These are some verification considerations. (in PDF form) Also see Uchaguzi Verification Overview for a best practices. (Note this page content was transferred from the retired Ushahidi community site.)
One of the challenges of a crowdsourcing tool or indeed any platform that relies on citizen reporting is the issue of verification: How can you tell whether information received is reliable or not? The answer to the question is generally a moving target depending on the context in which Ushahidi is being used. There are no quick solutions to the challenge, but Ushahidi offers a few ways, for instance, to rank the veracity of a particular source of information, and to incentivise users to report accurate information.
In most cases, a crowdsourcing tool’s, primary purpose is to help get the information out there. In some cases, it might not always be possible to verify a report, in which case you can mark a report as "unverified", until you receive information that suggests otherwise. If there is any doubt about a report, a good first step would be to mark the report as unverified, until you have the opportunity to investigate further.
In general, when you mark a report as verified it should be under the following circumstances:
numbers about the same incident.
You have a message from a different source (twitter, email, news, etc.) and also a text message confirming that report. You should try to make sure that the messages are actually from different sources first and not the same person using multiple media forms e.g. texting a report and then tweeting it.
You or a site administrator has spoken with the person on the ground to get more detailed information about the report.
Photos or videos have been supplied documenting the report.
You are directly in touch with the person reporting the information. For example: a trained community member, an election monitor who is part of a reputable network, or a trusted source (someone who has given credible information in the past or over a long period of time).
Things to avoid and be aware of during the verification process:
Be aware of poisoned data. Poisoned Data is data that is false, or created to function as bait or create a diversion.
For example: during an election, someone could create false reports about violence at a polling station, or supporters of one candidate could create false reports about intimidation by an opposing candidate.
Be aware that there is a separate "action taken" category to indicate when you know that someone has taken action on a report. For example, an aid agency or government entity has responded to the situation.
Also, it is important to anticipate the quantity of reports that you may receive to make sure you have capacity to keep up with report verification. If you are implementing a long-term campaign, you will need to consider sustainability, which may require training people to administer your instance and thinking about funding. You may want to consider recruiting additional volunteers to help with verification!