‘We don’t track Corona’

Good equity analysis is hard. At many companies, product is ultimately key. But for various reasons, it’s tricky for researchers to get close to the products.

Sometimes, analysts will dip a hand into the cookie jar of anecdotal data (cf. Elder, B., Further Reading, October 31). The limitations of such an approach are obvious, as much as we’d love to see the Wall Street equivalent of mince pie taste testing.

As a result, sellside analysis often sticks to the accounting, or resorts to unconventional measures to judge brand performance.

One example of the latter school of analysis landed in Alphaville’s inbox on Tuesday. Bank of America’s equity team has got its drank on (metaphorically), using data from Insta to “assess brand equity momentum for 50+ spirits and beer brands”.

Analysts Andreas Pistacchi, Victor Beltran-Segarra and Apoorv Gupta write:

1) tequila main share gainer in spirits, but decelerating; 2) Aperol best performing brand on IG, Jameson, Campari strong too

4) Casamigos declining YoY; 5) beer gaining share of alcohol posts in US; 6) Amstel, Tiger, Guinness, Carlsberg best in beer

Here’s the overall shift:

The killer line, for us, is in the report’s findings on beer (our emphasis):

Tiger and Amstel have been the fastest growing international beer brands (IG posts) in recent months, among the ones we track, with posts up 58% and 33%, respectively, in Jul-Sep 22, YoY. Guinness (+21%) and Carlsberg (+13%) have been strong too, while Bud Light, Beck’s and Stella declined on Instagram. The Heineken brand has lost some momentum too. (We don’t track Corona). Interestingly, the beer category continued to gain share, at an accelerating rate, of total alcohol posts in the US in the last 3 months. Wine (and now even spirits) losing share of posts.

Yes, no kidding!

Here are some samples from recent public posts with #corona in the description, based on a search on Tuesday. Unsurprisingly, the commonly-used shorthand for Covid-19 draws a lot of results that are not the AB InBev-brewed beer.

This kind of work can be complicated. Here’s an extract from BofA’s methodology:

[We] use data on Instagram posts provided by Social Standards, a social media analytics company, as we believe this is a good way to gauge brand health/momentum in alcoholic beverages. The data tracks the frequency with which Instagram users post about specific brands/categories.

Under points to consider, the report adds (our emphasis):

• In some cases, IG posts data could be distorted by various factors, not related to the health of the brand, e.g. COVID-19. We have excluded Corona from most of our analysis, given the distortions. The data included in the report is meant to be indicative only and not to be used as a measure of brand/company/category performance. Instead, it is meant to give an indication of consumer interest.

• We excluded spam and company-sponsored promotional content to ensure that the data is primarily driven by consumers.

• Share gains/losses for brands within a category do not add up to zero. This is particularly evident in categories like vodka or cognac in the US. This is explained by the way we calculate share of brands in a category. For example, in the case of Smirnoff and the vodka category in the US, we calculate share of Smirnoff by dividing “Smirnoff posts in the US” by “vodka posts in the US”. Some Instagram users, however, post on vodka (eg “enjoying a vodka-lemonade…”) without actually making reference to specific brands. Therefore, the sum of all posts mentioning vodka brands in the US will not equal the total number of vodka posts in the US.

Ignoring the ~past posts do not guarantee future purchases~ bit, the statement “we excluded spam” is a bold claim when the difference between a power consumer/influencer and a spammer is pretty tough to spot.

Clearly, these are muddy waters. Corona is particularly exposed, but we’d argue others including ‘Fireball’, ‘Beefeater’ and ‘Malibu’ would have problems (#greygoose, unsurprisingly, includes a smattering of nature photography). The value of this analysis pivots hugely on how well this absolute bucket of data has been filtered.

Social Standards, the Oakland, California-based company whose data underpins BofA’s analysis, has a frankly mind-boggling “Our Process” page. Here are some extracts (our emphasis):

Social Standards has developed an intricate system that captures and tracks topics that frequently co-occur on social media. Just like the co-occurrence of beer and diapers on receipts, the connection between two or more products or topics emerges over time . ..

Our team uses text analysis to detect differences, and patterns of differences, en masse. With social media text, we remove contextual commonalities or norms, to reveal a collective group’s differentiating features or preferences, among other collective groups or the larger whole. With a large collective social media as a standard background, we use discriminatory analysis to enable comparison between a smaller collective group and the standard background. The collective element makes possible the comparison between aggregate conversation about one brand versus another, such as comparing two Beverage Alcohol brands’ competitive landscaping in the hard seltzer category. Underneath that aggregate conversation is the essence of collective intelligence — the sheer amount of expression of individuals that reflect the commonality or norm that is common across any day, at any given time, in a dynamic manner.

We’d hate to be accused of cynicism, but we struggle to see how an analytical approach that can capture “the essence of collective intelligence” would be able to, for example, extricate the useful trends about Malibu rum from the 6.1mn posts with #malibu (which, on Alphaville’s inspection of a few hundred recent posts, include a lot of beaches, bikinis and puppies, some other drinks, but little rum), yet be completely foiled by the 30mn-ish #corona pics. Does it work at scale or not?