da-lj.livejournal.com]

Yes! Check out billmonk: they store your db of owed stuff (money, books, anything you want) and they send you emails based on changes. This program can parse the emails and submit new transactions via new emails to their system. (I believe they have blogged they want to add an API for doing this directly via the web, once they get a bit bigger). I would love to see this program transact via trusted email addresses. So, for non-mac users, they could get emailed transactions directly, or via billmonk (who will track the totals for you).

I had a 4am thought: this program itself could handle non-money items just as well. Make it transparently easy to add new "denominations" and shazam, if your "friends" category includes a transaction with the denomination "beer"... you can mark that you owe Dave a beer.

[The problem with email transactions is authentication; I expect billmonk will run into this as well, people forging the "from" and inserting bad URLs. This program can possibly get around that with SPF to validate the domain, in case your spam-filters don't do so automatically.]

[you probably don't care about that, but I wanted to write it down to be sure I remember myself. ;) ]

[da-lj.livejournal.com]

Yeah. Or, maybe I decide the day before the contest ends that I don't want to leave it for mydreamapp.com to do badly, and go off and do it with friends anyway! ;)

I would totally use this kind of thing.

Cool.

Anyone who does budgeting and tracking of detailed expenses or is managing property or runs a small business could use this.

That's the impression I want to give in the project description- it's for all of us. There's no difference between small-business and household, as far as the programming goes. (Just category names!)

[da-lj.livejournal.com]

Yeah. The OCR might be tough to do properly, especially if it's via low-res like the iSight. It's definitely the toughest part (other than generically deciding which extra features go into the first release, and which should be plugins.) OCR might rely on a dedicated scanner, for the first version (though that sucks too. Open scanner. lay down receipt. close scanner. press button. Lather/rinse/repeat.)

I like Delicious Library a lot, I downloaded it just to play with the iSight scanner thingy. I decided my time was better spent on other things than making a (very pretty) db of all our books. But yeah, it was one of my inspirations when I was thinking about this.

It would totally kick ass if machine-readable receipts were bar-coded so we didn't have to muck around with OCR.

[da-lj.livejournal.com]

I agree about the name, too. ;)

[da-lj.livejournal.com]

On the opportunity for customers:

I'm quite amused to discover just now that Intuit has a blog for Quicken. The most recent entry was Oct. 13, 2005, an announcement of an update to Quicken 2006.

http://quicken.typepad.com/my_weblog/2005/10/new_quicken_200.html#comments

It was taken over by mac users complaining politely about how badly they're being treated.

[dawn-guy.livejournal.com]

Ka-Ching could work to your advantage, given what I know of Apple employees. I'd be inclined to a more "fun in function" name like BuckTrace or Bank Shot, but I'm not steeped in Apple culture.

It'd be nice to have trends (for budgeting) and exceptions (highlight unusual activity), perhaps at the expense of the "Checkbooks are 20th century" verbiage.

[da-lj.livejournal.com]

Intuit might be heading in this direction, with a new web-service they started, called zipingo (apparently mostly for rating and finding vendors, which shows up as ratings within Quicken. which is an odd thing to bundle into a checkbook program, but I digress.) It also seems they are going to allow users to categorize vendors on the site, so if they thought of it, they could possibly pick up categories from vendor within quicken.

http://zipingo.typepad.com/my_weblog/2006/08/scattegories.html#more

I don't know how far-fetched this is, since the zipingo site seems to be down this morning. Anyway.

[psychedelicbike.livejournal.com]

I'm not so sure I like the Ka-Ching name. It seems, um, kind of trite and childish for an application that performs such a "serious" task (no offense intended). What about something that follows the Apple branding conventions, like:

iBank
iBalance
iBudget

Anyway, the app does sound very nice. Good luck with your entry!

[melted-snowball.livejournal.com]

I like iBank.

[da-lj.livejournal.com]

Those are all excellent ideas.

Buck... Bucko.lico.us? (*hork*) (I need a nap).

I dunno much about apple employees, so I'm not sure what advantage I was getting from the name. I do know that the Woz is supposed to be somehow involved with grading the submissions in the contest. (eek!!!)

[da-lj.livejournal.com]

thanks. Ibank would be my first choice if it weren't already taken by a piece of mac software. Hmmm.

http://www.iggsoftware.com/ibank/

I haven't tried theirs out yet, maybe I'll decide theirs is really good enough after all.

[da-lj.livejournal.com]

Well, theirs nicely illustrates how an attractive mac application can be semi-useful. And, what a big job producing a money app actually is, if it's supposed to be compatible with quicken QIF files..

