Maintenant si je vous dis auto-école en ligne vous pensez à quoi ?

Ok, vous pensez qu’il n’y a rien de très excitant, n’est-ce pas ?

J’avais le même a priori lorsque j’ai entendu parler pour la première fois d’éducation routière et d’Ornikar. Pourtant, j’étais loin d’imaginer à quel point les enjeux étaient nombreux et passionnants.

Pourquoi décider donc de rejoindre Ornikar alors qu’il y a des centaines de startups qui à première vue ont l’air plus “sexy” ?

Spoiler alert : rejoindre Ornikar c’est comme rejoindre 4 startups en une !

Enjeu clé n°1 : Tech & Product

Construire des produits webs et mobiles ultra-performants

Le succès d’Ornikar est avant tout basé sur son produit. L’enjeu est de construire des produits webs et mobiles permettant de rendre 100% digital un parcours de formation de plusieurs mois. Notre produit est bien évidemment une plateforme de révisions du code de la route mais pas que. Notre plateforme orchestre l’expérience des candidats de manière à leur faire obtenir leur permis de conduire le plus facilement possible : cela implique de synchroniser dans le bon timing à la fois la formation théorique au code, les réservations de leçons de conduite mais aussi le parcours administratif pour l’inscription aux examens officiels. Les candidats doivent arriver prêts pour réussir leur permis. Mais Ornikar c’est aussi une application mobile pour les enseignants qui leur facilite la gestion et suivi pédagogique des candidats. La technologie nous permet ainsi de ne rien laisser au hasard pour assurer les meilleurs taux de réussite.

Les besoins et interfaces de nos deux cibles sont différents mais notre ambition est la même : être capable de s’adapter en permanence aux besoins des différents utilisateurs pour construire la plateforme qui englobe toute la complexité de l’examen le plus passé en France (aussi le plus passé dans le monde!).

💻 Vous l’avez compris, si vous êtes développeur front-end ou si le développement de nouveaux services et APIs est votre dada, vous ne serez pas déçu en rejoignant Ornikar !

Enjeu clé n°2 : Data

Scaler une marketplace en équilibrant l’offre et la demande

Vous l’avez compris, Ornikar c’est une startup qui en plus de commercialiser un produit (le code de la route), commercialise un service (des leçons de conduite) via une marketplace qui met en relation les candidats avec des enseignants partenaires qui proposent sur notre plateforme des leçons de conduite. Si Ornikar est devenu aujourd’hui le leader de l’éducation routière en France, c’est parce que nous avons réussi à nous positionner comme un véritable tiers de confiance entre nos 400 000 candidats et plus de 400 enseignants partenaires. Avec une croissance exponentielle de l’offre et la demande de leçons de conduite aux cours des derniers mois, le matching est devenu un enjeu crucial chez Ornikar. Comme dans toutes les marketplaces, l’enjeu n’est pas comment mettre en relation deux personnes, le vrai enjeu est comment mettre en relation des milliers de personnes simultanément de façon optimale.

La data est au coeur de cet enjeu. Elle nous permet d’affiner de plus en plus notre algorithme de matching mais aussi d’identifier les noeuds de tension dans notre marketplace pour que les équipes en charge des enseignants partenaires puissent réagir au plus vite. Les enjeux opérationnels sont encore nombreux pour continuer à assurer le bon équilibre entre l’offre et la demande : maintenir la qualité du réseau actuel d’enseignants, continuer à recruter les meilleurs enseignants de France et accélérer le déploiement géographique pour couvrir l’ensemble du territoire français et bientôt européen !

👫 Donc si vous êtes avez la fibre commerciale et le sens du contact, vous devriez trouver votre bonheur chez nous.

Enjeu clé n°3 : Marketing

Alimenter la croissance et construire une marque leader

Avec 1,5 millions de candidats qui se présentent aux épreuves du permis chaque année, Ornikar est loin d’être dans un marché de niche. Les enjeux en termes de branding et d’acquisition sont nombreux pour conquérir ce marché.

Cela implique tout d’abord de construire une marque forte qui parle aux jeunes et dépoussière l’image de l’auto-école. Le design chez Ornikar est actuellement en pleine évolution pour coller à cet objectif (on vous réserve des surprises à ce sujet ! 🤫). Nous voulons que notre design dédramatise le passage du permis et rende l’expérience plus agréable et ludique, à l’image du produit que nous construisons.

