Down to 4.9 Stars for No Reason?

[dereck_s]

It's a little scary to think about how crucial systems are being tested a week away from this system being rolled out as our new system. I still don't even know what my actual score is because the bug cancellation issue still exists. This bug will cause me to lose all of my levels and the progress I've made over a decade and it's still not fixed a week away from launch? Are they even trying to fix it or are they testing this new rating system both of which don't seem to be in a state that could be released.

[vickiespencer]

6 hours ago, uptwee said:

My rate went from 5 to 4.9. @vickiespencer If you can kindly screenshot the 5 metrics for me, I would very much appreciate it. I can only see 3 from my end. 

Sorry, my posting was limited for 24 hours. 

[smashradio]

On 3/7/2024 at 5:54 PM, emmasuebowtique said:

Exactly! If my rating went from a 5 to 4.9, then why is my success score a 4... Yet I am being told my success score is no longer at level 2 level because of private reviews. So it docks my reviews by .1 however I dropped to a level 0... All I can do is laugh. I legit got a fiverrs choice award order yesterday. Guess a success score of 4 is deserving of that LOL! 

Yeah. It's almost like they just... implemented this without knowing how it would go. But nooo, they wouldn't. Would they? 🤣

 

[graveyardshiftx]

Welcome to Fiverr 2.0, where 5 means 3, 3 means 5, yes means no and "I love my new logo" means "You suck at your job".

Basically, this new rating system is shrouded in mystery and it's gonna stay that way because it's in Fiverr's interest to do so. Now they can just blurt out "Because algorythm!" whenever something like this happens and since we're not privvy to that information, all we can do is... well nothing. We can do nothing.

I'm beginning to suspect that this "algorythm" is in fact 1 part AI, 1 part private feedback and 3 parts unicorn queefs because it makes absolutely no sense.

[desmond_aubery]

For a New Seller, this is beyond frightening... 🥸

[zeus777]

So aside from my rates going down to 4.9 which aint all that bad, my response rate has gone down to 94%.

I'm sure it was 95 or 96 a few days ago...? I guess I'm not the only one.  I know its not that much of a big deal but still mildly annoying I guess.

[emmasuebowtique]

2 hours ago, desmond_aubery said:

For a New Seller, this is beyond frightening... 🥸

Run! 🫠

[emmasuebowtique]

2 hours ago, zeus777 said:

So aside from my rates going down to 4.9 which aint all that bad, my response rate has gone down to 94%.

I'm sure it was 95 or 96 a few days ago...? I guess I'm not the only one.  I know its not that much of a big deal but still mildly annoying I guess.

My response rate has been all over the place too, it will be 98% then later that day it will be 96%, then next day go back. 

[newsmike]

Fiverr showed their cards when they stated "4.9 ratings are more trustworthy than 5 star ratings," so their response has been to manipulate the data to arrive at that arbitrary number, regardless of how they get there. The goal seems to be to present buyers with manufactured 4.9's where 5's existed, you know, to appear more trustworthy. 

If simply "appearing" trustworthy as opposed to actually "being" trustworthy is Fiverr's goal, then they should rethink their position on using VPN to fake being a US seller as well. After all, it appears "more trustworthy", so why bother with that whole "truth" concept? 

The practice of simply aggregating the buyer's ratings and presenting them honestly, without weighting and manipulation is being sacrificed in order to fake it and present what Fiverr believes is a "more trustworthy" appearing metric based on some focus group.

They are no longer collecting data and presenting the results, they are instead, deciding what result they want, and then torturing the data collection and analysis to arrive at their predetermined conclusion. 

They are manipulating the metrics for marketing purposes, and putting the worst AI on the planet in charge of it. Google, X and others have interesting experiments with artificial intelligence running, but Fiverr is the only platform I'm aware of that is developing "artificial stupidity."

Edited March 9 by newsmike

[newsmike]

21 hours ago, smashradio said:

Yeah. It's almost like they just... implemented this without knowing how it would go. But nooo, they wouldn't. Would they? 🤣

 

[maitasun]

23 minutes ago, newsmike said:

They are manipulating the metrics for marketing purposes, and putting the worst AI on the planet in charge of it. Google, X and others have interesting experiments with artificial intelligence running, but Fiverr is the only platform I'm aware of that is developing "artificial stupidity."

Two different "me"s... I guess Fiverr's almighty AI found my long-lost twin I didn't know about. 🥳 🎊🎉

Gig A:

Gig B

 

[newsmike]

5 minutes ago, maitasun said:

Two different "me"s... I guess Fiverr's almighty AI found my long-lost twin I didn't know about. 🥳 🎊🎉

Gig A:

Gig B

 

But that's all within acceptable parameters.  😆 At least Google had the honesty to say, "we're pausing our AI" till we learn how to make it work.  Fiverr won't do that, they are all in no matter the damage, because admitting otherwise would not 'appear trustworthy."

