top of page
  • Writer's pictureLobster

Evaluating online in the world of shopper media…

While online marketing activity is ever growing in its importance, accurately measuring the true impact for brands presents an on-going challenge. Blanket coverage, much like press and TV activity, across grocery e-commerce sites means there is no robust control to compare against. No ‘new news’ there then.

In what should be a phenomenally data-rich channel, it seems that the appetite amongst the Major Mults to uncover what works best for which brand is somewhat absent. On the other hand, brands themselves are asking more than ever the question, “Is my online shopper spend seeing a payback?” . As the ability to robustly measure in-store activity through test and control methodologies becomes embedded into FMCG businesses, the focus turns to measuring online. So how can it be measured?

There are three methodologies to be considered.

Methodology One | Test vs. Control Over Time

This method is one that most will be familiar with: by plotting sales for a SKU or brand over time before overlaying secondary space, promotions and online media, we can identify ‘test and control periods’. The aim is to find the same promotion and secondary space, with and without online media support. Incremental uplifts driven by the media can then be isolated.

So, what are the watch outs? Firstly, as test and controls are identified at various times seasonality can influence the results. Secondly, smaller SKUs or brands that have highly fluctuating sales can make it difficult to identify control periods and therefore isolate the influence of the media.

Methodology Two | Dark Stores

Another methodology that takes brands a step forward in evaluating online is through ‘dark stores’. Dark Stores are effectively warehouses that are used only by ‘pickers’ to fulfil online orders; shoppers do not have access to these stores. Therefore, using sales from these dark stores a test period can be established when online media is live, versus a control period when there is no online activity. Uplift %, Uplift £ and ROI can then be measure and analysed at a Featured SKU and Brand level.

Are there any watch outs here? Yes, as dark stores are limited in number, there can be some issues with robustness.

Methodology Three | Geo-targeted Banners

Finally, and hot off the press, is geo-targeting online banners. Only recently available at, but no other supermarket websites, this method allows the first attempt of robust test and control evaluations of online media.

Banners are served to shoppers within a certain drive time to store, these stores become the Test Stores. Test stores are then matched to control stores, whose customers do not see the online banner (but are otherwise almost exactly the same to the test stores), to create a matched pair.

Once matched pairs are established – Uplift %, Uplift £ and ROI can be established at a featured SKU, Brand and Category level. This final approach is definitely the biggest step forward in terms of understanding how online activity is working for brands on grocery e-commerce sites and one we are excited about at Lobster.


Although there continues to be hype around online, it must be put in perspective – yes, it is growing and yes there is continued innovation in this area. However, online remains a small part of the pie – Tesco, for example, sees c.10% of sales attributed to online which, when worked out in value, is equivalent to something like 40 Extra stores. So next time you look to invest even £10,000, remember that equates roughly to £250 ‘cost per store’. Understanding online cost per store equivalents, coupled with the difficulties of measuring what works for brands means that justifying investment can be difficult for shopper teams. Even if methodologies are not as robust as the classic test vs control matched pairs approach, brands must look test and learn where possible – the above methodologies are a good starting point.

42 views0 comments


bottom of page