Le challenge est de taille : faire connaître notre marque auprès du grand public pour créer le « réflexe Ornikar » chez tous les potentiels candidats au permis de conduire. Ornikar a bénéficié d’une forte médiatisation à son lancement, mais nous ne comptons pas nous en arrêter là. Vous avez peut être vu les campagnes télé et métro que nous avons effectué à la rentrée. Notre objectif est maintenant d’accélérer pour continuer à déployer notre marque et développer la notoriété d’Ornikar sur les différents canaux online et offline.

🌍 Si vous êtes ambitieux, passionné par l’analytics et l’acquisition online, ou si voulez participer au déploiement d’une marque innovante à grande vitesse vous aurez toute votre place chez nous.

Enjeu clé n°4 : Expérience client

Passer de la satisfaction client à l’enchantement

Combien de startups une fois après avoir vendu leur produit accompagnent leurs clients pendant plusieurs semaines voire des mois ? Le passage du permis peut être un moment stressant et nous sommes très fiers chez Ornikar de l’enchantement avec lequel nous traitons toutes les demandes de nos candidats. Offrir une expérience client exceptionnelle dans un secteur où l’expérience client était très insatisfaisante est la raison même de la naissance d’Ornikar. L’enchantement à l’Ornikar est un peu notre secret sauce et c’est la raison pour laquelle nos candidats nous aiment autant :

Notre plus grand enjeux en ce moment est de continuer à enchanter nos candidats et maintenir notre niveau de satisfaction élevé malgré le nombre toujours plus important d’utilisateurs sur notre plateforme.

❤️ Si vous êtes passionnés par la relation client B2C et que vous cherchez un défi pour nous aider à scaler notre service clients, Ornikar est définitivement fait pour vous.

Vous l’avez compris, le terrain de jeu est très grand. Mais au-delà de la diversité des challenges passionnants auxquels nous nous attaquons au quotidien, Ornikar est une startup avec un vrai impact sur la vie de milliers de personnes.

Peu importe la raison pour laquelle chaque candidat souhaite avoir son permis (maximiser les chances de trouver un travail, accéder à sa liberté en termes de mobilité, etc), nous avons entre nos mains la responsabilité d’accompagner et former les futurs conducteurs de demain pour permettre à chacun de s’ouvrir de nouveaux horizons. Et savoir qu’on se lève tous les jours pour accomplir cette belle mission, ça n’a pas de prix !

Vous n’aviez jamais réalisé que la sécurité routière pouvait être si passionnante, n’est-ce pas ? 😉

Ne vous inquiétez pas, vous êtes encore à temps de monter dans la fusée dans cette team de choc ! 🚀

N’hésitez pas à me contacter pour savoir plus sur les différents postes qui sont ouverts ou m’envoyer vos CV à paola@ornikar.com

They are all based on speech recognition technology that has become quite popular over the last few years. Although speech recognition is an old discipline, deep learning leapfrogged the field into the 21st century and caused a significant decrease in the error rate and an increase in performance.

Deep WHAT???

I know, those terms were quite unfamiliar to me as well a couple of months ago. In this article, based on the Speech Recognition course given by the Facebook Artificial Intelligence Research Lab at École 42, we’re going to decipher how speech recognition systems work and understand their current limitations.

What’s behind speech recognition technology?

A system like Google Home basically does 4 things:

1. Identify your voice — speaker identification

2. Recognize what you are saying — speech recognition

3. Understand what you meant — natural language processing

4. Renspond — speech synthesis

We’re going to focus on speech recognition to understand how machines are able to transcribe human speech by using deep neural networks.

How do speech recognition systems work?

First of all, we need to understand how the human body produces speech.

Speech is nothing more than air expelled by our lungs that travels through the trachea, then the vocal folds, and is finally expelled by the nose, the mouth, or both, where it is shaped by articulators (teeth, tongue, palate, lips) to form sounds. To capture these sounds, microphones measure the variations in air pressure that we create when we speak. Microphones represent these sounds as a waveforms, which look something like this:

Trying to replicate the human ear

Trying to extract the word transcription (the word “cat” for instance) only by looking at the waveform is impossible. Instead, we’re going to extract the phonetic transcription, how the word actually sounds. However, since a waveform only plots the amplitude of a sound (it’s loudness) over time, it is very hard to differentiate between particular types of phonetic content (phonemes such as k, æ, t). We’re going to extract acoustic features of that signal which will characterize it with values that have connection to its physical properties. For the sake of simplification, we’re going to say that sound has basically two physical properties:

• Pitch (frequencies)
• Loudness (amplitude)

To be able to picture sound, we need to plot both parameters with respect to time. We’re going to use a spectrogram, a visual representation that plots frequency versus time with color representing the amplitude.

