Media Breaks: What Retyping That PDF Really Costs

The symptom

Every business has them and nobody calls them that: the points where information falls out of one system and gets typed back into another. The order arrives as a PDF attachment and is transferred into the ERP. The supplier invoice is printed, signed off, retyped. Timesheets come off the shop floor on paper and are entered into payroll. The delivery note is photographed and the quantity booked by hand.

Ask why it works that way and you nearly always get the same answer: "It's only a few minutes." True. Which is exactly why nobody has ever costed it.

The mechanism

A media break costs on three levels, and only the first is obvious.

Time. Four minutes per transaction, at 60 transactions a day, is four hours. Half a position, spread across several people so that it appears in no capacity calculation. It only becomes visible when someone is off sick.

Error rate. People who transfer numbers by hand mistype some of them, however careful and experienced they are. For this calculation, as in the example below, assume that at least one field arrives wrong on 0.9 per cent of lines. At 60 transactions a day, each with one line, that's two to three errors a week on the arithmetic. Most get caught. The ones that don't are the expensive ones.

Delay. This is the item nobody has on the invoice. Retyping doesn't happen immediately, it happens in batches: evenings, mornings, Fridays. Between the arrival of the information and its availability in the system lie hours or days. During that window, somebody else makes a decision without it.

The three levels reinforce each other. People retyping under time pressure make more errors. Correcting errors takes more time. And because both are unpleasant, work gets batched rather than done immediately, which increases the delay.

A side effect I see regularly: media breaks breed secondary systems. Since the transfer is manual anyway, "while we're at it" a spreadsheet gets maintained alongside, holding extra information for which the ERP has no field. That spreadsheet becomes the real working basis, and the media break has produced a shadow ERP.

The cost

An illustrative example with invented figures. Take a wholesaler of industrial supplies that receives a substantial share of customer orders as PDFs or scanned faxes. Two internal sales staff transfer those orders into the ERP, averaging 85 line items a day.

The business keeps a record for six weeks. In the example, pure data entry time comes to around 3.2 hours daily. The rest is more interesting: on 0.9 per cent of entered lines, at least one field deviates from the source, mostly quantity or article number. With 85 lines a day over 250 working days, that comes to roughly 190 incorrect lines a year.

The example breaks the downstream cost of those 190 lines down like this: 61 lead to incorrect deliveries with collection and reshipment, averaging €84 per case. 44 are caught before dispatch and corrected internally, at a cost but without external visibility. The rest splits between price differences and credit notes. In total that's around €21,000 a year in direct cost. Of that, €5,124 comes from the 61 wrong deliveries, which leaves roughly €15,900 across the other 129 lines. On top come two customer complaints that lead into framework contract renegotiations.

Data entry time itself, valued at labour cost, runs to about €29,000 annually. So the error share costs nearly as much as the work.

The fix

The obvious solution would be to ask customers to order electronically. In the example that works only with the largest. For the rest, the break isn't eliminated but bridged.

Incoming order PDFs are read automatically and presented in the ERP as a proposal. The employee no longer types; she checks and confirms. That sounds like a small difference and is a large one: checking is faster than entering, and the error type changes. Instead of typos you get recognition errors, and those are conspicuous rather than plausible.

Two plausibility rules come on top: quantity outside the historical range for that customer and article, and article numbers with no ordering history for that customer. Either triggers a query before release. Automating order entry from PDF and email walks through such a flow step by step, and what it needs first.

The example assumes that entry time then falls to around 50 minutes daily and the error rate on entered lines drops below 0.2 per cent. The freed capacity goes into active customer contact, not into headcount reduction.

The point is the sequence: find and measure the break first, then decide whether to eliminate it or bridge it. That is precisely what a process analysis delivers. Fixed scope, fixed duration, fixed price.

The next step

Count, over a single day, how often numbers get typed from one screen or sheet of paper into another system in your business. That figure alone surprises most managing directors more than any analysis.