So we’re a mid-sized marketing agency and honestly the invoice chaos is real. We’re talking dozens of media vendors, ad networks, random suppliers — and every single one sends documents in a completely different format. PDFs, screenshots, whatever got faxed and then scanned three times. We have zero control over what comes in.
What OCR tools have actually worked for people in this situation? Specifically trying to pull vendor invoice data reliably without having to beg every vendor to change their formatting. Would love to hear what’s actually worked in practice vs. what just sounds good in a demo.
Ugh, been there. The vendor diversity problem is genuinely what makes media agency invoicing so painful compared to other industries. You’re not dealing with one or two invoice layouts — you’ve got broadcasters, digital platforms, production houses, print vendors, all doing their own thing. Template-based OCR just falls apart in that environment.
In my experience, the only thing that actually holds up is AI-powered OCR that doesn’t need templates at all. We use Lido for this — you just upload whatever comes in and it figures out vendor name, invoice number, billing period, amounts, payment terms, all of it. No setup per vendor. That matters a lot when you’re onboarding new suppliers constantly.
The multi-format thing is huge too. We get PDFs, screenshots, scanned docs, stuff that was clearly faxed at some point in its life. Rather than trying to get vendors to standardize (good luck), you just need a platform that handles all of it and spits out consistent output into Sheets or Excel.
A few things specific to media buying that trip up a lot of tools: broadcast invoices with multiple spots at different rates, digital invoices mixing CPM and CPC line items, production invoices that are basically just a wall of text. If the structure varies at all, templates choke. Integration was pretty painless for us — maybe a week or two to get it talking to our accounting system properly. Accuracy has been solid, 95%+ on most stuff.
8 months in and still going strong, nice. We’re about 5 months in with Lido and yeah, not flawless but the time savings are real. I’d say we’re clawing back maybe 10 hours a week which for a team our size is genuinely significant. The manual process before was just… not sustainable.
Same here, honestly. We’ve been on ABBYY for a little over three months and yeah, it’s not flawless — we still catch errors here and there — but compared to what we were doing before (two people manually entering data from PDFs like it’s 2005) it’s night and day. I’d say we’re in that same ballpark, maybe 12-15 hours saved weekly for us specifically. The first few weeks felt rough but once you dial in the templates it gets a lot smoother.
Jumping in here because this is SO true and nobody talks about it enough. We spent weeks tweaking software settings trying to figure out why our accuracy was garbage and it turned out our old scanner was just… bad. Upgraded to something halfway decent and it was like a completely different tool. That said, we’re running at 300 DPI and still hovering around 91-92% on some of the older print ads we process — so I think document condition plays a role too, not just scan quality. But yeah, start with the hardware before you blame the software.