The role of feature extraction is to transform the signal into a representation of the acoustic signals that will be relevant for machine learning, making thus learning easier for our algorithm. This is in fact done to try to reproduce the human ear perception. But it turns out that the human ear is more sensitive to variations in low frequencies than in high frequencies. Without going into many technical details, Mel-filterbanks logarithmic scale is going to be applied to the spectrogram to reproduce this.

Now that we have a visual representation that matches as much as possible the human ear perception we need to slice it into short-time segments. Indeed, to be able to extract speech features from the spectrogram, we cannot process a whole sentence at once; we need to slice it into small fixed-length analysis frames (typically 10 milliseconds) shifted with a fixed timestep. This is where deep neural networks will enter into the game.

Deep acoustic model, where the fun part starts

The deep acoustic model is going to extract higher-level representations of speech features and compute the probability P that in that particular frame we’re looking at, a phoneme (k, a, æ, t) is pronounced, based on the features that our system is looking at. It takes features as input and outputs probabilities of phonemes. But how are deep neural networks used here? If you are a maths person, you probably know we are using conditional probabilities. If you are not a maths person, this deep acoustic model is just a stack of hidden layers that are going to do several computations on top of each other. At the end it is going to output the probability of a phoneme after taking the probability of the previous layer as an input.

I won’t go into many details but today there are 2 main types of deep neural networks used for speech recognition:

1. Recurrent acoustic model: the system sequentially sees one vector after the other. When processing a vector, it will not only use information that it sees in the features but it will also use the information from the previous prediction. For instance, it will take into account the fact that the probability that there is a vowel after a consonant is higher than having a consonant again because that’s usually how words are built.
2. Convolutional acoustic model: it’s the same type of neural network as the one used in computer vision. It will look at speech features as if we were looking at an image and it is going to use some context around the speech features to compute probabilities.

From phonemes probabilities to phonetic transcription

Our deep acoustic model gave us the probabilities of phonemes. We still need to extract the phonetic transcription of this before transcribing it to words. Since the small fixed-length analysis frames we used in the acoustic model doesn’t necessarily correspond to a single sound — because a sound might occupy several frames— we need to extract the maximum probability for each phoneme in each vector to train our system. However, it is very expensive to correctly label each vector with the correct phoneme because it needs human manual annotation done by professionals. Instead, we can use a simple technique called the Connectionist Temporal Classification. Basically, instead of maximizing the probability of one particular phoneme, it will compute each path’s probability and maximize the valid paths to give us the phonetic transcription of the word. It will consider the outputs from the acoustic model as a graph and take each probability of a phoneme as a node. Each path will have its own probability, which will simply be the product of the probability of the nodes.

Crossing the finish line: word transcription

Now that we have phonetic transcription we still have one more step to go to transcribe it into words that have a meaning. As easy as this might sound it is not, since we could have similar phonetic outputs but totally different words and meanings (“Madam” and “I’m Adam” are phonetically very similar but don’t have the same word transcription nor meaning). To take the phonetic transcription and transform it into a word transcription, we will use a language model. It will basically take the phonetic transcription, use the frequency of words (computed by analyzing big corpuses of text) and how likely they are to appear together to correct the errors of the acoustic model and the phonetic transcription.

Well, not so fast my friend…

Even if the best systems are now on par with humans on conversational speech in English (with an average 11% Word Error Rate), machines are quite far from being as efficient in speech recognition as humans for several reasons:

1. Accented speech and noisy speech (with heavy environmental noise) are still very challenging and can totally break the efficiency of a speech recognition system.
2. Languages without standard orthography such as Moroccan Arab are problematic since two different people won’t write a sentence in the same way because there is no standard way to write words.
3. Language evolution is a problem since new words and accents continuously appear in the language and the speech recognition system needs to be retrained to incorporate these evolutions.
4. Unlike humans, lots of data are needed to train a speech recognition system (more than 12 000 hours of speech recorded and annotated vs between 200 and 2 400 hours of non annotated speech for a kid at the age of 3). Unfortunately that huge amount of data is not available for all languages, for example low resources languages (with low amount of speakers).

