Half the software shipped at zero

Pricing a licence when nobody knows the price of the box

The new product replaces a family of hardware configurations with a single host plus software, and switches features and module integrations on after the sale. That is a sound commercial model with one prerequisite. You have to know what the existing line transacts at, because the licence price has to be set against it.

The stated problem was that the current models sold across a wide price range, and that the range depended on the sales representative and on country dynamics rather than on anything written down. The sprint was scoped as a proof of concept, to establish whether the transaction data could support a pricing decision at all before anyone built a model on top of it.

Four European markets were taken as a cross section, and one full year of transaction data was analysed at item level.

What the data would and would not carry

Hardware came through clean. Host orders went from 666 to 629 after removing zero priced records and outliers, and host units from 2,511 to 2,377, retaining 94 and 95 per cent. That is a reliable base to price from.

Modules did not. Order level retention fell to 67 per cent, with 330 orders removed for carrying no price at all. Unit retention held at 85 per cent, which tells you the discarded orders were the small ones. Several module cells in individual markets rest on one, two or three transactions, and a range built on one transaction is not a range.

One inference had to be flagged rather than fixed. Care area, meaning whether a monitor went to an operating room, an intensive care unit or emergency and recovery, is not recorded in the transaction. It was inferred from the software package sold alongside the host. The number of hosts sold is about 20 per cent lower than the number of packages sold, so that allocation is an estimate built on a proxy and is labelled as one wherever it appears.

The same box, the same country, four times the price

Within a single host configuration in a single market, the observed transaction range after outlier removal ran to roughly four times between the cheapest and the dearest unit. That is not a mix effect and it is not a currency effect. It is the same product sold to different buyers at very different numbers.

The pattern across markets was consistent enough to be useful. One market priced lowest and most tightly and also sold the most units, which is usually what a disciplined channel looks like. Another priced highest and loosest, and it was also the market where the most records had to be discarded as outliers, at 18 per cent of units against nought to two per cent everywhere else. Loose pricing and dirty data turned out to be the same phenomenon.

Actual price levels are not reproduced here. They are live commercial terms for an identifiable product family, and publishing them would hand the client customers and competitors its country pricing. The ratios carry the finding without doing that.

The finding that mattered for a software led model

Of 11,624 software units shipped across the four markets in the year, 5,367 carried a price. The other 54 per cent went out at zero. In two of the four markets, individual licences were sold at zero as a matter of routine.

That is the single most important number in the study for a business about to make software the product. Every one of those units is a feature the customer already receives and has never been asked to pay for. Repricing them is not a price increase in the ordinary sense. It is charging for something that used to be handed over, and the internal resistance to that will be higher than any competitive resistance.

Where software was priced, the ranges sat an order of magnitude below hardware and still varied widely. The widest of the top selling licences ran from 12 to 616 in one market for the same item.

What the client was left holding

Two macro enabled pricing tools, one for hardware and one for software, that take transaction level sales data and return minimum, maximum and weighted average price by product and by market in a single summary table. Pivot sets holding the full year of ranges for all four markets. And the cleaning rules that make the output trustworthy: one consistent source, zero and negative records removed, outliers removed, one currency, and focus on the items that actually carry volume.

This was a proof of concept and the deck says so. It produced no price recommendation and no revenue effect, and none is claimed here. What it produced was the ability to set a price, in any market, from data that had been sitting there unused.