Measuring the impact of prices and promotions

Introducing the Evo Pricing Impact Tracker Tweet This

I want to ask you a question: when you change prices, or launch a promotion, how do you measure its impact?

As reported by the Guardian, in 1861 a shopkeeper in Philadelphia revolutionised the retail industry. John Wanamaker opened his department store in a Quaker district of the city, introducing price tags for his goods along with the high-minded slogan: “If everyone was equal before God, then everyone would be equal before price.”

The practice caught on. Up until then high-street retailers had generally operated a market-stall system of haggling on most products. Their best prices might be reserved for their best customers. Or they would weigh up each shopper and make a guess at what they could afford to pay and eventually come to an agreement.

Wanamaker’s idea was not all about transparency, however. Fixed pricing changed the relationship between customer and store in fundamental ways. It created the possibilities of price wars, loss leaders, promotional prices and sales. For the first time people were invited to enter stores without the implied obligation to buy anything (until then shops had been more like restaurants; you went in on the understanding that you wouldn’t leave without making a purchase). Now customers could come in and look and wander and perhaps be seduced. Shopping had been invented.

On top of his invention of fixed pricing, another famous quote was attributed to John Wanamaker. He was noted for being a great innovator and he was one of the first big merchants in the USA to employ advertising agencies to successfully promote his department stores.  Yet, even while outpromoting his rivals, he realized he was still leaving money on the table. “Half the money I spend on advertising is wasted; the trouble is I don’t know which half.”

At the turn of the 20th century that lack of precise targeting was normal. There just weren’t the tools available to give Wanamaker the insight he wanted. 

Now, more than a century later, we have amazingly powerful analytical tools that Wanamaker could have only dreamed of. They can predict how sales, of even a single product, react to new price points, at a particular time of year, under different weather conditions.
This being the case, why is it that 59% of grocery retail promotions destroy shareholder value according to the Tesco’s data consultancy branch, dunnhumby?

The answer is very simple: lack of feedback.

Many retailers don’t have a systematic method to measure the impact of prices and promotions taking into account seasonality, the type of store, or product substitution, to determine optimal price points. And therefore they struggle to design and measure tests, slowing down the creative test & learn process at the core of sound management practices. This is a mistake, because taking a more global view and achieving greater accuracy does help boost profitability.

Some managers may for example think that, when only a relatively small number of units are involved, price testing is not effective because the signals are too small to analyze. Or, when a promotion is sluggish, they might blame the price when the issue might have been linked to the weather or a low footfall.

We have spent many years to develop specialistic expertise in this field, and will soon launch a dedicated Impact Tracker offering to fill this critical void in strategic pricing. But in truth there are 3 different approaches to measuring the impact of prices and promotions, and 2 of the 3 can be replicated by any practitioner in a simple yet effective, powerful manner:

  1. Pre-vs-post: suitable when a low percentage (<25%) volume of SKUs undergoes a change in price. This method allows for immediate  feedback, on a short-term basis, and has the substantial advantage of being simple and transparent. The volume impact is measured by comparing the performance of each item in the period preceding the price change (“PRE”, typically set at 4 weeks) to the period following the change (“POST”, from 1 up to 4 weeks maximum), using the volume change of all other items that kept the price constant, as a baseline. So if for example the “no price change” volume has increased +5%, but the “price change” volume has changed by +2%, then the “price impact” will be set at -3% (the difference between the two). The revenue and margin impacts can then be derived using the pre/post pricing information (revenue equals volume times price!). This method accounts for seasonality, but not for item substitution, and only works when items have historical comparability (for example, if a new bundle offering across products is introduced, this will not be comparable to a “pre” period).                                                                                                                      
  2. Baseline-vs-forecast: suitable up to a medium-term time horizon, up to 3 months after the change, even when price changes affect more than 25% of SKU, and particularly fit for the analysis of promotions as it accounts more properly for substitution and cannibalisation, when compared to the simpler “pre-vs-post” method. In this case the volume impact is measured by comparing actual sales to those forecasted had the price remained unchanged. This is a more challenging method to implement, and relatively ‘black box’ - requiring advanced mathematical modelling that impedes a simple ‘pen and paper’ explanation. Whereas the ‘pre-vs-post’ results can be easily replicated using a pocket calculator (or Excel), a proper ‘baseline forecast’ requires a deeper understanding of the interplay among variables affecting sales, and the use of dedicated software. This is the kind of work that we thrive on at Evo Pricing: much like Google has been working on mastering ‘search’ (a never-ending quest), we aim at becoming the world’s best at ‘forecast’ - a challenging pursuit.                                                                                                                                                                                                                                    
  3. Long-term: impact tracking for a price change can be extended all the way until the end of the items’ product lifecycle or until a new price is applied (possibly years later), by using the elasticity measurement derived with the previous 2 methods to compute the baseline volume of daily sales at the ‘old price’, and can therefore be extended over time. This approach is simple and appropriate for multiple waves of price changes over long periods of time, but its accuracy declines as the length of the forecasting time horizon increases. Clearly, long-term monitoring will only be as accurate as the elasticity measurements that feed into it, and this is an ‘approximation of the truth’: elasticity is a great descriptive tool, but a poor prescriptive method due to volatility and lack of predictive power. However when time horizons are long, elasticises tend to be well-behaved and therefore fit for purpose to the measurement of cumulated impact.

Having a systematic approach to impact tracking leads to identifying patterns and possibilities in a transparent way, requiring less manual work and adjustment, accelerating test&learn, and suggesting opportunities for optimization, even mid-promotion. And, as new decisions are implemented, automated learning enables feedback to further refine accuracy and therefore enhance impact. 

The ability to assess the outcome of one’s pricing strategies is critical to unleash the true power of analytics, and foster the alliance between data-driven systems and human intuition. At Evo Pricing for example we use 3 methods to formulate price recommendations:

  • When responding to rapid changes in sales and movement of top and bottom performers, the recommendations are driven by the SKU trend. Strong performers can sustain a price increase, poor performers should have a price decrease or promotion.                                                                                                                          
  • Elasticity driven recommendations are issued when opportunities to increase overall margins are identified based on the simulated sales at a new price versus baseline forecast at the old price.                                                          
  • Past results of price changes are the third driver, behind recommendations aimed at automatically reviewing and reverting the issues observed with the implemented pricing strategy, to capture the maximum impact. 

A systematic impact tracker can therefore fully unlock the hidden value behind consumer reaction to prices and promotions and unveil which half of the budget was being wasted - John Wanamaker, eat your heart out!


About the writers

 
Martin Luxton is a writer and content strategist who specializes in explaining how technology affects business and everyday life. Big Data and Predictive Analytics are here to stay and we have only just begun tapping into their enormous potential.

Martin Luxton is a writer and content strategist who specializes in explaining how technology affects business and everyday life.


Big Data and Predictive Analytics are here to stay and we have only just begun tapping into their enormous potential.

 
Manuela Vaturi is a Junior Product Manager at Evo Pricing. She is currently undertaking an MSc in Management at Cass Business School and working on introducing Evo Pricing's products to the industry.

Manuela Vaturi is a Junior Product Manager at Evo Pricing.

She is currently undertaking an MSc in Management at Cass Business School and working on introducing Evo Pricing's products to the industry.