This is the channel that will discuss influencers marketing using surveys.

Why focus on surveys?
New tools that succeed in a particular field are often those that solve a particular problem or offer an easier method of achieving goals that has not yet been adequately addressed by majority of tools on the market.
Even if the new tool has most features offered by other tools, one or two differentiating features of such kind can often lead to faster adoption by the clients.
Survey are excellent sources of information in identifying areas where users are encountering difficulties with current tools or areas which present a substantial concern for them.
Additional survey benefit is providing time evolution of topics. If the survey is showing that a certain area of concern is being addressed and its importance decreasing then it presents an important information for a new developer to not focus on a problem whose importance is already being diminished. If you want a FOSS online tool for surveys, this is a very good choice: https://github.com/LimeSurvey/LimeSurvey. Data obtained with surveys can be used for many purposes, e.g. they can be part of recommender systems.

2. Surveys

2.1 The State of Influencer Marketing 2019 : Benchmark Report
I will first address the influencer marketing survey from Influencer Marketing Hub. Full report can be download here:
https://influencermarketinghub.com/influencer-marketing-2019-benchmark-report/
I will focus only on segments from survey that are relevant in the context discussed in section 1.

Influencer fraud
Influencer marketing fraud is a concern to 64% of our survey respondents. This is because 63% of respondents have personally experienced influencer marketing fraud in their campaigns.
The 64% share is relatively high and indicates that fake follower detection could be an important future feature that users may be motivated to pay extra as it could address their concerns. Fake follower detection can be done by using an AI model e.g. using a deep learning model.
The interest in area is also indicated by monthly visits number from Similarweb analytics for Hypeauditor when compared with other major tools or sites in field-

Website Monthly visits (June 2019)*
Hypeauditor 646K
Neoreach 171K
Influencermarketinghub 1.89M
Buzzsumo 1.07M
Upfluence 171K
* source: similarweb

Majority find influencer discovery moderately difficult
18% of respondents consider selection of appropriate influencers to be very difficult, while only 20% of respondents consider influencer selection easy. Majority of respondents consider the problem of finding suitable influencers to be of medium difficulty.
Although this issue seems to be diminishing in importance (with influx of new tools) in recent period it indicates that a new, simpler and better way of finding influencers could make an important differentiation.
One method to address moderately difficult influencer discovery would be adoption of a recommender engine, perhaps involving images similarly as in Thread.com. Thread.com is using a data science approach to recommending the cloths to their users. Here is a nice tool that can help you with fashion in AI: https://arxiv.org/abs/2005.08847.

Concern about brand safety
Nearly half (49%) of respondents believe brand safety can occasionally be a concern when running an influencer marketing campaign. 30% gave a more definitive belief that brand safety is always a concern.
In another similar major survey – The 2019 State of Influencer Marketing Report by Global Marketing Influencer Marketing Agency Relatable (see later sections below) 84.4% of marketers consider brand safety to be a concern on some level, with nearly 40% saying it’s always a concern, brand safety can be managed by using a machine learning model service.

I see one potential method to address this concern in terms of a new feature.
Keep your brand safe with due diligence of influencers
Each influencer is audited by scanning their posts as well as comments for profanity words, adult content, violence related words. Influencer can be then tagged correspondingly and include this category in filters.
In later versions due diligence based on text could be supplemented with one that also uses AI image analysis. By vectorizing images e.g. using Imagenet and then performing analysis.

In relation to brand safety and content filtering, another important way how one can manage this is by purchasing the URL categorization database and then building an app that uses it for content filtering. More about building such URL database can be found here: https://www.alpha-quantum.com/blog/url-database/url-database/

When influencers decide which products they want to sponsor they can use a product database API which helps them quickly decide which domains are appropriate to be considered as potential sponsors and which not. This is done by using text classification models like RNN to determine categories of products and based on this then the categories of domains. Product database API used for classification models can be an easy way to better annotate products. Here is an example of research paper on A Retail Product Categorisation Dataset: https://arxiv.org/abs/2103.13864 This product database API is usually based on preclassified products that was done already prior and is not done in real-time. Sometimes namely look for pre-defined offline database for products already categorized.

Majority Believe Influencer Marketing Can be Automated, although Significant Numbers Disagree
Survey shows that 57% believe in the automation process. However, a sizable minority, 43%, don’t think that automation can result in successful influencer marketing.
I see this as people receptive to some kind of digital assistant in this field that could automate several or all steps in the processes in influencer marketing. Probably way too early at the moment but worth keeping in mind.

2/3 of Influencer Marketing Campaigns are Campaign Based
Survey findings: there is a trend towards brands cultivating more long-term relationships with influencers.
This could be addressed with a new feature where each influencer is analysed in terms of longevity of their campaigns. So for each tag detected that shows association with a brand, latest and earliest date of posts with this brand are determined. This number is then averaged for influencer/brand combination to indicate if the influencer tends to engage in long-term or short-term campaigns.
A new category or tag could be defined offering the searchers additional filter – to search for influencers already engaging in long-term campaigns.

2.2 The 2019 State of Influencer Marketing Report
Second survey to be addressed is from Global Influencer Marketing Agency Relatable. Report is accessible at:
https://www.relatable.me/the-state-of-influencer-marketing-2019?mc_cid=715645ec41&mc_eid=fb5c170c72
I will focus only on segments from survey that are relevant in the context discussed in section 1.

76% of marketing teams are operating their influencer marketing manually without any tools
This along with many other indicators shows there is potential for growth in the tools segment.
4 out of 10 marketing teams having little to no experience with influencer marketing
Case study of tremendous success of this approach in a similar field:
I started in SEO in 2007 and the major players at the time were Majestic (founded 2004) and Moz (founded 2004). In 2011, 7 years later than the aforementioned two, Ahrefs was founded and in the last years it really seems that it overtook them. So what is their secret of success? They actually published what they did differently that could explain their rise.

Influencers analytics can also help one if interested what are the best stores in given vertical. You could find best stores e.g. by gathering all stores mentions by all influencers and then find out which are the most popular best stores.

An interesting tool that we recently used as part of online stores project is a reverse IP API to find collections of domains that share the same IP address. It is surprising to see how many different domains can be on the same IP. This is due to the fact that hosting companies often put many different domains on the same server in order to have desired profitability and margins. Sometimes this can also be due to using of cloudflare: https://www.cloudflare.com/en-gb/