Humans 1 - Speech recognition systems 0

I must admit it is quite unfair to compare machines and humans. Unlike speech recognition systems, we are not just an ear on legs. Our speech recognition system uses:

1. Our senses: in particular our eyes to read the lips.
2. Context: for instance, when we’re in a classroom we are more likely to hear vocabulary related to the classroom.
3. References: for example when someone points out an object and a bus passes by, we can decipher the world the person was saying.
4. Social queues: help us understand the meaning of the words, for instance if a dad smiles to his baby when he’s telling him something he did was good.

Opportunity for research

In a try to reproduce the efficiency of human speech recognition, there is a lot of research done in the following areas:

• Multi-modal learning: use new ways to recognize speech such as computer vision systems through lip reading.
• Untrained dataset: instead of having annotated speech, use incomplete annotations to train the system (words with no particular order).
• Semi-supervised learning: use a small data set very well annotated and a lot of speech without annotations.
• Transferred learning in speaker recognition: use a model trained in another language and generalize it to another one.
• Computer vision systems and convolutional neural networks: try to remove the need of speech features to be able to train speech recognition systems directly from the waveform up to the phonetic transcription.

So don’t worry, we still have a small advance over machines ;)

If you want to learn more about speech recognition I recommend you to read Speech and Language Processing and Computational Analysis of Sound Scenes and Events. A special thanks to Facebook Artificial Intelligence Research Lab and Neil Zeghidour for his course on speech recognition.

Did you learn something? Click on the 👏 to thank the author and help other non-tech people understand speech recognition technology.

Have you ever wondered how Amazon Echo, Siri or Google Home work? was originally published in Coinmonks on Medium, where people are continuing the conversation by highlighting and responding to this story.

Hormis le dépaysement linguistique, climatique et visuel (je n’ai jamais vu un endroit aussi gris que la banlieue de Troyes) pour lequel nous étions mentalement préparés, nous n’aurions jamais pu imaginer que les démarches administratives auxquelles nous allions faire face allaient ressembler à vrai un parcours du combatant.

Tout au long de l’année et demie où j’ai vécu en France, j’ai été témoin des nombreux aller-retour à la préfecture, des menaces de me déporter toute seule (mes parents ayant le droit de rester en France mais pas moi…allez comprendre comment la dame du guichet qui était aussi peut-être elle-même mère a pu prononcer ses mots sans laisser transparaître la moindre trace de compassion), du débarquement de la police chez nous qui venait mesurer la taille des placards dans notre appartement parce que “nous savons que vous les étrangers, vous êtes habitués à vivre entassés, mais c’est pas le cas ici”. ENTASSÉS. Si le policier avait su qu’au Mexique je vivais dans une maison, modeste certes, mais qui faisait 3 fois la taille de notre appartement en France…

Aux nombreuses demandes de support effectués auprès de son directeur de thèse, ma mère a eu une réponse qui résumait très bien l’accueil qui nous avait été réservé en France : “Nous ne pouvons rien faire pour toi. De toute façon tu n’auras jamais raison ici pour la simple et bonne raison que tu es étrangère”. Comme si cela n’avait pas déjà été assez difficile pour ma mère d’abandonner tout ce qu’elle avait connu dans sa vie pour aller étudier à l’autre bout de l’Atlantique, quitter sa famille et ses amis, voir diviser par 3 son pouvoir d’achat…on devait également porter le fardeau d’être des étrangers !

Face à l’absence de support de l’université où ma mère faisait son doctorat, j‘ai fini par rentrer au Mexique avec mon père. Ma mère a sombré dans une dépression qui l’a faite obtenir son doctorat en 5 ans au lieu de 3 ans comme c’était initialement prévu.

L’humiliation, la xénophobie, l’indifférence.

Trois mots qui m’ont marqué dès le plus jeune âge. Celle-ci est mon histoire mais c’est aussi l’histoire de milliers d’étrangers, en France et ailleurs, qui pour des motifs professionnels, familiaux ou humanitaires quittent leur pays de façon provisoire ou définitive.