 

Edited March 9 by newsmike

[priyank_mod]

41 minutes ago, newsmike said:

The goal seems to be to present buyers with manufactured 4.9's where 5's existed, you know, to appear more trustworthy. 

Classic corporate scenario where strategic decisions are often made on the basis of outsourced Market Research and dubious FGD's (Focused Group Discussions).🤦‍♂️

Someone in a fancy suit walks in and presents uncharacteristic+unheard findings as results and then entire system starts working with fervour to align itself with those findings!! 🫠

[maitasun]

11 minutes ago, newsmike said:

But that's all within acceptable parameters.  😆 At least Google had the honesty to say, "we're pausing our AI" till we learn how to make it work.  Fiverr won't do that, they are all in no matter the damage, because admitting otherwise would not 'appear trustworthy."

The problem here is how much more that stu... adorable AI wants me to communicate with my buyers. Until they get sick and tired of me and kick my 🍑 for pestering them??? 😡

[mandyzines]

53 minutes ago, newsmike said:

They are no longer collecting data and presenting the results, they are instead, deciding what result they want, and then torturing the data collection and analysis to arrive at their predetermined conclusion. 

It’s very dangerous, Mike, because a lot of what you’re seeing as attacks on Fiverr quite frankly are attacks on science, because all of the things that Fiverr has spoken about consistently from the very beginning, have been fundamentally based on science. 

[newsmike]

10 minutes ago, mandyzines said:

It’s very dangerous, Mike, because a lot of what you’re seeing as attacks on Fiverr quite frankly are attacks on science, because all of the things that Fiverr has spoken about consistently from the very beginning, have been fundamentally based on science. 

 

We must believe anything labelled as "science," no questions. 

 

 

https://en.wikipedia.org/wiki/Arthur_Godfrey#Death

 

Edited March 9 by newsmike

[vickieito]

Oh, this is interesting. I can now see the ratings breakdown for all of my gigs. My worst gig (Writing Services), has an average rating of 4.7 stars, with "Quality of Delivery" at 3.5 stars:

Quote

Public Reviews (all 5-stars) didn't give me any insight into the 3.5 score, but the Success Score gave some insight:

Quote

The 2-3 Key Areas that Fiverr shows for each gig seems to be the lowest-scored Key Areas (based on the 6 used by Fiverr). I have zero order cancellations, so the reasons would fall on "Effective Communication" and "Client Satisfaction" as the top reasons for my low "Quality of Delivery."

I'm guessing that since this Writing Gig often requires that I provide constructive criticism of my clients' writing (especially if proofreading or editing), there may have been clients unhappy with my critical feedback or AI might have picked up possible argumentative language that I've had in the order chats with ESL buyers who wanted me to correct their grammar in the order chats (several had learning styles of "arguing it out until the concept is understood"). I was pretty blunt in those conversations if the buyers were wrong.

[emmaki]

32 minutes ago, vickieito said:

I was pretty blunt in those conversations if the buyers were wrong.

I had to stop telling buyers when they were wrong. I almost lost pro because of it. Even though one of those buyers later came back and said I was right. Now I just agree and keep my right(eous)ness to myself. If the customer wants to be right, I am not sacrificing my metrics for them, even if it could cost them thousands of dollars - that game works both ways. Well done, Fiverr! I now deliver worse work for better reviews! 

It doesn't make any sense how all 5-star gigs can result in a 3.5 rating. Anyway, "quality of delivery" is just Fiverr's botched rewrite of UW's "quality of work", which as I'm sure we all know is a very different concept.

Still, who cares? Toss everyone on the bonfire! 

[vickieito]

10 minutes ago, emmaki said:

I had to stop telling buyers when they were wrong.

But the order requirements were basically, "I will drop my writing into the order chats. You will tell me what's wrong." This was more for writing advice/feedback session. So I would say what's wrong and show my corrections. Then we would "argue" back and forth until the buyer understood why. It worked for me and my buyers. These buyers don't want to be right and they didn't want me to let anything slip. But as you said, maybe AI isn't looking at these chats to determine the score. Most likely it was the private reviews that brought this score to 3.5.

