Here's the state of the art for RFID tagging: "At a case-by-case basis, as I go down through our process, I validate that case. A tag is on it. I encode the tag and then I immediately validate that, and I capture those statistics. If it doesn't read, I can pull that out and put another tag on until we get it read."

That's what Ray Hagedorn, an IT executive at Sara Lee (maker of cheesecake, clothes and other things), told Computerworld recently. And it's not just that the tags are unreliable, Hagedorn says. The middleware is half-baked, the consultants aren't much help, and a real return on investment is nowhere in sight.

See? It's not just you. Everybody's RFID project is like that. This, despite there being plenty of guidance on planning.

The technology really doesn't work the way it should. You really can't point an RFID reader at a pallet and be sure of getting a response from the tags on each carton. All those dreams of RFID visionaries may come true someday, but not in time for your project. RFID just doesn't work yet. This stuff really isn't ready [which may not be much comfort to parents visiting Legoland, which uses it to track kids - Editor].

Rise to the challenge
So if you're facing an RFID mandate from Wal-Mart or the US Defense Department or some other big customer, that means you have to make a choice.

You can look at your RFID project either as a death march - or as an adventure.

This is no ordinary IT project. One in five RFID tags dies. The signals are blocked by metal in products, by shiny packaging, even by thick liquids. If you expect RFID to work as advertised, your project will be a long, horrible waking nightmare. And it will fail. And everyone connected with it will want to kill you for putting them through it.

But you've still got to meet that mandate. What can you do? You can forget about this as a conventional IT project and treat it like what it is: an expedition into the unknown.

So don't pick a team of perfectionists. Pick spit-and-bailing-wire tinkerers and shade-tree mechanics. Warn them in advance that the hardware is unreliable and that the problems are intermittent. The ones for whom failure isn't an option will run away screaming.

The ones who are left will see it as a challenge. Failure will be their daily companion - right up until it turns into success.

You know these people. For them, this won't be an endless nightmare. It'll be a challenge.

Smart workarounds beat the glitches
The tags don't work reliably? These people will test 'em as they go and keep testing until they can sniff out the bad ones early. They'll add checksums and redundant data to the tags' contents and put the smarts to verify and do error correction into the software. It may not be a pretty solution, but they'll get it working.

The signals can't make it all the way through a palletload of cartons? They'll work around it, maybe by scanning each carton as it's added to the pallet, then linking all that RFID information in the database. That way, once that pallet is wrapped and strapped, if they can successfully scan even one tag, they can identify everything on the skid.

The middleware isn't up to the job? The database is wrong for the task? Here's where you'll probably have to rein in their urge to kludge. Eventually RFID technology will work, and when that happens, you'll need a database that scales up and middleware that hasn't been hacked to a fare-thee-well.

But for everything else, give your RFID people plenty of room to experiment. Their brainstorming ideas may sound off-the-wall. But let them try. If they discover some oddball trick that works, you've got a competitive advantage. If they don't, well, you already know that a conventional approach to RFID is doomed to grim failure.

You don't need another death march, not now. Don't let your RFID project turn into one.

Make it an adventure.