A 18 ans j’ai décidé moi aussi de venir étudier en France. L’accueil que j’ai reçu a été moins malheureux que celui qui nous avait été réservé 13 ans auparavant. Mais les démarches administratives auprès de la préfecture n’ont, en 7 ans de vie en France, jamais arrêté d’être une source d’angoisse, peur et confusion.

Plus jamais ça.

Angoisse, peur et confusion mais aussi indignation, rage et tristesse quand je constate les conditions dans lesquelles les étrangers (dont un membre de notre équipe) sont reçus depuis plusieurs mois à la préfecture de Nanterre, devant passer la nuit dehors à faire la queue “Place des Droits de l’Homme” à partir de 22h dans l’espoir d’obtenir un ticket le lendemain. Des conditions déplorables et un traitement inhumain qui n’est pas à la hauteur des valeurs que porte la France ni de l’amour et la gratitude que nous éprouvons pour cette France qui est aussi la nôtre.

Macron l’a imaginé…

Lors de sa conférence au salon VivaTech le 15 juin dernier, Macron déclarait vouloir faire de la France une “startup nation" et annonçait le lancement du French Tech Visa, censé faciliter les procédures administratives pour attirer entrepreneurs, investisseurs et employés de startup en France.
Dans ses promesses de campagne, M. Le Président a également pris l’engagement de réduire les délais d’examen des demandes d’asile de treize à six mois environ.

La théorie c’est très bien, la mise en place c’est encore mieux.

Nous partons du constat qu’aujourd’hui quand un étranger fait une première demande de titre de séjour, un renouvellement ou un changement de statut, il est confronté à 3 problèmes principaux :

1. 😳 Processus confus 😖: les informations officielles sont souvent incomplètes, obsolètes et hétérogènes (parfois au sein de la même préfecture en fonction de l’agent), ce qui crée de la confusion. De plus, une veille constante sur les mises à jour de la loi, des interprétations et des applications de celle-ci ou des exigences propres à chaque préfecture est nécessaire, ce qui fait perdre beaucoup de temps et nous met dans une situation d’insécurité juridique.
2. ❓ Manque de suivi ❓ : aujourd’hui c’est très compliqué voire impossible de suivre le statut de son dossier ainsi que de joindre facilement des personnes au sein des administrations concernées qui répondent à nos questions de façon rapide et claire. Demander des retours d’expérience à ses amis, sur des forums ou des blogs est souvent frustrant car nous ne trouvons pas une réponse claire et adaptée à notre situation. De plus, les délais sont très longs tout au long du processus (prise de RDV, traitement de dossier) et nous mettent souvent en situation d’illégalité, portant atteinte à nos droits les plus basiques.
3. ⚖ L’accès à des conseils juridiques est une affaire de riches 💰 : faire valoir nos droits en matière d’immigration et se faire accompagner par des experts en la matière est très compliqué. Faire appel à des associations d’aide juridique est très frustrant car elles sont débordées et font un traitement à la chaîne pour essayer de répondre à une forte demande. La seule solution qui reste est de faire appel à un avocat, qui coûte très cher et dont l’expérience client est loin d’être terrible.

…nous l’avons fait.

My Visa Angel est le premier service qui accompagne les étrangers dans les démarches de titre de séjour en France. My Visa Angel :

1. ⚖ T’informe gratuitement : via la plateforme de chat Facebook Messenger, notre robot Manu se charge de centraliser, mettre à jour les informations et donner des conseils sur le titre de séjour le plus adapté à ta situation personnelle et professionnelle.
2. 😎 Te met en contact avec un expert : qui t’accompagne dans la constitution du dossier, t’aide à remplir les formulaires, répond à tes questions et te soutient jusqu’à l’obtention de ton titre de séjour.

Toi aussi tu as la phobie administrative ?

Viens nous voir ! 😇

Et si tu veux nous envoyer des mots d’amour, nous donner du feed-back ou tout simplement nous rencontrer, écris-nous à hello@myvisaangel.com 🚀

(from a startup employee point of view)

Congrats, you’re making at least one of the following mistakes. Chances are you are not really aware of it, but you could seriously be undermining your team. In a startup, the most important thing is not your customers, it’s your employees. This may sound weird or counterintuitive to some of you. But think about it this way: if you want to grow, you are certainly not going to achieve this on your own. Take care of your employees who in turn are going to take care of your customers. It’s time to get slapped in the face by the truth!