Most of these buyers were from my newbie seller days and I no longer work with those types of buyers. So the score can only get better from here (as long as I work on communication and improving my client satisfaction). 🤞

Edited March 9 by vickieito

[mandyzines]

41 minutes ago, vickieito said:

Oh, this is interesting. I can now see the ratings breakdown for all of my gigs. My worst gig (Writing Services), has an average rating of 4.7 stars, with "Quality of Delivery" at 3.5 stars:

Public Reviews (all 5-stars) didn't give me any insight into the 3.5 score, but the Success Score gave some insight:

The 2-3 Key Areas that Fiverr shows for each gig seems to be the lowest-scored Key Areas (based on the 6 used by Fiverr). I have zero order cancellations, so the reasons would fall on "Effective Communication" and "Client Satisfaction" as the top reasons for my low "Quality of Delivery."

I'm guessing that since this Writing Gig often requires that I provide constructive criticism of my clients' writing (especially if proofreading or editing), there may have been clients unhappy with my critical feedback or AI might have picked up possible argumentative language that I've had in the order chats with ESL buyers who wanted me to correct their grammar in the order chats (several had learning styles of "arguing it out until the concept is understood"). I was pretty blunt in those conversations if the buyers were wrong.

That's interesting. We both have a 9 on the gig. I too have had no cancelations, but have a "positive impact" rather than a "strong..." like you do.

My client satisfaction's lower than yours, but quality of delivery is the only thing that's different in the rating breakdown.

None of this makes sense.

 

 

[maitasun]

55 minutes ago, vickieito said:

The 2-3 Key Areas that Fiverr shows for each gig seems to be the lowest-scored Key Areas (based on the 6 used by Fiverr). I have zero order cancellations,

Sorry to disagree. I also have zero order cancellations and, yet, all my gigs show "positive impact" in that area, even if the gig has a SC of 10. Fiverr is just throwing things at the wall to see what sticks...

Edited March 9 by maitasun

[emmaki]

9 minutes ago, vickieito said:

AI isn't looking at these chats to determine the score

Yeah, that is different (your gig). Although it would be disastrous if the AI was looking at the chats to determine scoring based on sentiment, as the AI really isn't smart on that front and can't tell if the badwords are being said seriously or as part of an exercise. Then again, isn't that how it is (partially) looking at revisions and gig extensions to determine whether it's OK or not? It all feels a bit big brother and unnecessarily, even pointlessly intrusive (I'm fine with Fiverr having access to all convos etc, it's more 💩 AI systems that get my goat)

 

6 minutes ago, mandyzines said:

None of this makes sense.

It makes perfect sense if you think the worst of Fiverr 🙂 This has to be the worst product launch they've ever done in terms of bugginess/intransparency/brokenness/words I'm not allowed to say.

Ah well, only a few more days until we can all start melting down all over again.  

[vickieito]

7 minutes ago, maitasun said:

Fiverr is just throwing things to the wall to see what sticks...

I thought they just turned on an AI switch and left it on. 😊

We can see a little more data, but we can't see our scores in all six areas (only 2-3 key areas). For the others, we don't know how those key areas are rated by the AI (i.e., strong negative impact, negative impact, room for growth, positive impact, or strong positive impact). And even further, we don't know how AI breaks down each of those ratings.

There are so many factors that sellers (and Fiverr) don't know are influencing these scores.

[domenikbrenner]

20 minutes ago, vickieito said:

But the order requirements were basically, "I will drop my writing into the order chats. You will tell me what's wrong." This was more for writing advice/feedback session. So I would say what's wrong and show my corrections. Then we would "argue" back and forth until the buyer understood why. It worked for me and my buyers. These buyers don't want to be right and they didn't want me to let anything slip. But as you said, maybe AI isn't looking at these chats to determine the score. Most likely it was the private reviews that brought this score to 3.5.

Most of these buyers were from my newbie seller days and I no longer work with those types of buyers. So the score can only get better from here (as long as I work on communication and improving my client satisfaction). 🤞

AI looking over chats and analysing them for our seller score is a bad idea.. Even if a human looks over it, it doesn't really tell how the client was picked up and made understood or how they felt about it. Sometimes a bit of tough love is necessary and even expected, at least I can imagine if someone hires Vickie for her literary criticism, which is literally that: My text is bad and I want it to become better, so you need to tell me what's wrong and fix it for me.

[vickieito]

5 minutes ago, domenikbrenner said:

AI looking over chats and analysing them for our seller score is a bad idea..

The only inkling that I have that AI might be doing this is in Fiverr's "Success score, explained" article where it keeps referring to "tone" when advising sellers how to communicate with buyers (e.g., polite and friendly). But let's hope that Fiverr isn't that reliant on AI...

Grammarly tells you immediately in the form of emojis how your messages sound. I was wondering if they used the same technology to "guess" how a conversation is going.

Edited March 9 by vickieito