[da-lj.livejournal.com]

So do they. ;)

[psychedelicbike.livejournal.com]

Curses! Foiled again!

iFinance? iFi for short :)

[ng-nighthawk.livejournal.com]

You know, I think this could be done online over SSL, too, just using MySQL and PHP. But then you'd actually have to build it yourself.

However, if you ever decide to pursue this on your own, I'd be interested in contributing effort toward it. I offer expertise in testing, SQL, and PHP, plus I have a low flake factor. I wouldn't say my life's dream is to come up with a secure, customizable personal finance tool, but it'd be an interesting project.

[ng-nighthawk.livejournal.com]

Oops, except for OCR on receipts. That part would require a local app. Never mind me.

[melted-snowball.livejournal.com]

Oh...hm.

[da-lj.livejournal.com]

One of the fascinating things I've discovered is that mac is blurring the lines between web apps and local apps. OS X includes SOL (sqlite) and dashboard widgets are really locally executed javascripts. For example:

{
    if(window.widget)
    {
        document.getElementById("outputText").innerText = 
            (widget.system("/usr/bin/uptime", null).outputString);        
    }
}

That piece from /Developer/Examples/Dashboard/Uptime/Uptime.wdgt/Uptime.js from the Xcode download.

I've browsed Apple's beginner tutorials and if Tiger weren't going to make things so much easier to generate widgets, I think I'd be doing some widget programming in my free time right now.

[da-lj.livejournal.com]

Their software is OK but it oversells. You can't actually import from quicken properly, and you can't compare accounts with different currencies. But it shows promise.

It's actually quite useful for illustrating how this putative project shouldn't be done. ;)

But too bad they got the name first. :)

[ancawonka.livejournal.com]

"I've browsed Apple's beginner tutorials and if Tiger weren't going to make things so much easier to generate widgets, I think I'd be doing some widget programming in my free time right now."

See, that's the problem I have... There are so many fun things to program, but why should I bother if someone was going to do it in a few months anyway? Other than to hang out with cool people and do it. :)

[da-lj.livejournal.com]

Yes, exactly. This is one of the reasons I've got a perlmongers group to hang out with (perl folks are good for this social programming stuff). I just got myself subscribed to a Toronto based mac-programming group, so I'll see if that adds anything to the mix.

Unfortunately, there's this thing called "work" that seems to interfere with a lot of these kinds of projects. ;)

[earthling177.livejournal.com]

OCR might rely on a dedicated scanner, for the first version (though that sucks too. Open scanner. lay down receipt. close scanner. press button. Lather/rinse/repeat.)

Well, yes and no. I guess the part that sucks with the scanner is the overhead (it will take about the same time per linear inch of the flatbed whether you have one or 4 receipts). That part can be solved by allowing multiple receipts to be scanned at once, at least one app (CanonScan) does that already for the picture part, I have not tested the OCR part, but I'm almost sure that the OCR part just takes documents, so the scan part can just break say, 6 receipts into 6 docs and feed them consecutively to the OCR.

Actually, not only that makes scanners more useful than they used to be, if you can now do finances too, but maybe one or more of the scanner manufacturers might be interested in bundling your app with their scanner -- at this point, they are all basically a commodity, selling from 50-150 bucks and desperately in need of anything to make them look different and better -- Canon scanners, for example, advertise they can be used vertically, LOL... while it's true and nice that they are thin enough to be put away between two external hard drives until you need to scan something, thus saving table space, it's very hard to use it on its side, not to mention it makes it impossible to scan stuff that is more than a few pages thick, like books.

[da-lj.livejournal.com]

Getting off the topic a bit... but: have you seen the newer wand scanners? I heard of one a few months ago, but I don't think I wrote down the brand name. That would be wonderful if they worked well..

[earthling177.livejournal.com]

I remember seeing one model, in passing, at my local CompUSA -- I didn't even stop or see what brand it was either, my first thought was "oh, they're trying that one again". I sure hope they work well, or at least better than the previous couple of tries since 1986. The reputation is that it takes a very steady pace to get good results. Granted, now we have enough computing power to embed in them that they might actually be working well. Then again, if you skew the scanner just a bit, the distortion is hard to correct. I guess they are a little desperate, unless one is in a very specialized application, like libraries or journalism, scanners don't have as high an appeal as they once had, now that photography is digital.

PS: Google seems to point to a lot of different brands, like Wizcom InfoScan, IRISPen and Planon System Solutions DocuPen. They all seem like OCR pens, no pictures (not that I expected pictures).

[da-lj.livejournal.com]

hm. here's a mystery: the one I heard of was operated differently than those three: you waved the long side of the wand against the page. I'll see if I can figure out what it was.