1. Blurry vision & values

Employees join a startup because they feel in line with its vision and values. Vision is like your northern shining star. It is the reason why you and your employees come to work every morning. Values will define your culture, the people you should hire, promote and fire and it is specially useful when you or your employees have hard choices to make. As a CEO, it is your responsibility to clearly formulate and explain your vision and values. You should constantly ask yourself how you can make it clear why what your employees do matters. If you fail to provide a strong vision and embody your values or you suck at communicating them, your employees will feel lost. You will have a hard time to focus and scale. You and your team will have the feeling that you are working your ass off for nothing. And there is nothing more demotivating than not being able to see the light at the end of the road! Besides that, there are high chances that this will be perceived from the outside. You are probably going to have a hard time recruiting and your clients will have a hard time to understand who you are, where you are heading and why they should follow you. After all, vision and values are the foundations of any startup.

2. Poor leadership

Startups need leaders, not managers. Leaders inspire the people they work with. Leaders don’t treat their employees as little soldiers who merely execute orders and then report back results. They neither exploit their teams, nor put them at a distance. As a leader, you should show the example and be ready to make the same sacrifices you ask your employees to do, if not more. Lead by example, not fear. Work hand in hand with your team and nurture a collaborative approach. Leaders also don’t try to control everything, because they know how important it is to promote an environment where employees feel safe and confident enough to be able to decide for themselves. See your team as a football team. If you want to build a kick-ass team, more than micro managing and supervising what each one of your team players is doing, your priority should be to identify their strengths and understand what motivates each one of them. Emotionally connect with them at a personal level and understand their needs. Let them know you trust them, learn to delegate and give them big responsibilities. You will be amazed by how much they can do. If you are not fully convinced, you should get familiar with the concept of “pygmalion effect”. As a startup CEO, your responsibility is not to empower them, because everything people need to be motivated, they already have it inside them. Your responsibility is to focus on creating the best environment possible to allow your employees’ intrinsic motivation to come to light. The key to your success depends on your ability to be a leader and not a manager.

3. Not giving feedback

Your employees need feedback as much as you need feedback from your customers. Not giving enough feedback might be interpreted as a lack of interest and recognition for your employees’ work. As a startup CEO, your main responsibility is to help your team grow, develop and attain their goals. Feedback, whether done formally or informally around a casual beer, is part of your team’s development process. Ignoring this could compromise your team’s evolution and success. No matter if your employees are doing it right or wrong, you should take the time to point out what they are doing correctly and work out with them to learn from their mistakes. Please celebrate success as much as you focus on failures (French people specially suck at doing this…). By the way, have you ever asked yourself if your team has the feeling their work is not recognized? Make yourself a favor. Start by saying thank you. And if you say it, mean it. Of course you pay them for the work they do. But hey, a simple “thank you” or “congratulations” doesn’t cost anything and it may mean the world to them.

4. Failing to listen

Listening is often an overlooked skill. As a startup CEO you are probably overbooked. You don’t have the time — or you probably don’t want to take the time. Are you the kind of CEO that is on the phone or doing something else while in a meeting? Are you the kind that cancels meetings at the last moment without apologizing or giving a valid reason? Let me tell you something: your time is as valuable as theirs. Sitting at a meeting without actually being fully engaged is one of the best ways to devalue your employees’ efforts, kill engagement and promote a blasé attitude. Being a good listener is hard. Indeed, it takes a lot of humility, open-mindedness and empathy to listen. Listening entails being able to accept to hear things that are hard to acknowledge. It means being able to ask the right questions, and show gratitude for the time and sincerity of the opinions shared. Failing to do so will break communication between you and your team. They will perceive you as the know-it-all CEO, unable to accept criticism and challenge himself. In a startup environment where everything is permanently changing, this attitude could kill you.

5. Over promising

Don’t make promises you won’t be sure you will be able to keep. Actually, don’t make promises at all. By definition a startups is fragile and everything is uncertain. Making promises in such an unstable environment is counterproductive. As a startup CEO and leader, your word is taken as a fact. However, promises in a startup context mostly depend on external factors that you cannot control. Best case scenario, you will be able to keep your promises and your team will take it as granted. Worst case scenario, you will not be able to keep the promises and they will feel betrayed. People will feel like they were fooled and will start to doubt about all the promises you made them, which will lead to disappointment. Expect trust to take a pummeling.

