When the details have to be right

“Why do our unit numbers differ from controlling’s? We both have the same data from the same ERP.” Questions like that destroy trust faster than any wrong forecast. If you have two numbers for the same thing, in the end you believe neither.

This month was about the details that this depends on.

A bundle counts once

When we recalculated controlling’s return rate, the numerator matched but the denominator didn’t. The cause: bundles appeared twice on the receipt – once as the bundle line and again as the individual components. In one product group the unit numbers were five to thirteen per cent too high as a result.

Now a bundle sold counts once: as one unit with one revenue. The components are kept separately and remain on record. Pythia also now calculates margin cleanly per unit: revenue minus cost of goods per unit times units.

One channel, many spellings

A second finding was even bigger. In the ERP the same channel appeared in different spellings – sometimes capitalised, sometimes not, sometimes with a trailing space, sometimes with an umlaut and sometimes without. To Pythia each spelling was a separate channel, and when building the figures it hit one of them at random. A considerable share of revenue ended up under “Unknown”, and some large channels looked almost empty.

The error wasn’t in the calculation. It was in the assumption that the same things are written the same way.

Pythia now brings channel names into a normal form before comparing them. Genuine typos and synonyms are a different matter – they deliberately stay a maintenance task for people and are not guessed automatically.

Parcel carriers on the timeline

Every figure can now be split by parcel carrier. The “no parcel carrier” segment – essentially freight – can be clicked, and the drill-down leads from carrier to channel. Because credit notes know no carrier, Pythia assigns them to a remainder segment and says so in a note.

An expectation that can explain itself

Finally, the expectation itself: on a click, Pythia shows how it came about – per previous year the actual, factor, weight and contribution, plus the shopping weather in both periods and the calendar events. Anyone who sees a red candle can check straight away whether the expectation was plausible before looking for causes in the business.

Next month: ready for operations – roles, sign-in and a data build that is three times as fast.