Why did our approved project never go live?

That is what I mean by decision rights: not only who is allowed to overrule a team, but whether the decisions the project depends on are made before the money is committed. Authority used after approval, one block at a time, can leave the people overruled obstructive for the rest of the project.

The published claim, and what it assumes

In March 2017, McKinsey published its view of what dynamic pricing does for a retailer. Where the tool is customised to the business and adopted by the people who set prices, the article says, it typically brings “sales growth of 2 to 5 percent and increases of 5 to 10 percent in margins”.

I have no evidence against that range. The same article names one reason the gain may not arrive: when pricing staff cannot see how the model reached its answer, “The pricing staff thus ends up rejecting them entirely because they don’t trust the recommendations.”

The range rests on a word in its own sentence: adopted. Adoption is where decision rights come in. A model that recommends a daily price is worth its range only in a business that can show a daily price in its stores, change it in its systems, square it with its media plan, and have its pricing team accept it.

Three explanations, and the one that leads

When an approved project stalls, three explanations are on offer. The technology was not ready: real, but readiness is something a plan can schedule, so it explains a delay rather than a project that never starts. The people resisted: real too, and McKinsey’s explanation is a version of it, but resistance can be a team defending a process it built and is measured on.

The third is decision rights, and I think it leads because it explains how the other two add up. A scheduling problem and a trust problem each have a fix. Four of them at once, each owned by a different team, do not get fixed by overruling them one by one, because each team overruled stays a drag on the next step. What would prove me wrong is a project whose blocks were all settled before approval that stalls in the same way.

What the wait costs

The numbers here are an illustration. Take a business with £30m of gross margin a year. McKinsey does not say whether “5 to 10 percent” means percentage points or a relative increase, so take the smaller reading: margin 5% higher than before is worth £1.5m a year. Every month the approved model sits switched off costs £125,000 of that. The cost is easy to miss, because the project is approved and the money committed. The benefit has simply not started.

The case I know best is a machine-learning pricing model that was reviewed, costed and approved, and never went live. It waited on shelf labels, a middleware and ERP upgrade, a media plan built on fixed price points and a Commercial team unconvinced that a system could improve on its own experience. None of the four was about the model. Any one of them could have been overruled, but the teams overruled would have stayed obstructive along the way. It was the combined impact of the four that killed the project.

What to do before the next approval

Write down every team whose way of working has to change for the project’s output to be used, the block each one holds, and the owner of each. Then settle each block before approval: agree the system change, the media change and the process change with the people who own them, rather than leaving them to be overruled once the money is spent. If a block cannot be settled before approval, the board has learned something useful before spending anything.

The cheaper middle route is a pilot, which is how McKinsey’s own examples began, but for a multichannel retailer the obvious pilot is the wrong one. An online-only test sounds simple, and it was discussed on the project described above. It was not appropriate: there were very few online-only products, and the business had a duty to its customers to sell the same product at a consistent price in both channels. A pilot that lets the website and the stores show different prices for the same product tests the model at the cost of the customer’s trust. The pilot has to cover a category across both channels, which means the shelf labels, the systems and the media plan for that category have to be ready first. That is a smaller version of the same four blocks, and it is a better place to settle them than the whole business.

How you will know

Run the pilot categories against matched categories that keep the old pricing, matched for season and stock, with the same promotions in both. Measure gross margin and units sold in each, with customer complaints about price as the guardrail. The pilots in McKinsey’s article were read after a few months. Agree a read date of that order before you start, and agree the rule: if the pilot categories do not beat the matched ones on margin without losing units, stop.

The decision

The decision for the board is whether a project is approved with its blocks settled or with them still open, because open blocks compound after approval. The question for Monday: for the last project this board approved, how many teams had to change for it to work, and how many of them agreed to before it was approved?

This rests on one retailer’s project and one consultancy’s published range. It cannot tell you how often approved projects stall for this reason, and I have not seen anyone count.

My verdict on the 2017 claim: I have no evidence against the range, and one approved project’s worth of evidence that it applies only to a business that has settled every block between the model’s price and the customer before it starts.

Sources

1. Gadi BenMark, Sebastian Klapdor, Mathias Kullmann and Ramji Sundararajan, How retailers can drive profitable growth through dynamic pricing, McKinsey, 27 March 2017

Back to Thinking