They are loved and hated in equal measure. Embraced and dismissed depending on the outcome, opinion polls have become a mainstay of Kenya's politics.
As the clock ticks to the August 9 General Election, opinion polls have come thick and fast.
After initially leading the early polls, popularity surveys done recently have shown that Deputy President William Ruto has been overtaken by his rival Raila Odinga in a two-horse race to succeed President Uhuru Kenyatta.
However, another opinion poll released last week by Intel Research Solutions (IRS) has shown that Raila trails Ruto.
Raila was unhappy with a recent poll which portrayed as losing grip of Bungoma. "An opinion poll which was carried out last week shows that Raila is trailing Ruto in Bungoma. Ruto is at nearly 60 per cent and Raila is at 40 per cent. Shame on you. I feel slighted because these are my people," he said.
Pollsters say the dismal of opinion poll results by candidates is a campaign tactic and a charade because most of the politicians have commissioned their own polls to gauge their popularity and are aware of the reality.
Political opinion polls have been a subject of criticism for politicians and have often been dismissed as being inaccurate and having no importance to the contest. Some of the issues that have been raised concern the ownership of the research firms and the sponsorship of the polls.
For instance, DP Ruto has consistently dismissed opinion polls as a campaign tool for his opponents. The United Democratic Alliance candidate says the polls have been concocted to show that he was trailing.
However, survey firms that belong to the Market Survey Research Association of Kenya (MSRA) must adhere to the association's laid down procedures, particularly with regard to polls that relate to an election.
The procedures require the firms to be politically neutral to avoid situations that create the perception of being partisan.
Maggie Ireri, the CEO of research firm TIFA said survey firms that conduct political opinion polls are required to ensure the questions they ask have minimal bias and ambiguity, to avoid misinterpretation by respondents.
The phrasing and positioning of questions should be balanced so as not to influence responses. The association recommends questions such as “If presidential elections were held today, whom would you vote for if that person were a candidate?” or “If you have decided, whom are you intending to vote for as president and deputy-president in the forthcoming election?”
But the industry is facing a problem of research firms that from the outside look purpose-built to solve political problems. "There are others that have emerged and there is a general suspicion that they emerged in an election period and they seem to be publishing opinion polls left, right and center," she said.
All opinion polling samples are required to be statistically representative of and proportionate to the population recorded during the census or a published voter register.
The MSRA recommends a minimum sample size of 1,500 respondents for an election-related national survey and a margin of error that is +/-2.5 per cent.
To ensure transparency the association requires public releases of opinion polls to include details of the sampling method used (such as whether random and if not, on what basis), the basis of sample size determination, achieved sample, geographical distribution, margin-of-error, and confidence level, the demographic details of respondents, the interview method and the fieldwork dates.
The association further requires the identity of the sponsor of any election-related survey where the sponsor is a single client, the percentage of respondents who refused to answer, and the percentage of those (registered to vote) who do not intend to vote in the election.
There have been concerns about the credibility of opinion polls and some pollsters. There are polls that may be faulty because those undertaking them don't have sufficient technical capacity for correct methodology and in presenting the data.
But a darker concern exists about polls which are deliberately falsified either because of political ethnic bias on the part of who owns the company or for financial gain.
In some polls, the criticism has been that the methodology used in the sampling of voters can skew the outcome of the poll.
For instance, a sample used by pollsters could over-represent a particular coalition's supporters and under-represent another. Politicians have constantly attacked unfavorable opinion polls, helping to create skepticism about them
Tom Wolf, said when he appeared for an interview at KTN News that because the level of understanding of the science of polling is less familiar in Kenya, it encourages those who were especially unhappy with results to make personal attacks.
Opinion polls on social media are, however, a different ball game. Polls on social media by political analysts and opinion leaders have also emerged as a method to determine the popularity of presidential candidates.
Twitter polls have been a go-to and often feed into the social media strategy employed by the candidates. Infotrak was forced to abandon and delete a poll it was conducting on Twitter after discovering that some of the engagement was not organic.
The use of an autonomous programme on social media site that interacts with users (bot) and (sock puppet accounts) whose actions are controlled by another user is prevalent and could be used to deliberately skew the results of the poll.
"When we did this Twitter poll we realised there were bots and we decided to pull it down," Infotrak CEO Angela Ambitho said.
She said that while a Twitter poll was not scientific, it could gauge the opinion of users of the social media site.