6. Negativity and lack of recognition

You see the worst of everything. You assume your people always make bad decisions, ask silly questions or take poor initiatives. Have you asked yourself if maybe you recruited the wrong person, didn’t set up clearly your expectations and their objectives? Of course not, it is never your fault! If you fail to reward initiatives (good or bad), don’t be surprised if your team starts doing the minimum required and stops taking risks. You will gradually undermine their morale, kill motivation and boost disengagement. If when someone makes a mistake, a deadline is missed, or a plan goes wrong, you look for scapegoats and lay the blame on others, don’t be surprised if you have a high turnover. No one wants to work in such a poisonous environment. Negativity is the enemy of all working relationships. You will earn a reputation for intimidating your teammates and evading your responsibility. So please start by recognizing your employees’ efforts, no matter how well they are performing. Create a positive working environment and promote a culture where people are grateful with each other. Let your team know you value their work and teach others to do so as well.

7. Indecisiveness

In a startup time is money. As a startup CEO, you certainly have a million things to do, and failing to act quickly could be the difference between life and death. If you are paralyzed by the inability to find solution or delay in the hope a magical solution will appear, you are dead. You keep looking for excuses — you are too busy, it is not the right time, you lack information, etc. — and you keep finding them. You are certainly afraid to make the wrong decision but your inability to make up your mind will send the wrong message to your team. And your indecisiveness almost always comes hand in hand with your lack of consistency. One day you decide one thing, the next day you decide the contrary. Your team feels lost and trapped. People will start talking behind your back. You will be perceived as someone weak, who doesn’t know where he wants to go. And no one wants to be led by someone who seems to be lost. So please stop losing your time and money. The faster your make a decision, the sooner you will know whether or not it was the right one. Most of the time you will fail. It doesn’t matter. You will learn, adapt and correct.

Remember what Thomas A. Edison said, “I have not failed. I’ve just found 10000 ways that won’t work.”

8. Bad communication

Keeping an important amount of information in the hands of some privileged employees is wrong. It leads to distrust and defiance. Transparency may not be your thing but it is key in order to keep speculations and gossips away. Avoid playing Chinese telephone. Give you a chance to explain the motivations behind your decisions before your employees learn this from someone else. What’s worse than one of your sales team member learning from one of his colleague that you have decided to put in place a new remuneration system for the sales team? “Why? When? How come I’m not informed? Am I not important enough to be taken into consideration?” Learn to involve people by keeping them well informed. Trust them enough to give them important information that directly affects their work. If you don’t trust them, what are they doing here? Of course you cannot control every single exchange that takes place between your employees but you have at least the responsibility to clearly establish the boundaries of what kind of information is acceptable to share and what is not.

9. Inability to clearly define your employees’ scorecard

If you are not explicitly defining and documenting the scorecard of each role in your startup, there are high chances you are dealing with misunderstandings, project overlaps and ultimately jalousies within your team members. In Who: The A Method for Hiring, defining a scorecard for each position in your company is the blueprint for success. Scorecards define the individual objectives of each employee for the purpose of the startup’s common strategy. It will allow you and your team to clearly outline the job mission, the expectations and the skills needed to do this. Having clear roles and scopes will keep politics away (and trust me, you really want to keep politics away from your startup as long as possible!). It will help your employees to feel ownership of their work and responsibilities. It will avoid your startup losing time because two people are working on the same project and will ultimately boost collaboration since they will exactly know who they can rely on for certain type of projects.

10. Hiring mistakes

Hire fast, fire faster. Does that sound familiar to you? All CEOs have done a hiring mistake. One of the most common ones is hiring by skill, not by team and culture fit. On paper the candidate looked great; top education, solid track record, good technical skills. But once it’s time to get to work, you realize how expensive that hire really was. Expensive in time and effort, for you and your entire team. Your team needs to evolve in a culture of mutual blind trust, where they can rely on each other. They don’t have to worry about ego battles, free riders, gossips, politics and so on. Remember that a negative attitude can spread very quickly among your team members. Don’t underestimate the damage of a wrong hire. Even the six-month intern you decided to hire must be totally aligned with the startup’s culture and values. Your responsibility as a CEO is to quickly identify and fire that toxic employee if you don’t want to hinder your team’s dynamics. The longer you will wait to fire them, the higher the cost, and one of the biggest costs is regret.

Like this post? Share it with your friends and fellow entrepreneurs.