[Tribune] Pitfalls to avoid when setting up your Social Room

The purpose of a Social Room? Monitor what’s happening online using large screens and numerous dashboards. Will abundant KPIs, good dataviz, reliable technical support and your persuasion capacity be enough to make this operation a success?

1st pitfall to avoid: rely on a single tool

Marketing Manager, Digital Manager or Director of Communications, perhaps you have no experience in monitoring social networks, but you rightly believe that it is essential to monitor online conversations by mentioning your brand and your competitors. You may have heard of social listening or analytics tools before, as there are many on the market (Talkwalker, Brandwatch, Digimind, Meltwater, Netbase etc …) and think it’s probably a good starting point for your Social Room.

After several weeks of benchmarking, numerous meetings and a few quotes, your choice is made: choose a service provider who can configure the tool for you. In fact, it will accompany you to determine the keywords and create the request that will allow the social listening tool to scan the contents on the web citing your brand or that of your competitors.

After a few weeks of use, your colleague from the community management department takes stock of the latest communication campaign. Problem: Its figures are much higher than those announced by your social listening tool. You tell yourself that there is a flaw in the instrument …

Indeed, it is very common that the announced figures are not identical from one platform to another. Likewise, some data internal to your brand will be difficult to access for an external tracking tool. For example, community managers, administrators of the group’s Facebook or Instagram page, have more data due to the limitation of external access to the APIs of these different social networks. There are solutions (token management, page configuration, etc.) but the first observation that can be made is that of the need to associate the internal data of the company (so-called proprietary data) with external data (obtained through social listening) .

However, variations in the volume of external data can still be observed depending on the sources. Why ? Each social listening tool performs a real-time web scan based on the various requests made by all its customers. Some tools will work best on Twitter, others on forums, or even online print. The set of all the searches carried out constitute the backdata of each instrument. Therefore, when trying to extract data on one or two years of history, it is this database that is consulted. Completeness on the web does not exist. Each tool is complementary and has its own database based on the requests made by its customers.

The amount of content produced online every second is such that claiming completeness when it comes to crawling the web is a myth. Therefore, it is easy to understand that no tool can provide identical figures. Without multiplying the subscriptions, it is important to leave the necessary freedom to be able to change the instrument if necessary. The best way to assess the relevance of a solution in your industry is to be able to test it before signing up for any subscription. A software can indeed prove to be quite relevant on a given topic and provide much less results in another field of activity.

2nd pitfall to avoid: skipping the qualitative analysis

Conversation volume, share of voice, tone percentage, virality index, Net Promoter Score, Earned Media Value … all KPIs are available. Averages, percentages, rates, your management swears by the numbers and your dashboards are there to meet those expectations.

However, when it comes to formulating a recommendation, action plan, or strategy, the numbers are of little help. And when asked about the reasons for your competitor’s increased voice sharing, you find yourself somewhat helpless. You must therefore “ dip your hands into the data to be able to extract real information. The first step is obviously to read and literally analyze a sample to understand how Internet users express themselves.

For example, among all these verbatim, isolate the 5 most viral publications, analyze their contents and if you find that they all come from Instagram, that they use the same lexical field and that they are published by accounts with a large audience, then it is probably an influencer campaign marketing. Now you can see that your competitor has seen a double-digit increase in their share of voice on social networks over the past 3 months, but you also know how.

It now remains to decide how to counter this competitive communication campaign: are you also running an influence campaign? Replicate by other means? If the final answer remains in your hands, you now have all the cards to judge the situation correctly. Numbers alone, without a qualitative and in-depth human analysis, would not have allowed these decisive insights to be extracted.

Third trap to avoid: trusting artificial intelligence

Imagine: This morning the social listening tool detected an increase in positive conversations of about 12%. No specific communication operations today, you wonder where this increase could come from. The idea of ​​sharing information in the communications department crosses your mind, but you know you don’t ignore qualitative analysis. Then you look at the data and you did well.

Instead of positive conversations, you notice mocking tweets, ironically mentioning your brand. Why did the tool qualify them as positive conversations? These are the limits of artificial intelligence. Currently, the software is not yet capable of capturing the second degree, context, irony, or sense of humor. However, humor is the main vector for the dissemination of information on social networks. It is therefore clearly not an element to be overlooked in data analysis.

The error rate during automatic tone qualification is between 20% and 50%. A tone error of about 20% to 50% can completely change a ranking among brands, analyze biases and lead to bad strategic decisions. Without human verification of automatic qualification, it is therefore always essential to take a step back from the figures communicated by the tool.

4th trap to avoid: underestimate the noise

Now take your AI tweezers and know that manual qualitative analysis is required before issuing strategic recommendations directly from dashboards. Despite tweaking the tool and improved filtering to exclude “spam” content, it turns out that irrelevant information is always present in the results. You still think this noise must be in the minority.

Be careful though, because if you take the time to analyze these conversations one by one, you may find that the percentage of irrelevant data is perhaps not that small. Obviously it will depend on the topic analyzed, but what are the strategic decisions based on KPIs comprising 65% of noise? 65%: this is the share of noise detected by Groupe 361 on the data provided by the social listening tools. And this, despite artificial intelligence and optimized settings.

Manual sorting of all this data takes time. However, this work can be outsourced to an external service provider. Otherwise, you will always need to step back and be careful when reading dashboards based on data that have not been manually corrected.

5th trap to avoid: allocating resources for a one-shot project without considering the medium-term implications

After a few months of use, you may find that LCD screens will only have cosmetic features. Everyone will therefore have become accustomed to their presence despite their strategic position. Because even though dashboards are automated, it will really be necessary to put the people behind this engineering so that the data being transmitted can truly be exploited. As you can see, setting up a Social Room without a dedicated team wouldn’t make much sense.

If the purpose of the Social Room is to monitor its E-reputation, this is not a goal in itself. Likewise, dedicating a team without assigning specific goals would bring little value. Why do you want to monitor your electronic reputation? If a crisis is detected through these channels, do you have the resources to react? with which levers and which processes? If an opportunity is detected thanks to dashboards, would you be able to seize it? How will the marketing departments use the indicators that will be sent to them? A social room? So what? This is the whole question that needs to be resolved before allocating technical and human resources to a long-term project. Appropriately anticipated, the establishment of a Social Room can really position itself as a central and driving element of the marketing, communication and digital departments.

You now have all the cards in hand to successfully set up your Social Room. You know it’s important to cross multiple data sources, qualitatively analyze a sample of conversations, step back from the insights provided by artificial intelligence and, if possible, manually correct irrelevant data and publication noise.

These steps are time-consuming and require the experience of true professionals. You can count on the help of an expert agency who will be able to support you in this type of project. Do not hesitate to contact the AACC with any questions.

Tribune by Lise Duchatel, Social Listening Manager, Groupe 361 agency, member of the AACC.

Leave a Comment